Application of Swat

Stockholm Environment Institute, Technical Report - 2012

Application of SWAT and a Groundwater Model for Impact Assessment of Agricultural Water Management Interventions in Jaldhaka Watershed: Data and Set Up of Models Devaraj de Condappa, Jennie Barron, Sat Kumar Tomer and Sekhar Muddu

Application of SWAT and a Groundwater Model for Impact Assessment of Agricultural Water Management Interventions in Jaldhaka Watershed: Data and Set Up of Models Devaraj de Condappa, Jennie Barron, Sat Kumar Tomer and Sekhar Muddu

Stockholm Environment Institute Kräftriket 2B SE 106 91 Stockholm Sweden Tel: +46 8 674 7070 Fax: +46 8 674 7020 Web: www.sei-international.org Head of Communications: Robert Watt Publications Manager: Erik Willis Layout: Richard Clay Cover Photo:

This publication may be reproduced in whole or in part and in any form for educational or non-profit purposes, without special permission from the copyright holder(s) provided acknowledgement of the source is made. No use of this publication may be made for resale or other commercial purpose, without the written permission of the copyright holder(s). Copyright © March 2012 by Stockholm Environment Institute

ABSTRACT This study contributes to the understanding of potential for Agricultural Water Management (AWM) interventions in the watershed of Jaldhaka river, a tributary of the Brahmaputra river, located in Bhutan, India and Bangladesh. An application of the Soil Water Assessment Tool (SWAT) and of a simple lumped groundwater model was developed for the Jaldhaka watershed. The rst stage of this work was to collect a large dataset to characterise the natural and agricul tural contexts of the Jaldhaka watershed. The watershed has a contrasting topography, with mountains upstream and large plains downstream. It experiences high rainfall with a monsoonal pattern and an average of 3,300 mm/year. The river ow is seasonal, with a sustained ow during the dry season, high ows during the monsoon and recurrent ood events. The soils are sandy loam (upstream) to silty loam (downstream), with little permeability. The aquifers in the region are alluvial and the groundwater levels in the watershed are shallow and stable. This study contributed to the development of a precise landuse map which identies the natural vegetation, the water bodies, the settlements / towns, the tea plantations and the different cropping sequences in the agricultural land. Agricultural statistics were gathered at administrative levels for cropping sequences and crop yields. The irrigation in the watershed is predominantly from groundwater, with diesel pumps, to irrigate rice during summer and potatoes during winter. SWAT and the groundwater model were adjusted in an interactive manner: SWAT was calibrated against the observed streamows while the groundwater model was calibrated against the observed groundwater levels and the interaction aimed at the convergence of both models. The performance was satisfactory for modelling the watershed on an average monthly basis. However, the model set-up failed to reproduce adequately the crop yields. This paper ends with a discussion of the modelling setup and data collection for agro-hydrological modelling. This set-up was applied in an accompanying research report to study the current state of the hydrology in the Jaldhaka watershed and the impacts of two types of AWM scenarios.

CONTENTS Abstract

iii

List of abbreviations

viii

1 Introduction

1

2 Introduction to the modelling softwares

3

2.1 Soil and Water Assessment Tool (SWAT) 2.2 Groundwater model 3 Biophysical data of the Jaldhaka watershed 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8

Digital Elevation Model Streamflow data Climate data Soils Groundwater data Land-use Agricultural Irrigation

4 Modelling set up 4.1 Initial setting of SWAT 4.2 Calibration of the groundwater model and SWAT

3 3 6 6 6 12 16 20 23 29 37 40 40 45

5 Discussion 5.1 … on the input dataset 5.2 … on the model set up

56 56 57

6 Conclusion

58

Acknowledgements

60

Annex

62

References

70

LIST OF FIGURES Figure 1: Location of the Jaldhaka / Dharla river watershed (in purple). The delineation of the Jaldhaka / Dharla watershed were generated in this work. Figure 2: Scheme of the modelling Figure 3: Digital Elevation Model from the Shuttle Radar Topography Mission and locations where climatic and streamflow data was available Figure 5: Topographic profile of the transect defined in Figures 3 and 4 Figure 4: Slope derived from the DEM, the two local meteorological and streamflow gauge stations Figure 7: Available time-series for streamflows measured at Taluk-Simulbari and Kurigram stations, unfiltered (left) and average monthly streamflow, filtered (right); the vertical error bars indicate the statistical standard deviation of daily streamflows Figure 8: Zoom around Kurigram on Google Earth where are visible the infrastructures for water diversion as well as the neighbouring rivers, in particular the massive Brahmaputra. Figure 9: Representative average rainfall for the Jaldhaka watershed, as calculated by SWAT, and average streamflow at Kurigram (period 1998 – 2008) Figure 10: Rainfall at Jalpaiguri and Cooch Behar stations (period 1988 - 2008). Top: daily rainfall. Middle: annual rainfall. Bottom: average monthly rainfall, the vertical error bars in red indicate the statistical standard deviation of daily rainfall (in mm/day) Figure 11: Average climatic data at Jalpaiguri and Cooch Behar stations (period 1988 - 2008). Top: temperature. Middle: wind. Bottom: humidity. The vertical error bars indicate the statistical standard deviation of daily data Figure 12: Distribution of the average annual rainfall in the sub-watersheds, as represented in SWAT (period 1998 - 2008) Figure 13: The georeferenced soil map in the region of the Jaldhaka watershed Figure 14: The Harmonised World Soil Database and its soil units in the region of the Jaldhaka watershed. Figure 15: Plot in soil textural triangle of the United State Department of Agriculture Figure 16: Location of the observation wells for groundwater level measurement. CGWB stands for Central Ground Water Board and SWID for State Water Figure 17: Measured groundwater levels in the Jaldhaka watershed. In pale: level of different wells. In black: average of all the wells Figure 18: Typical groundwater levels in the Jaldhaka watershed. The wells are located on Figure 16. The vertical error bars indicate the statistical standard deviation Figure 19: Interpolation of average piezometric levels observed by the State Water Investigation Directorate (SWID) (period 1994 - 2009). Figure 20: Satellite images acquired for high resolution landuse mapping. Note the demarcation between the north and south view Figure 21: Location of the groundtruthing sites visited in April 2010 and draft unsupervised classification of the landuse. Right: zoom on the transect (note on this view the discrepancy Figure 22: Calendar of the main cropping sequences in the Jaldhaka watershed Figure 23: High resolution (10 m) landuse map of the Jaldhaka watershed (year 2008). Figure 24: Photos of the spots identified on the landuse map (Figure 23) Figure 25: Modified version of the landuse map (Figure 23, year 2008) entered in SWAT (90 m resolution) vi

1 3 7 7 7

9

10 11

13

15 16 17 17 19 21 22 22 23 24 25 26 27 28 29

Figure 26: Area of the major crops in administrative blocks containing the Jaldhaka watershed Figure 27: Yield of the major crops in administrative blocks containing the Jaldhaka watershed. Mind the different vertical scale Figure 29: Average monthly reference evapotranspiration calculated from difference sources Figure 30: Calibration with respect to the actual evapotranspiration ETa. Monthly value of the different landuse vegetation categories (average over the calibration

32 33 46

period, 1998 – 2008). 47 Figure 31: Piezometric levels simulated at a monthly time-step by the groundwater model vs. observations 48 Figure 32: Calibration with respect to the recharge of the shallow aquifer (GW_RCHG), average for the Jaldhaka watershed over the calibration period (1998 – 2008) 49 Figure 33: Calibration with respect to the shallow groundwater baseflow (GW_Q), average for the Jaldhaka watershed over the calibration period (1998 – 2008) 50 Figure 34: Streamflow simulated (FLOW_OUT) at Kurigram in the initial run over the calibration period (1998 – 2008) 51 Figure 35: Streamflow simulated (FLOW_OUT) in the final calibration (calibration run n°100) over the calibration period (1998 – 2008). 53 Figure A.1: Example of the groundtruthing form (site GT 35) filled by the field assistants 69

LIST OF TABLES Table 1: Table 2: Table 3:

Table 4: Table 5:

Table 6: Table 7: Table 8: Table 9:

Topographic regions of the Jaldhaka watershed Available number of measurements at Taluk-Simulbari and Kurigram stations. Source of data: Bangladesh Water Development Board. Available climatic time-series and gaps in the datasets. RMC stands for Regional Meteorological Centre (Kolkata) and NCC for National Climate Centre. Annual rainfall at Jalpaiguri and Cooch Behar stations (period 1988 - 2008) Available measured groundwater levels in the Indian part of the watershed. CGWB stands for Central Ground Water Board and SWID for State Water Investigation Directorate. Distribution of the landuse categories (Figure 23) within the Jaldhaka watershed. Distribution of the landuse categories entered in SWAT (Figure 25). Available agricultural statistics. Average yields in the administrative blocks containing the Jaldhaka watershed, period 1998 – 2008. In bracket the average dry yield of rice for period 2004 to 2008. Source of data: Bureau of Applied Economics and

7 8

12 14

20 28 30 30

Statistics and Directorate of Agriculture. 31 Table 10: Typical cropping sequences and associated irrigation schedules in the Jaldhaka watershed. 34 Table 11: Indicative distribution per sub-watershed of the cropping sequences within the landuse units AAAJ and AWJJ. Derived with data from the Bureau of Applied Economics and Statistics and the Directorate of Agriculture. 36 Table 12: Estimated irrigation per crop. Sources of data: groundwater pumping duration from Mukherji (2007) and diesel pump discharge from TERI (2007) 38

vii

Table 13: Indicative areas irrigated from surface sources in each sub-watershed, derived from DPDWB (2005) for year 2004/5 Table 14: HRUs generation stages. Table 15: Landuse distribution considered in SWAT after pre-processing by ArcSWAT, with respect to the discretisation in HRUs, and management operations for each category. Table 16: Estimation of irrigation areas and amount for 2008, with respect to the discretisation in HRUs. Table 17: Initial values for undetermined SWAT’s parameters. Table 18: Average annual reference evapotraspiration calculated from different sources. Table 19: Calibration with respect to the evapotranspiration ETa. Annual values of the ratio ETa / ET0 for the different landuse vegetation categories (average over the calibration period, 1998 – 2008). Aman: monsoon rice, Boro: summer rice, Aus: pre-monsoon rice. Table 20: Values of the calibration indicators defined by Eq. (10) to (13). Table 21: Watershed-average dry crop yields simulated by SWAT in the final calibration (calibration run n°100) over the calibration period (1998 – 2008). Table 22: Simplifications and limitations of the modelling. Table A.1: Soil parameters. Light orange: data from the srcinal soil map from the Indian National Bureau of Soil Survey and Landuse Planning. Table A.2: SWAT vegetation / crop parameters. Table A.3: Sources of irrigation per administrative blocks containing the Jaldhaka watershed, year 2004/5 Table A.4: SWAT calibration steps.

LIST OF ABBREVIATIONS ALAI_MIN

SWAT parameter, minimum LAI for plant during dormant period [L2/L2]

ALOS ALPHA_BF ArcSWAT AVNIR AWC AWM B BLAI CGWB CH_K(1) CH_K(2) CH_N(1)

Advanced Land Observing Satellite SWAT parameter, baseflow alpha factor [-] ArcGIS interface for SWAT Advanced Visible and Near Infrared Radiometer SWAT parameter, available water capacity (AWC, [L3/L3]) Agricultural water management Baseflowintothestreams[L/T] SWAT parameter, maximum potential LAI [L2/L2] Central Ground Water Board SWAT parameter, effective hydraulic conductivity in tributary channel alluvium [L/T] SWAT parameter, effective hydraulic conductivity in main channel alluvium [L/T] SWAT parameter, Manning's value for the tributary channel [-]

CH_N(2) CHTMX CN2 DEEPST DEM Dnet DPDWB EPCO ESCO

SWAT parameter, Manning's value for the main channel [-] SWAT parameter, maximum canopy height [L] SWAT parameter, initial soil curve number for moisture condition II [-] SWAT parameter, initial depth of water in the deep aquifer [L] Digitalelevationmodel Netgroundwaterdraft[L/T] Development & Planning Department - West Bengal SWAT parameter, plant uptake compensation factor [-] SWAT parameter, soil evaporation compensation factor [-]

viii

38 40

42 43 44 46

47 52 53 55 62 65 66 68

ETa ET0 FLOW_OUT GIS GPS GT GW_DELAY GW_Q GW_RCHG GW_REVAP GWQMIN h HRU HWSD I IDC IWMI LAI M masl mbgl NS NShigh NSlow O P PGIS PHU RCHRG_DP RDMX REVAPMN RG SEI SHALLST SOL_K Sol_Z SOL_ZMX SRTM SURLAG SWAT SWID Sy t T_BASE T_OPT WISE WTF

Actual evapotranspiration [L/T] Reference evapotranspiration [L/T] SWAT ouput, average daily streamflow out of reach during time step [L/T] Geographical Information System GlobalPositioningSystem Groundtruthing SWAT parameter, groundwater delay time [T] SWAT ouput, groundwater baseflow contribution to streamflow [L] SWAT ouput, recharge entering the shallow aquifer [L] SWAT parameter, groundwater evaporation coefficient [-] SWAT parameter, threshold depth of water in the shallow aquifer required for baseflow to occur [L] Groundwaterpiezometriclevel[L] Hydrologicresponseunit Harmonised World Soil Database Irrigation[L/T] SWAT parameter, land cover / plant classification International Water Management Institute Leafareaindex[L2/L2] Biasindicator[-] Meterabovesealevel[L] Meter belowground level [L] Nash and Sutcliffe (1970) efficiency [-] Modified version of NS to emphasise on high flows [-] Modified version of NS to emphasise on low flows [-] Netgroundwaterunderflow[L/T] Rainfall[L/T] ParticipatoryGIS SWAT parameter, total number of heat units or growing degree days needed to bring plant to maturity SWAT parameter, deep aquifer percolation fraction [-] SWAT parameter, maximum root depth [L] SWAT parameter, threshold depth of water in the shallow aquifer required for evaporation or percolation to the deep aquifer to occur [L] Total groundwater recharge[L/T] Stockholm Environment Institute SWAT parameter, initial depth of water in the shallow aquifer [L] SWAT parameter, Soil conductivity, [L/T] SWAT parameter, depth from soil surface to bottom soil layer [L] SWAT parameter, Maximum rooting depth [L] Shuttle Radar Topography Mission SWAT parameter, surface runoff lag coefficient [-] Soil Water Assessment Tool State Water Investigation Directorate (West Bengal) Specificyield[L3/L3] Time [T] SWAT parameter, minimum (base) temperature for plan growth [°C] SWAT parameter, optimal temperature for plan growth [°C] World Inventory of Soil Emission Potentials Water Table Fluctuation

ix

x

Stockholm Environment Institute

1

INTRODUCTION

Agricultural Water Management (AWM) interventions are often a rst step towards increasing smallholder farmers’ yield levels, their incomes and household food security, in many developing countries. Globally, smallholder farming systems may have the potential to increase current yield levels 2-4 times, and water productivity gains potentially more than double (Rockström 2003). The AgWater Solution project (http://awm-solutions.iwmi.org/) is systematically assessing opportunities to invest in agricultural water management interventions at local to continental scale, to enhance smallholder farmers livelihoods. However, agricultural development and intensication can also unintentionally impact various social and environmental dimensions where the interventions are adopted. This report considers the AgWater Solution project watershed of the Jaldhaka river, also known as Dharla, a tributary of the Brahmaputra river. It is a transboundary river srcinating in Bhutan, ow ing through India and joining the Brahmaputra in Bangladesh (Figure 1). The Jaldhaka watershed is one of four project watershed sites, subject to a suite of assessments on agro-hydrological, livelihood and institutional contexts undertaken to identify what potential opportunities there are at a local (watershed) scale and how potential interventions may impact the environment, in particular water resources, and livelihoods.

Figure 1: Location of the Jaldhaka / Dharla river watershed (in purple). The delineation of the Jaldhaka / Dharla watershed were generated in this work. Images adapted from Google Earth

1

Appl ication of SWAT and a Groundwater Model for Impact Assessment

The focus of this work is the development of an application of the Soil Water Assessment Tool (SWAT) and of a simple lumped groundwater model to study the impacts on hydrological balance and crop production under different scenarios of agricultural interventions. This working paper presents the methodology deployed for data collection and agro-hydrological modelling of the surface and groundwater resource of the Jaldhaka watershed. The accompanying research report de Condappa et al. (2011) applies the modelling to analyse the current state of the hydrology and agricultural water management scenarios. The following sections introduce the chosen modelling software (section II.), the input dataset (section III.), the set up of the models (section IV.) and will end with a discussion (section V.).

2


2

INTRODUCTION TO THE MODELLING SOFTWARES

The primary hydrological model selected for this purpose was the Soil and Water Assessment Tool (SWAT) developed by the United State Department of Agriculture and Texas A& M University. SWAT simulates the different surface and ground hydrological components as well as crop yields. Since its modelling of the groundwater is extremely simplied and the groundwater is a prominent water resource for agriculture in the Jaldhaka watershed, the groundwater model developed by Tomer et al. (2010) was also employed to specically describe the groundwater processes. The groundwater model interpreted the available groundwater levels, which is not possible with the used version of SWAT , and guided subsequently the setting of SWAT’s groundwater parameters. The strategy of the modelling that will be detailed in the following sections is illustrated in Figure

Figure 2: Scheme of the modelling

2.1

Soil and Water Assessment Tool (SWAT)

For the application to the Jaldhaka watershed, version 433 of SWAT 2009 was used and it was operated through the interface ArcSWAT version 2009.93.4. General information on this model can be found on the website http://swatmodel.tamu.edu and in the references Arnold et al. (1993), Srinivasan and Arnold (1994), Arnold et al. (1995) and Arnold et al. (1998)

2.2

Groundwater model

The groundwater model considered here was developed by Tomer et al. (2010). It is based on a combination of groundwater budget and the Water Table Fluctuation (WTF) technique. The WTF technique has widely been applied to link the change in ground water storage with resulting water table uctuations through the storage parameter (specic yield). The WTF is a lumped model based approach suited when limited hydraulic head measurements made at a nite number of observation wells and also little hydrological, geological and meteorological information is available. It was rst used to estimate ground water recharge (e.g., in West Africa by Leduc et al. (1997), in Korea by Moon et al. (2004)) and has been extended to estimate change in groundwater storage (e.g., in California by Ruud et al. (2004)) or the ground water recharge and the specic yield (e.g., in India by Maréchal et al. (2006)) with the combined use of groundwater budget.

3


The main limitations of the WTF modelling method are: (i) it requires the knowledge of the specic yield of the saturated aquifer at a suitable scale, (ii) its accuracy depends on both the knowledge and representativeness of water table uctuations and (iii) it does not explicitly take into account the spatial variability of inputs, outputs, or parameters and considers the catchment as an undivided entity and uses lumped values of input variables and parameters. This approach was however relevant to this work as observed time-series of groundwater levels were available at different locations while very few other hydrogeological data were obtained. As SWAT’s groundwater module does not simulate piezometric levels, the groundwater model enabled the use of the available measured groundwater levels. The mathematical expressions at the core of the groundwater model developed by Tomer et al. (2010) is the groundwater budget: (1)

where Sy [L 3/L3] is the specic yield, h [L] is the groundwater piezometric level, t [T] is the time, RG [L/T] is the total groundwater recharge due to rainfall and other sources including irrigation and recharge from streams, D net [L/T] is the net groundwater draft, B [L/T] is the baseow into the streams and O [L/T] is the net groundwater underow from the area across the watershed boundary. The term represents the total discharge (Tomer et al., 2010). In this work, we assumed that O represented regional deep aquifer processes and was nil at the scale of the Jaldhaka watershed. Moreover, following the approach of Park and Parker (2008), a linear relationship was assumed between the baseow B and the level h: (2)

where λ [1/T] is a rate coefcient. Replaced in Eq. (1) it gives: (3)

The equation (3) is a linear ordinary differential equation, which can be solved analytically. Following the guidance of Simon (2006), the analytical solution was converted into a discrete equation for the ease of modelling, which can be written as: (4)

where A [-] is called the discharge parameter, k is the index for time and the discharge was equal to: (5)

As commonly assumed, the recharge RG is calculated linearly from rainfall P [L/T] and irrigation I [L/T]:

4


(6)

where r [-] is the recharge factor. Note that the irrigation I is only equal to Dnet if groundwater is the only source of irrigation water (e.g., no irrigation from river). As in the Jaldhaka watershed the groundwater data did not show a constant recharge factor, a time varying recharge factor was assumed, calculated from the rainfall P and irrigation I: (7)

where a [-] and b [T/L] are recharge parameters. Finally, the total recharge RG is expressed as: (8)

and Eq. (4) and (8) are used during the calculations of the groundwater model.

5


3

BIOPHYSICAL DATA OF THE JALDHAKA WATERSHED

Biophysical data were gathered by conducting eld work, request to relevant organisations and using publicly available data on internet.

3.1

Digital Elevation Model

3.1.a

Source

3.1.b

Analysis

The source of the Digital Elevation Model (DEM) is the Shuttle Topography Mission (SRTM), as pre-processed by Jarvis et al. (2008).

In Figures 3 and 4, three topographic regions can be identied in the watershed (Table 1): •

mountainous upstream (18 per cent of the watershed), where elevation ranges from 500 to more than 4,000 meter above sea level (masl) and slope from 3 to 40 degrees (within the watershed),

•

piedmont upstream (22 per cent of the watershed), where elevation ranges from 100 to 500 masl and slope from 1 to 3 degree,

•

and plain middle and downstream (60 per cent of the watershed), where elevation ranges from 100 to 18 masl and slope less than 1 degree. The prole of transect dened on Figures 3 and 4 is placed in Figure 5. A striking characteristic of this watershed is the atness in the plain which:

•

makes the delineation of the watershed boundary downstream highly uncertain,

•

and entails invasive oods from neighbouring rivers, in particular from the Teesta river bordering the watershed in the west; during these events, the delineation is further uncertain as rain falling outside of the watershed’s border contributes to the ood of the Jaldhaka river, hence the dened watershed boundaries varies with rainfall and ood events.

3.2

Streamflow data

3.2.a

Source

Obtaining streamow from Indian organisations (Central Water Commission, Irrigation and Waterways Department) was impossible during the span of this study. Instead the Bangladesh Water Development Board provided ow of the Jaldhaka at two gauges stations, downstream in the Bangladeshi part: Taluk-Simulbari and Kurigram (Figure 6). The measurements were instantaneous readings of height, i.e., no average over a time-span, from August 1998 to June 2009 with variable frequencies: •

at Taluk-Simulbari: almost daily up to June 2002, then weekly,

•

at Kurigram: about twice a month.

6


Figure 3: Digital Elevation Model from the Figure 4: Slope derived from the DEM, the two Shuttle Radar Topography Mission and locationslocal meteorological and streamflow gauge where climatic and streamflow data was stations available

4,500 Crest of the watershed

Table 1: Topographic regions of the Jaldhaka watershed

4,000 t n o m d ie P

Mountains

3,500 l) 3,000 s a m ( 2,500 n o it 2,000 a v e l 1,500 E

Plains

1,000

Boundary of the watershed

Topography

Area (km²)

Share (%)

Mountains

1,031

18

Piedmont

1,270

22

Plains

3,494

60

Total

5,795

100

500 0

0

20

40

60

80 100 120 140 160 180 200

Distance (km)

Figure 5: Topographic profile of the transect defined in Figures 3 and 4

7


3.2.b

Inspection of the data

The number of measurements is higher at Taluk-Simulbari, especially during the low ow season (Table 2). The hydrographs of both stations were compared and some measurements at Kurigram appeared suspicious (Figure 7); these were removed. In total, 878 measurements at Taluk-Simulbari and 197 at Kurigram were used to calibrate the SWAT application for the Jaldhaka watershed.

Table 2: Available number of measurements at Taluk-Simulbari and Kurigram stations. Source of data: Bangladesh Water Development Board. Originaldata Kurigram #

Filtereddata

Taluk-Simulbari

Kurigram

Taluk-Simulbari

% of days

#

% of days

#

% of days

#

% of days

Low flows November to April

112

6%

674

34%

97

5%

674

34%

High flows May to October

112

5%

204

9%

100

5%

204

9%

Total

224

5%

878

21%

197

5%

878

21%

As Kurigram is located downstream of Taluk-Simulbari, the ow at Kurigram should be greater than at Taluk-Simulbari. The average monthly ow during the months July to March seems to satisfy this criteria (Figure 7). However during April to June, when the rst major rain events occur, this is not any longer the case. Without a detailed knowledge of the local conditions, we speculated that this may be due to diversion infrastructures that are visible on Google Earth, which divert part of the raising water (Figure 8). Another apparent anomaly is that peaks of the discharge at Kurigram are much greater than at TalukSimulbari. Three possible explanations could be: •

there are errors in measurements of peak ows at Taluk-Simulbari and / or at Kurigram; we think that these errors are most likely at Kurigram where the standard deviation is very important for measurements in July (Figure 7),

•

in this at zone the tributaries of the Brahmaputra (in particular the Teesta and Torsa rivers) converge, hence it is possible that ood water from these neighbouring rivers invade the part of the watershed between Taluk-Simulbari and Kurigram, creating much higher ows at Kurigram,

•

Kurigram is just 20 km upstream from the massive Brahmaputra, hence it could be that ood water from the Brahmaputra travel upstream; however, according to the DEM, the difference in elevation from Kurigram and the conuence is about 9 m, which does not favour this possibility.

3.2.c

Analysis of the ltered data

The ow of the Jaldhaka is perennial at both locations, although there is an important seasonal contrast (Figure 7). The low ow season is between November to April, the average base discharge is 8


4,000

Kurigram

Taluk-Simulbari

3,500 3,000 ) s / 2,500 3 m ( e 2,000 g r a h c s i 1,500 D

1,000 500 0

Jun-98 Oct-99 Mar-01 Jul-02 Dec-03 Apr-05 Aug-06 Jan-0 8 May-09

3,000 Kurigram Taluk-Simulbari 2,500

) s / 2,000 3 m ( e g r a 1,500 h c s i D

1,000

500

0

1

2

3

4

5

6

7

8

9 1 0 11 12

Month Figure 7: Available time-series for streamflows measured at Taluk-Simulbari and Kurigram stations, unfiltered (left) and average monthly streamflow, filtered (right); the vertical error bars indicate the statistical standard deviation of daily streamflows Source: Bangladesh Water Development Board

9


Figure 8: Zoom around Kurigram on Google Earth where are visible the infrastructures for water diversion as well as the neighbouring rivers, in particular the massive Brahmaputra.

114 m3/s (coefcient of variation of 52 per cent) at Kurigram and about 11 per cent of the annual ow occurs during this period. The high ow season occurs between May to October with a sharp raise of the discharge in June. Occurrence of ood is common, often in the month of July, with a maximum of 3,700 m3/s measured at Kurigram in July 2005. According to a personal communication from the Irrigation and Waterways Department – Cooch Behar, the maximum ow measured at Mathabhanga (the inlet of sub-watershed 13 on Figure 6) is 10,120 m 3/s, about three times more the maximum observed during the time period August 1998 to June 2009 (3,700 m3/s). The force of the oods can be such that the main river of the region, the Teesta, shifted its course from the Ganges to the Brahmaputra only in the last three hundred years (Kundu and Soppe 2002). The comparison of rainfall and streamow patterns (Figure 9) shows that both signals reach their maximum in the same month (July) but there is a lag: the increase in streamow at the beginning of the monsoon is not as rapid as for rainfall and the decrease of ows is buffered at the end of the rain season. Moreover, there is a noticeable baseow during the dry season.

3.2.d

Processing for modelling – generation of the sub-watersheds

The DEM (section III.1.) was used as such in ArcSWAT and the Automatic Watershed Delineation routine of ArcSWAT generated the set of sub-watersheds for the modelling. As the downstream part of the watershed is very at, the tributaries downstream were digitised on Google Earth and included in the Automatic Watershed Delineation so as to carve their river beds in the DEM. The same was done with tributaries of the neighbouring Teesta river system (Figure 1) for a correct demarcation. As we had no data on the ow of the Jaldhaka at the conuence with the Brahmaputra, the outlet of the watershed chosen in this work was not this conuence (6,140 km²) but the Kurigram station (5,795 km²) (Figures 2 and 3). 10


1,200

1,000

Jaldhaka watershed rainfall Streamflow at Kurigram

800 ) th n o m 600 / m m ( 400

200

0

1

2

3

4

5

6

7

8

9 1 0 11 12

Month Figure 9: Representative average rainfall for the Jaldhaka watershed, as calculated by SWAT, and average streamflow at Kurigram (period 1998 – 2008) The vertical error bars indicate the statistical standard deviation of monthly values (in mm/month).

An important parameter of the delineation routine is the minimum area of the sub-watersheds. As sub-watersheds carry the climatic characteristics of the watershed, it is advisable to consider a sufcient number of sub-watersheds (S. L. Neitsch et al. 2005). However the observed streamow was only available at two gauge stations and climatic data about every 0.5° (cf. section III.3. and Figure 3), the minimum area threshold was chosen equal to 200 km², which generated a sufcient number of 14 sub-watersheds. Four additional sub-watersheds were manually created (Figure 6): •

two having the Indian streamow gauge stations as outlet (NH-31 and Mathabhanga), in case data from these stations become available at a later stage (sub-watersheds n°3 and 10),

•

one for the Taluk-Simulbari station (sub-watershed n°17),

•

and a last to represent climatic data from Jalpaiguri meteorological station (sub-watershed n°8).

11


3.3

Climate data

3.3.a

Sources

Two sources of climate data were obtained (Figure 3): •

daily rainfall, min and max temperatures, wind and humidity at Jalpaiguri and Cooch Behar stations, period 1988 – 2008, from the Regional Meteorological Centre, Kolkata;

•

gridded daily rainfall at a resolution of 0.5°, period 1971 – 2005, from the National Climate Centre, Pune (Rajeevan and Bhate 2008); 8 pixels from this dataset are within and near the watershed. The coverage of these dataset is summarised in Table 3.

Table 3: Available climatic time-series and gaps in the datasets. RMC stands for Regional Meteorological Centre (Kolkata) and NCC for National Climate Centre. Variable

Time period

Resolution

Rainfall

1988 2008

Daily

Rainfall

1971 2005

Daily

Min temperature

1988 2008

Daily

Max temperature

1988 2008

Daily

Humidity

1988 2008

Daily

Wind

1988 2008

Daily

3.3.b

Missing measurements

Available measurements (days)

(days)

( per cent)

6,040

1,631

21

CoochBehar

7,034

637

8

Jalpaiguri

12,784

0

0

Whole India, resolution 0.5°

6,036

1,635

21

CoochBehar

5,868

1,803

24

Jalpaiguri

6,039

1,632

21

CoochBehar

7,027

644

8

Jalpaiguri

6,017

1,654

22

CoochBehar

6,353

1,318

17

Jalpaiguri

3,656

4,015

52

CoochBehar

4,467

3,204

42

Jalpaiguri

Location

Source

RMC

NCC

RMC

RMC

RMC

RMC

Analysis

The annual rainfall is important in the Jaldhaka watershed with values uctuating from 2,000 to almost 5,000 mm/year, with an average of 3,500 mm/year at Cooch Behar station, during period 1988 to 2008 (Figure 10 and Table 4). The rainfall has a monsoonal seasonal pattern, with a relatively dry season from November to March and a rainy season from April to October (Figure 10). Almost all of the annual rain falls in the rainy season (98 per cent), especially between June to September (80 per cent). Daily rainfall intensities can be very high during the peak of the monsoon, with in average about 33 mm/day per rain event, occurrence every year of events greater than 100 mm/day and a maximum recorded in the magnitude of 470 mm/day in 1999. Some daily rain events usually occur every month of the dry season, but of much lesser magnitude. 12


Jalpaiguri

Cooch Behar

500

500

450

450

) y 400 a d / 350 m300 (m ll 250 a f in 200 a r liy 150 a 100 D

) y 400 a d / 350 m300 (m ll 250 a f in 200 a y lir 150 a 100 D

50 0

50

87 88 89 90 91 92 93 94 95 96 97 98 99 0 0 01 02 03 04 05 06 07 08 09

0

87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09

Year

Year

Jalpaiguri

Cooch Behar

5,000

5,000

)r 4,500 a e 4,000 y / 3,500 m 3,000 (m ll 2,500 a f in 2,000 ra l 1,500 a u 1,000 n n A 500

)r 4,500 a e 4,000 y / 3,500 m 3,000 (m ll 2,500 a f in 2,000 ra l 1,500 a u 1,000 n n A 500

0

88 89 90 91 92 93 94 95 96 97 98 99 0040102 03 04 05 06 07 08

Year

88 89 90 91 92 93 94 95 96 97 98 99 0040102 03 04 05 06 07 08

Year

Jalpaiguri ) 1,000 th Monthly n 900 o Daily m 800 event / 700 m 600 (m 500 ll fa 400 in 300 a r ly 200 h t 100 n o 0 M 1 2 3 4 5 6 7 8 9 101 11 2

0

80 70 60 50 40 30 20 10 0

Month

) y a d / m m ( t n e v e in a r ly i a d e g ra e v A

Cooch Behar ) 1,000 h t Monthly n 900 o Daily m 800 / 700 event m 600 m ( 500 ll fa 400 n i 300 ra ly 200 th 100 n o 0 M 1 2 3 4 5 6 7 8 9 101 11 2

80 70 60 50 40 30 20 10 0

Month

) y a d / m (m t n e v e in ra ly i a d e g a r e v A

Figure 10: Rainfall at Jalpaiguri and Cooch Behar stations (period 1988 - 2008). Top: daily rainfall. Middle: annual rainfall. Bottom: average monthly rainfall, the vertical error bars in red indicate the statistical standard deviation of daily rainfall (in mm/day) Source: Regional Meteorological Centre, Kolkata

Temperature in the watershed is moderately warm with a winter season from November to February, where minimum monthly temperature is about 10°C in the plain, and a summer season from March to October, where maximum monthly temperature is about 32°C in the plain (Figure 11). Wind measurements from both local stations show some differences and after comparing with average data available 13


Table 4: Annual rainfall at Jalpaiguri and Cooch Behar stations (period 1988 - 2008) Source: Regional Meteorological Centre, Kolkata

Station Jalpaiguri Cooch Behar

Range(mm/year) 2,000 4,800 – 2,500 4,900 –

Average(mm/year) 3,375 3,500

CV(percent) 20 19

on the website of the Indian Meteorological Department, measurements from Jalpaiguri station appear erroneous. Values from Cooch Behar indicate that there is a clear variation of wind with time, with a maximum of about 6 m/s in April followed by a continuous decrease to 2 m/s in December. Monthly humidity is high and does not vary much monthly, taking a minimum of 60 per cent in March and a maximum of 85 per cent during several months in winter and summer. High humidity and moderately warm temperature imply that the reference evapotranspiration (R. G. Allen et al. 1998) is modest in the watershed, with an average value of 1,300 mm/year (period 1998 – 2008). As a consequence, the ratio [rainfall] / [reference evapotranspiration] is particularly high as it equals 2.5, which is a distinguishable characteristic of this watershed as compared to the other watersheds studied by the AgWater Solution Project. With respect to daily variability of the climate data within a month, unsurprisingly rainfall is the most variable followed by the wind, and less variable is the humidity and almost stable monthly-wise are the temperatures (Figures 10 and 11).

3.3.c

Processing for modelling

Modelling requires a continuous climate input dataset. As the data from the two local stations has some gaps (Table 3) and the daily rainfall gridded data from NCC ends in 2005, we tried to complement both dataset to have a continuous coverage in the period 1998 – 2008, which was the calibration period of the groundwater model and SWAT. More precisely we used multivariate regression method: •

during period 1998 – 2005 to ll gaps in rainfall data from the two local stations Jalpaiguri and Cooch Behar using gridded data as predictor, without considering any time lag in the rainfall of neighbouring stations,

•

during 2006 – 2008, on the contrary, gridded data at the 8 pixels were predicted from measurements at the two local stations.

As for temperatures, humidity and especially wind, a daily missing value was replaced by the average of contiguous days. Gaps of several days were replaced by the average value of the given month. During the processing of input data, the ArcSWAT interface selects for each sub-watershed the meteorological station which is the closest to the centroid of the sub-watershed. In our case, ArcSWAT chose for the whole Jaldhaka watershed the two local stations and three pixels of the gridded data. The equivalent distribution of annual rainfall is mapped on Figure 12. There is no clear pattern of the rainfall with this method of distributing rainfall, with a succession of higher or lower rainfall amount while one moves from upstream to downstream.

14


Jalpaiguri 40 35

Max Min

Cooch Behar 40

Max Min

35

) 30 C (° 25 re tu 20 a r e 15 p m10 e T 5

) 30 C (° 25 re u t 20 a r e 15 p m10 e T 5

0

0 1 2 3 4 5 6 7 8 9 1 0 11 12

1 2 3 4 5 6 7 8 9 1 0 11 12

Month

Month

Jalpaiguri 10 9 ) 8 /s 7 (m 6 d e 5 e p 4 s d 3 n i W2 1 0

Cooch Behar 10 9 ) 8 /s 7 (m 6 d e 5 e p 4 s d 3 n i W2 1 0

1 2 3 4 5 6 7 8 9 1 0 11 12

1 2 3 4 5 6 7 8 9 1 0 11 12

Month

Month

Jalpaiguri 100 90 80 ) 70 % ( 60 y it 50 id 40 m u 30 H 20 10 0

Cooch Behar 100 90 80 ) 70 % ( 60 y it 50 id 40 m u 30 H 20 10 0

1 2 3 4 5 6 7 8 9 10 11 12

Month

1 2 3 4 5 6 7 8 9 1 0 11 12

Month

Figure 11: Average climatic data at Jalpaiguri and Cooch Behar stations (period 1988 - 2008). Top: temperature. Middle: wind. Bottom: humidity. The vertical error bars indicate the statistical standard deviation of daily data Source: Regional Meteorological Centre, Kolkata

15


Figure 12: Distribution of the average annual rainfall in the sub-watersheds, as represented in SWAT (period 1998 - 2008)

3.4

Soils

3.4.a

Sources

Five sources were used: •

the scanned soil map of West Bengal prepared by the Indian National Bureau of Soil Survey and

Landuse Planning, courteously provided by the Indian Space Research Organisation; •

general data on soil texture and fertility for Cooch Behar district, courteously provided by the Principal Agricultural Ofcer, Cooch Behar;

•

the Harmonised World Soil Database which is an international database that combines regional and national soil information worldwide (SOTER, ESD, Soil Map of China, WISE) with the information contained within the FAO-UNESCO Soil Map of the World (HWSD 2009);

16


•

the qualitative description in Kundu and Soppe (2002);

•

and discussions with local soil scientists from the Agricultural University of Cooch Behar.

3.4.b

Analysis

The soil map of West Bengal was clipped to the zone of the Jaldhaka watershed, georeferenced and digitised in GIS (Figure 13). Unfortunately we did not have the detailed notice attached to the soil map and only the brief qualitative information from the legend of the map was available (Table A.1, Annex). Three topographical zones are dened in the soil map: •

the mountainous soils, W001 to W004, shallow, coarse sandy loam,

•

soils in the piedmont, W006 to W008, deep, sandy loam to loam,

•

soils in the plain, W010 to W028, deep, sandy loam to silty loam. The unit Riv additionally describes soils of the river beds.

Figure 13: The georeferenced soil map in the region of the Jaldhaka watershed Source: soil map of West Bengal, Indian National Bu-

Figure 14: The Harmonised World Soil Database and its soil units in the region of the Jaldhaka watershed.

reau of Soil Survey and Landuse Planning

17


The soil spatial units dened by the HWSD in the region of the Jaldhaka watershed (Figure 14) are those of the Digital Soil Map of the World, from the FAO-UNESCO. The HWSD reports additionally some quantitative information from the WISE database (FAO/IIASA/ISRIC/ISS-CAS/JRC 2009), like the textural percentage in sand, silt, clay (Table A.1 in Annex). The information from the Principal Agricultural Ofcer , Cooch Behar, were used to interpret the soil map (Table A.1, Annex). Local soil scientists indicated the same texture for these soils (Loamy Sand) and insisted that generally the clay percentage is low. Kundu and Soppe (2002) report that the mountainous soils (units W001 to W004) are sandy with high inltration rates – this information was used while entering the soil data in SWAT’s database. They also mention that in the plains top-soils are usually Sandy Loam with rather low inltration rates and that there is a textural transition at about 50 cm in depth for a sandier sub-soil.

3.4.c

Pcrocessing for modelling

The soil information had to be processed before entering it in SWAT. In particular, the qualitative description had to be transformed into equivalent quantitative values for the soil database of SWAT. The Table A.1 in Annex summarises the values entered in the soil database. The textural percentages from HWSD were plotted in the United State Department of Agriculture soil textural triangle to check if they were in accordance with the information from the soil map (Table A.1) and Kundu and Soppe (2002) (Figure 15). This was not the case as the top soils’ texture from HWSD was ner (Loam) and sub soils were even ner instead of getting coarser. Corrections were as follows: •

For soils in mountainous regions (W001 to W004), the percentages of clay and silt from the HWSD for top soils were too high while percentage for sand too low, hence 10 per cent was deducted to the percentage of clay and silt and 20 per cent was added to the percentage in sand. For sub-soils, srcinal values from the HWSD were ignored and instead the corrected values of top-soils were considered, reducing further the percentages of silt and clay in favour of the sand content.

•

For soils in piedmont and plains (W006 to W028), the HWSD clay percentage for top soils were too high (more than 20 per cent) compared to the information from local soil scientists, hence the top soil textures of Loamy Sand units were modied by deducing 10 to the clay percentage and adding it to the sand content; silt content was not modied. For the sub-soil, srcinal values from the HWSD were ignored as well and the textures were calculated by deducing 10 per cent and 5 per cent to the silt and clay content of the top soil, adding it to the sand percentage.

The corrected soil textures were indeed matching with description from the various sources (Figure 15). The texture of the unit Riv was not modied. Eventually the qualitative information of the soil map (Figure 13) was translated into equivalent quantitative values using the HSWD and local soil knowledge. The soil hydrologic group, required by the ArcSWAT interface, was derived from the soil texture: •

group A: coarse sandy loams, units W001, W002, W004 and Riv,

•

group B: ne sandy loams, units W003, W006, W008 to W025,

•

group C: loamy soils, units W007, W026 and W028.

18


Top soil

Top soil

Original texture from the HWSD

Corrected texture

Sub soil

Sub soil

Original texture from the HWSD

Corrected texture

Figure 15: Plot in soil textural triangle of the United State Department of Agriculture

After the soil texture, another important soil characteristic in SWAT is the soil Available Water Capacity (AWC, [L3/L3]). The HWSD provides approximate AWC but we preferred to enter values considering the texture of each soil unit (Figure 15) and adapting the capacities advised by Kundu and Soppe (2002). Moreover, there is a sort of continuity between of the soil AWC and the specic yield used by the groundwater model. The value of specic yield was often 0.15 in the plain, less in mountainous sub-watersheds and this was reected in the AWC, as it will be detailed in section IV.1.d.. Ultimately: •

unit Riv: top soil AWC = 0.08,

•

unit W001: top soil AWC = 0.10,

•

units W002 and W004: top soil AWC = 0.15, sub-soil AWC = 0.10,

•

units W003, W006, W008 to W025: top soil AWC = 0.20, sub-soil AWC = 0.15,

19


•

units W007 and W026: top soil AWC = 0.25, sub-soil AWC = 0.20,

•

unit W028: top soil AWC = 0.30, sub-soil AWC = 0.25.

The soil depths were also specied combining the information from the soil map and Kundu and Soppe (2002): •

shallow soils: top soil 50 cm, no sub-soil,

•

moderate shallow: top soil 30 cm, sub-soil 70 cm,

•

deep: top soil 40 cm, sub soil 100 cm,

•

very deep: top soil 50 cm, sub-soil 200 cm. The parameter SOL_ZMX was given the value of 300 cm.

The soil conductivity (SOL_K) was chosen with respect to the soil hydrologic group as advised by Neitsch et al. (2010). Finally, the soil map (Figure 13) was approximatively extended using the satellite imageries to cover all the delineation of the watershed.

3.5

Groundwater data

3.5.a

Source

Groundwater levels in the Indian part of the watershed, from Jalpaiguri and Cooch Behar districts, were obtained from two organisations (Table 5): •

IWMI provided reading from the Central Ground Water Board (CGWB),

•

and the State Water Investigation Directorate (SWID) of West Bengal.

Table 5: Available measured groundwater levels in the Indian part of the watershed. CGWB stands for Central Ground Water Board and SWID for State Water Investigation Directorate. Monitoring organisation

Number of observation wells Total Jaldhaka watershed

Period

Measurementfrequency

Total

Jaldhaka watershed

CGWB

49

12

1996 – 2006

1996 – 2006

Every 3 months

SWID

112

43

1988 – 2009

1994 – 2009

Every 3 months

3.5.b

Analysis

These measurements are mainly located in the plain, with some few in the piedmont (Figure 16). The spatial resolution of the measurement in the Indian part of the watershed is satisfactory with a sufcient number of the observation wells. However, the time resolution is scattered as measurements are not monthly but almost 3 months with some irregularities, hence we may miss some time variation 20


Figure 16: Location the observation for Central Ground of Water Board and wells SWIDfor forgroundwater State Water level measurement. CGWB stands

of the groundwater levels. In particular we may not know exactly when the groundwater levels are the deepest and the shallowest. Leaving aside the groundwater system in the mountains upstream where no observations are available, the groundwater regime can be categorized in two categories: ( i) in the plain downstream and (ii) in the piedmont area middle stream. In the plain downstream, the aquifer system is alluvial and composed of ancient sediments from succession of the Ganga – Brahmaputra river systems (Kundu and Soppe 2002; CGWB 2009). The groundwater levels are shallow. According to the scattered timeseries of groundwater levels, the water table is apparently deepest in April before the monsoon, uctuating from 1 to 5 meter below ground level (mbgl), and shallowest in August during the monsoon, uctuating from 0 to 3 mbgl (Figure 17). Typical groundwater levels are those of wells PTC-25 and PTC-9 (Figure 18). In the piedmont area, the aquifer is composed of more recent sediments carried by the tributaries from the mountainous upstream areas (Kundu and Soppe 2002; CGWB 2009). The groundwater levels are also shallow, although some wells show deeper level. The levels are apparently the deepest between February to April before the monsoon, uctuating from 2 to 15 mbgl, and the shallowest in August during the monsoon, uctuating from 1 to 11 mbgl (Figure 17). Typical groundwater levels are those of wells D-10 and D-12 (Figure 18). Watershed-wise, the average groundwater level is between 2 to 4 mbgl (Figure 17). The last 16 years of the groundwater levels time-series show little inter-annual water table uctuations (Figure 17 and 21


Source: Central Groundwater Board (CGWB)

Source: State Water Investigation Directorate (SWID)

Figure 17: Measured groundwater levels in the Jaldhaka watershed. In pale: level of different wells. In black: average of all the wells

In piedmont

In plain

In piedmont

In plain

Figure 18: Typical groundwater levels in the Jaldhaka watershed. The wells are located on Figure 16. The vertical error bars indicate the statistical standard deviation Source of data: State Water Investigation Directorate (SWID)

22


Figure 18): there is no noticeable trend to increase or to decrease, i.e., watershed-wise the groundwater levels have been stable during the period 1994 – 2009. This is consistent with observations from Shamsudduha et al. (2009). Interpolation of observed groundwater piezometric levels in India using the Inverse Distance Squared Weighting method shows that the levels follow the topography, which is expected for an alluvial aquifer system (Figure 19). The gradient is along the North to South orientation and direction in piedmont region and changes towards the South – East in the plain. The groundwater ow is therefore in the same orientation and direction of the surface water and converges towards plains in Bangladesh, in particular towards the Brahmaputra river. This is again consistent with Shamsudduha et al. (2009). If we compare the cases when the available observed piezometric levels are the shallowest (August) versus the deepest (April), there is a general shift of the contours along the ow direction, i.e., the topography, but the relative distribution does not change signicantly.

3.5.c

Pre-processing

The dataset from SWID contained more measurements than the one from CGWB (Figure 17), hence only readings from SWID were considered afterwards. Moreover the groundwater model focused on the wells located within the watershed, which amounted to 33 wells.

3.6

Land-use

A high spatial resolution landuse map of the Jaldhaka was generated within the context of the AgWater Solution Project. Satellites images were acquired and we contributed to the development of the landuse map by conducting the groundtruthing and helping in the landuse classication.

Figure 19: Interpolation of average piezometric levels observed by the State Water Investigation Directorate (SWID) (period 1994 - 2009). 23


3.6.a

Satellite images

Six images from the Advanced Visible and Near Infrared Radiometer type 2 (AVNIR-2) aboard the Advanced Land Observing Satellite (ALOS) were acquired by IWMI. The images are of 10 m spatial resolution and have four bands, three in visible and one in near infrared. The images do not cover the Jaldhaka watershed completely (Figure 20). The six images were for two locations with three replications for each location. The purpose of having multiple images from different seasons is to enable to better extract crop type and rotation information (Cai, personal communication, 2010). Among the three dates, October (31/10/2008) have the clearest views. The two views of January (31/01/2009) are slightly hazy while views of March ( i) are from different dates (15/03/2008 and 18/03/2009) and (ii) there is a spatial discrepancy between both views. This combination of dates was representative of the year 2008 and thus an additional difculty was that groundtruthing was carriedout in April 2010, at a date different to the satellite images.

3.6.b

Groundtruthing

Before the eld work, a draft unsupervised classication was produced on a southern view, to choose the location the groundtruthing (GT) sites (Figure 21). In total, 90 GT sites were visited in April 2010 and two types of observations were carried-out: •

precise in 40 (GT points 1 to 40) of these sites,

•

brief in the remaining 50 sites (GT points 41 to 90).

Taking inspiration from Cai and Sharma (2010), a GT form pertaining to the vegetation distribution and the crop system (sequence, growing period etc.) was prepared. The choices of the sites were as follows (Figure 21):

Date North view: 31/01/2009



Date South view: 31/01/2009



Figure 20: Satellite images acquired for high resolution landuse mapping. Note the demarcation between the north and south view

24


Figure 21: Location of the groundtruthing sites visited in April 2010 and draft unsupervised classification of the landuse. Right: zoom on the transect (note on this view the discrepancy

•

location of sites was decided in the eld while driving on roads referring (i) downstream to the different classes of the unsupervised classication and ( ii) upstream to the visible landuse as observed on srcinal satellite image and GoogleEarth,

•

sites were visited while walking along a transect of about 8.5 km; this transect was chosen on the draft classication so as to be representative of the downstream part of the watershed.

Precise observations meant a 360° assessment of the land cover (e.g., urban, natural vegetation, agricultural land) and interview of farmers on the spot to distinguish the cropping systems within agricultural lands. Two local eld assistants helped for the survey by interviewing farmers and lling the form. An example of this form is shown in Figure A.1 (Annex). Brief observations were quick observation without stopping the car, noting the main land cover categories (e.g., forest, settlement, light forest, tea plantation, agricultural land). Some photos of different landuse are placed in Figure 24. The distribution of GT sites is greater middle and downstream as the draft classication was only available for this part. This is acceptable as most of the agricultural land lies in this part of the watershed. From the GT observations, main cropping sequences were identied (Table 10). The cropping calendar of these sequences is illustrated on Figure 22.

3.6.c

Generation of the landuse map

The landuse map was generated in collaboration with IWMI. The rst task was to geo-rectify the satellites as some gaps have been observed as compared to GPS measurements. The GIS / remote sensing expert from IWMI ran an unsupervised classication on two separate sets: the rst set con taining the 3 northern views and the second set containing the 3 southern views. The advantage of processing separately images of different seasons is the changes in vegetation conditions across both 25


Rainfed Monsoon_Rice

Jute


Rainfed Monsoon_Rice Jute

Pre-monsoon_Rice

Irrigated Monsoon_Rice

Irrigated Monsoon_Rice Pre-monsoon_Rice

Jute


Potato

Summer_Rice

Wheat

1

2

3

4


5

6

7

Potato

Irrigated Monsoon_Rice Potato Jute


Irrigated Monsoon_Rice Summer_Rice


Irrigated Monsoon_Rice Wheat

8

9

10

11

12

Month

Figure 22: Calendar of the main cropping sequences in the Jaldhaka watershed

views (mainly crops) are taken into account while the pixels are clustered. The northern and southern classications created each 20 classes (Cai, personal communication, 2010). As this study focuses primarily on agricultural land, IWMI’s GIS / remote sensing expert tried to identify crop types as well as crop rotations. Temporal spectral changes of each agricultural class were analysed and the trend in vegetation development was assessed. They were then compared to the crop calendar (Table 10) to match with dominant crop types. GT points and Google Earth were used for to aid crop type determinations (Cai, personal communication, 2010). The fact that groundtruthing was conducted at a date different from the srcinal satellite images hindered the classication. This generated 40 classes: 20 for the northern view and another 20 for the southern view. In an attempt to validate, this classication was queried in a buffer of about 100 m radius around each GT location and extracted landuse was compared with GT observations. Discrepancies were important with, for instance, an over-representation of Wheat, an under-representation of Jute and the absence of a category for towns. Hence we tried to reduce the number of classes and enhance their representativeness by: •

merging equivalent northern and southern classes,

•

using Google Earth to recognise the GT observations and associate them with classication clusters,

•

differentiate more precisely Tea from Shrublands,

•

creating manually a new class for towns.

26


Eventually, a high resolution (10 m) landuse map with 11 classes representative of the year 2008 was generated (Figure 23): •

8 classes for non-agricultural lands with in particular three categories of habitation zones, from the smallest to the largest: (i) settlements, with few habitations around trees, ( ii) villages with a greater number of habitations and sparse vegetation and (iii) towns with urbanised areas,

•

and 3 categories of agricultural practices: - Monsoon_Rice → [Pre-monsoon_Rice], i.e., Monsoon_Rice possibly followed by Premonsoon_Rice, - Monsoon_Rice → [Winter Crop] → [Jute or Pre-monsoon_Rice], i.e., Monsoon_Rice, possibly followed by a Winter Crop, possibly followed by Jute or Pre-monsoon_Rice, - Monsoon_Rice → [Winter Crop] → Summer_Rice, i.e., Monsoon_Rice, possibly followed by a Winter Crop, followed by Summer_Rice.

Locally, the monsoon, summer and pre-monsoon rices are called Aman, Boro and Aus respectively. The schedule of each crop is placed in Figure 22. The photos taken at the locations identied on the Figure 23 are placed in Figure 24.

Figure 23: High resolution (10 m) landuse map of the Jaldhaka watershed (year 2008). Photos of the identified observation sites are placed in Figure 24.

27


Table 6: Distribution of the landuse categories (Figure 23) within the Jaldhaka watershed. Landusce ategory

Area (km²)

Forest

(percent)

740

Tea Light or Forest SmallTreesorShrublandorSettlement

510 786

17.2

Monsoon_Rice–>[Pre-monsoon_Rice]

174

Monsoon_Rice –> [Winter Crop] –> [Jute or Pre-monsoon_Rice]

1,356

Monsoon_Rice–>[WinterCrop]–>Summer_Rice

16.2 11.1

309

3.8 29.6 6.8

Village or Fallow

254

5.6

Town

8

0.2

Water

104

bed River

182

2.3 4.0

Cloud

151

3.3

Total

4,574

100.0

Figure 24: Photos of the spots identified on the landuse map (Figure 23)

3.6.d


The landuse map (Figure 23) cannot be used as such by the interface ArcSWAT as its cloud category has to be replaced by a relevant landuse and it does not cover all of the watershed. The cloud category was replaced referring to the landcover visible on Google Earth, ie., generally Forest upstream in Bhutan and Monsoon_Rice → [Winter Crop] → [Jute or Pre-monsoon_Rice] in India and Bangladesh. Before extending the landuse map to part of the watershed not covered, as the category River bed is a mixture of sand and gravels bare soil class (visible on satellite images) where hardly any vegetation grows, it cannot be matched to a SWAT land use class, henceRiver bed was replaced by the category Water. Finally, the map was extended using Google Earth and matching with the soil map:

28


•

upstream it was expanded as Forest,

•

while downstream it was a mixture of Monsoon_Rice –> [Winter Crop] –> Summer_Rice, Monsoon_Rice –> [Winter Crop] –> [Jute or Pre-monsoon_Rice] and Water.

As the GIS raster of the landuse map was to be processed by ArcSWAT, the raster was matched spatially with the DEM, which is the fundamental GIS information for ArcSWAT. This implied in particu lar that the spatial resolution of the enlarged landuse map (10 m) was downgraded to 90 m. The resulting landuse map is placed in Figure 25 and the tabulated areas are in Table 7: these are the landuse information eventually entered in SWAT. We dened 9 landuse categories for SWAT (e.g., FRSJ for Forest, AAAJ for Monsoon_Rice → [Pre-monsoon_Rice]) that were associated to a crop/ vegetation of SWAT’s database. The agricultural units AAAJ, AWJJ and AWBJ were generic crop classes, their associated crop sequences, that will be presented in section III.7., were entered in SWAT at a later stage, while dening SWAT’s management tables (section IV.1.b.).

3.7

Agricultural

3.7.a

Sources

Data on crop yields, area extent and productivity were obtained from 3 sources (Table 8): •

the website of the Development & Planning Department - West Bengal (DPDWB 2005)

•

the Bureau of Applied Economics and Statistics, Kolkata, India,

Figure 25: Modified version of the landuse map (Figure 23, year 2008) entered in SWAT (90 m resolution) 29


Table 7: Distribution of the landuse categories entered in SWAT (Figure 25). Landusceategory

In

Forest Tea or Light Forest

SWAT

Area (km²)

(%)

FRSJ

1338

23.1

TEAB

SmallTreesorShrublandorSettlement

510

FRMJ

Monsoon_Rice–>[Pre-monsoon_Rice]

Monsoon_Rice–>[WinterCrop]–>Summer_Rice

785

AAAJ

Monsoon_Rice –> [Winter Crop] –> [Jute or Pre-monsoon_ Rice]

8.8 13.5

174

AWJJ AWBJ

1,841 342

3.0 31.8 5.9

Village or Fallow

VIFA

256

4.4

Town

URMD

8

0.1

Water

WATR

541

9.3

5,795

100.0

Total

Table 8: Available agricultural statistics. Source

Timeperiod

Spatial resolution

Variables

Crops

Development & Planning Department West Bengal

2003 – 2004

Administrative Blocks of Cooch Behar and Jalpaiguri districts

Area, Yield, Production

Monsoon_Rice, Pre-monsoon_Rice, Summer_Rice, Potato, Jute, Wheat, various pulses

Bureau of Applied Economics and Statistics

1988 – 2009



Monsoon_Rice, Pre-monsoon_Rice, Summer_Rice, Jute, Wheat, Maize, various pulses

Directorate of Agriculture

1998 – 2009



Potato

•

and the Directorate of Agriculture, Kolkata, India.

Additionally, typical crop growing seasons were noted during the eld work described in sec tion III.6.b. (Figure 22) and are reported in Table 10.

3.7.b

Analysis

The major crops in the region of the Jaldhaka watershed are (Figure 22): •

Summer_Rice, called locally Boro: irrigated rice grown before the onset of the monsoon, from February to May.

•

Pre-monsoon_Rice, called locally Aus: partly irrigated rice grown at the onset of the monsoon, from April to June.

30


•

Monsoon_Rice, called locally Aman: rice grown during the rain season, from June/July to September/October, rainfed or partly irrigated depending on the case.

•

Jute: rainfed vegetable bre grown at the onset of the monsoon, from April to June.

•

Winter Crop: irrigated crop following the rain season, which is Potato (predominantly), Tobacco or Vegetables, from November to February; from now we will consider this crop to be Potato.

•

Wheat: irrigated during winter, from January to April.

The irrigation schedule of these crops are described in the following section III.8.. Although Tobacco is an important cash crop in the watershed, in particular in Cooch Behar district, no data could be gathered as this crop is not monitored by governmental organisations. The data from the Development & Planning Department - West Bengal were ignored as these were only for the year 2003/04, however their utility were to assure the homogeneity with statistics from the Bureau of Applied Economics and Statistics and the Directorate of Agriculture. Area and yield of these crops during the period 1998 – 2008 in the administrative block containing the Jaldhaka watershed is shown on Figures 26and 27. There is a gap in the data from the Bureau of Applied Economics and Statistics for the years 2001/02. A striking feature is that the area under Monsoon_Rice is much greater than the other cultivations and is quite stable. The tendency for area under Pre-monsoon_Rice was to gently decrease while potato to increase. The area under Summer_Rice increased sharply in the recent years in the administrative blocks located downstream. Extent of Jute and Wheat cultivation is relatively stable with a small area under Wheat as compared with the other crops. Among the yields, those of Potato are the most varying. The average crop yields in the watershed for the period 1998 – 2009 is estimated by calculating the average yields of the administrative blocks containing the watershed (Table 9). It is noteworthy that the yield of Summer_Rice is greater than Monsoon_Rice and Pre-monsoon_Rice. The agricultural statistics obtained for rice from the Bureau of Applied Economics and Statistics were in two parts: one for the period 1998/99 – 2003/04, which only mentioned for the rice the clean yield, and another for the period 2004/05 to 2008/09, which mentioned for the rice the dry yield and clean yield. Hence only the clean rice yield was available throughout the period 1998 – 2009 but the dry yield was also reported in Table 9 as it was required to compare with SWAT’s outputs (cf. section IV.2.f.).

3.7.c


We derived typical cropping sequences that will be considered in SWAT (Figure 22 and Table 10) from (i) observations during the landuse groundtruthing (cf. section III.6.b.), (ii) Participatory GIS

Table 9: Average yields in the administrative blocks containing the Jaldhaka watershed, period 1998 – 2008. In bracket the average dry yield of rice for period 2004 to 2008. Source of data: Bureau of Applied Economics and Statistics and Directorate of Agriculture. Monsoon rice (Aman) Clean yield (T/ha) 1.5(2.4)

Pre-monsoon rice (Aus) Clean yield (T/ha) 1.4(2.0)

Summer rice (Boro) Clean yield (T/ha) 2.2(2.9)

1.9

Jute Dry yield (T/ha)

Wheat Dry yield (T/ha)

1.8

Potato Yield (T/ha)

18.8

31


35

35 Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj

30

25 ) a h 0 020 ,0 1 ( a e r 15 A

Mathabhanga II I Mathabhanga Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi

10

5

0

Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj

30

25 ) a h 020 0 ,0 1 ( a e r 15 A

Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi

10

5

0

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

Years

Years

Pre-monsoon_Rice (a.k.a. Aus)

Monsoon_Rice (a.k.a. Aman)

35

35 Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II

30

25 ) a h 0 020 0 , 1 ( a e15 r A

10

Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi

5

0

Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II

30

25 ) a h 0 020 0 , 1 ( a e15 r A

10

Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi

5

0

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

Years

Years

Jute

Summer_Rice (a.k.a. Boro) 35

35 Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II

30

25 ) a h 0 020 ,0 1 ( a e r 15 A

10

5

Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II

30

25 ) a h 0 020 ,0 1 ( a e r 15 A

10

5

Sitai Sitalkuchi

0

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

Sitai Sitalkuchi

0

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

Years

Wheat

Years

Potato

Figure 26: Area of the major crops in administrative blocks containing the Jaldhaka watershed Source of data: Bureau of Applied Economics and Statistics and Directorate of Agriculture

32


4.0

4.0

3.5 ) a h / 3.0 (T ld ie2.5 y e ic r 2.0 n a le C

Average of administrative blocks in the Watershed

Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri

1.5

3.5 ) a h / 3.0 (T ld ie2.5 y e ic r 2.0 n a le C


1.5

Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi

1.0 0.5 0.0

Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi

1.0 0.5 0.0

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

Years

Years

Monsoon_Rice(a.k.a.Aman)

Pre-monsoon_Rice(a.k.a.Aus)

4.0

4.0 Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I

3.5 ) a /h3.0 T ( d l ie2.5 y e ic r 2.0 n a e l C

1.5

1.0 0.5


Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi

0.0

Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I

3.5 3.0 ) a /h2.5 T ( d l e i 2.0 y y r D

1.5

1.0 0.5


0.0

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

Years

Years

Summer_Rice (a.k.a. Boro)

Jute

4.0

40 Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I

3.5 3.0 ) a h / 2.5 T ( d l ie2.0 y y r D

1.5 1.0 0.5

Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri


0.0

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

Dinhata I Dinhata II Sitai Sitalkuchi

Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I

35 30 25

) a h / T (20 d l ie Y

15 10 5


0

98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09

Years

Wheat

Dinhata I Dinhata II Sitai Sitalkuchi

Years

Potato

Figure 27: Yield of the major crops in administrative blocks containing the Jaldhaka watershed. Mind the different vertical scale Source of data: Bureau of Applied Economics and Statistics and Directorate of Agriculture

33


Table 10: Typical cropping sequences and associated irrigation schedules in the Jaldhaka watershed. Irrigation duration (days)

Chosen irrigation events

Cropping systems

Landuse category in SWAT

Start

End

Grow ing days

Total pumped (mm)

Rainfed Monsoon_ Rice

AAAJ and AWJJ

June

September

122

-


AAAJ and AWJJ

July

October

123

-

April

June

91

-

July

October

123

247

100

12 mm every 5 days

April

June

91

617

80

15 mm every 2 days

July

October

123

247

100

12 mm 5 every days

November

February

120

247

100

10 mm every 4 days

April

June

91

-

June

September

122

247

100

12 mm every 5 days

February

May

120

1,233

100

25mm every 2 days

January

April

120

247

100

12mm every 5 days

June

September

122

247

100

12 mm every 5 days

Jute Irrigated Monsoon_ Rice

AAAJ and AWJJ

Pre-monsoon_ Rice Irrigated Monsoon_ Rice

AAAJ and AWJJ

Potato

Jute Irrigated Monsoon_ Rice

AWBJ

Summer_ Rice Wheat

Irrigated Monsoon_ Rice

34

AAAJ and AWJJ

100


(PGIS) analysis conducted by SEI (de Bruin et al. 2010) and (iii) the landuse map (Figure 25). The irrigation of these sequences will be explained in the forthcoming section III.8. The next critical steps was to estimate the area of each cropping sequences from the agricultural statistics at administrative block level. These areas were indeed used prior to the modelling to differentiate the agricultural units AAAJ and AWJJ of the landuse map (Figure 25); the unit AWBJ was assigned the sequence Irrigated Monsoon_Rice - Summer_Rice, hence no winter crop was eventually supposed to grown in this unit. As the area growing Monsoon_Rice is much greater than the other crops (Figure 26), we took the assumption that all the elds of units AAAJ and AWJJ grow Monsoon_Rice during the rain season, which may be followed by fallow or crop(s). We did as follows in each administrative block: •

We calculated the average area of each crops (i.e., Monsoon_Rice, Pre-monsoon_Rice, Summer_Rice, Jute, Wheat, Maize, various pulses, Potato) over the last 5 years.

•

The areas of Maize and various pulses were small, hence Maize was combined with Wheat and various pluses with Potato (approximative same growing season).

•

As this differentiation concerns units AAAJ and AWJJ, where there is no Summer_Rice, for each administrative block we subtracted to the Monsoon_Rice area the Summer_Rice extent.

•

We took the additional assumption that each cropping sequence introduced in Table 10 are distinct, hence the area calculate in step 3 is composed of all the cropping sequences dened in Table 10, except Irrigated Monsoon_Rice - Summer_Rice.

•

We expressed the area growing Pre-monsoon_Rice, Jute, Wheat and Potato (calculated in step 1 & 2) as a percentage of Monsoon_Rice’s area estimated in step 3. This percentage was always less than 100 per cent.

•

Then it was decided to distribute the cropping sequences as follows: Area ( per cent) of...

… was equal to...

Irrigated Monsoon_Rice Jute

→

Winter Crop

Percentage of Potato computed in step 5

→

Irrigated Monsoon_Rice Irrigated Monsoon_Rice Rice Rainfed Monsoon_Rice

→

W heat

Percentage of Wheat computed in step 5

→

Pre-monsoon_

Percentage of Pre-monsoon_Rice computed in step 5

→

J ute

Maximum [0,Percentage of Pre-monsoon_Rice computed in step 5 - percentage of Irrigated Monsoon_ Rice Winter Crop Jute] →

RainfedMonsoon_Rice

→

Remainingto100percent

Subsequent to this allocation per administrative block, the percentages were proportionally distributed per sub-watersheds (Table 11). It is reminded that these shares concern SWAT’s landuse units AAAJ and AWJJ, and that the unit AWBJ was assigned the sequence Irrigated Monsoon_Rice – Summer_Rice. This table is indicative as the distribution actually considered during the modelling depends on the spatial discretisation in SWAT (cf. forthcoming section IV.1.b.). Referring to the agricultural statistics (Figure 26), Rainfed Monsoon_Rice is logically the predominant cropping sequence, fol35


lowed by Irrigated Monsoon_Rice → Potato → Jute. Table 11 was utilised while determining SWAT’s management table (cf. section IV.1.b.). The last step of the crop / landuse processing was to dene the relevant SWAT crop / vegetation parameters. The Table A.2 (Annex) summarises the value of the crop parameters of the landuse map. Except for rice, we created a specic category for the vegetation of the Jaldhaka watershed from an existing category of the SWAT database, modifying some parameters as mentioned in the Table A.2 : •

Large trees (FRSJ): we created from the existing SWAT class Forest-Evergreen, modifying the optimal temperature from 30 to 25°C and the maximum height from 10 to 15 m.

Tea (TEAB): from Range-Brush, modifying the maximum LAI from 2 to 3 (cf.http://www.gisdevelopment.net/application/agriculture/yield/rishpf.htm). We used the defaults crop parameters for rice, the minimum LAI from 0 to 0.7 as it’s a perennial crop, the maximum root depth from 2 to 1 m.

Table 11: Indicative distribution per sub-watershed of the cropping sequences within the landuse units AAAJ and AWJJ. Derived with data from the Bureau of Applied Economics and Statistics and the Directorate of Agriculture. Subwatershed


Irrigated Monsoon_ Rice – Premonsoon_ Rice

Rainfed Monsoon_ Rice – Jute

Irrigated Monsoon_Rice – Potato – Jute

Irrigated Monsoon_Rice – Wheat

1

55%

7%

0%

16%

22%

2

46%

11%

0%

23%

20%

3

39%

18%

0%

30%

13%

4

18%

27%

0%

38%

17%

5

11%

35%

0%

25%

29%

6

7%

34%

0%

38%

21%

7

37%

16%

1%

34%

12%

8

49%

19%

0%

22%

10%

9

53%

12%

5%

20%

9%

10

35%

19%

0%

33%

11%

11

61%

7%

1%

25%

6%

12

53%

11%

8%

19%

9%

13

53%

10%

13%

17%

7%

14

52%

10%

19%

12%

7%

15

55%

10%

1%

27%

8%

16

39%

5%

12%

33%

11%

17

52%

18%

7%

14%

10%

18

39%

5%

12%

33%

11%

Watershed

43%

15%

5%

25%

12%

36


Medium trees (FRMJ): from the Large trees category above, reducing the maximum LAI from 5 to 4, the maximum root depth from 3.5 to 2 m, the maximum height from 15 to 10 and the optimal temperature from 25 to 30°C as these trees are more in the plains. Potato and Wheat: from Potato and Spring Wheat, adjusting the temperature factors to match with the Jaldhaka region. Jute: no matching crop was found hence the Jute class was created from the generic agricultural class, taking a maximum LAI equal to 5 (it’s a leafy crop) and maximum root depth equal to 1 m. The last parameter to be estimated for the agricultural crops (non perennial) was the total heat units required for plant maturity (PHU) [°C] dened as (S. L. Neitsch et al. 2005): (9)

calculated over the growing period, where Ti [°C] is the average daily temperature and Tbase [°C] is the plant base temperature. PHU was assessed using average temperature from the two local meteorological stations (cf. section III.3.) and the crop growing periods (Table 10).

3.8

Irrigation

3.8.a

Sources

There were three sources of data. The rst was Mukherji (2007) which provides typical duration of pumping for Summer_Rice, Monsoon_Rice and Potato, from diesel pumps (Table 12). The author coined these gures by conducting eld works in various region of West Bengal, in particular in Cooch Behar district. The second source was DPDWB (2005) which identies the source of water for irrigation in each administrative blocks for year 2004/5 (Table A.3, Annex). The third source was the irrigation frequency per crop (Table 10) obtained during the groundtruthing eld work for landuse mapping (cf. section III.6.b.).

3.8.b

Analysis

Participatory GIS analysis conducted by SEI within the AgWater Solution Project in the downstream part of the Jaldhaka watershed (de Bruin et al. 2010) reveals that farmers largely pump groundwater with diesel pumps from shallow tubewells. Figures from DPDWB (2005) conrm this characteristic and more precisely that (i) the main source for irrigation is groundwater pumped from shallow tubewells and (ii) this is especially true in Cooch Behar district, where lies the downstream part of the watershed. According to this same reference, the next irrigation source is deep tubewells but the statistics for this category seem suspicious if the number is deep tubewells is compared with the area these wells are supposed to irrigate; hence we ignored this category. The following irrigation source is surface water, from canals or river lifts, especially in Jalpaiguri district, where lies the upstream and middlestream part of the watershed. It is noteworthy that there is schematically a spatial differentiation of irrigation source in the watershed: •

the upstream and middlestream part of the watershed mainly irrigates from surface water source, more precisely by diverting and lifting the river (canals and river lift);

37


•

while the downstream part uses mainly groundwater (diesel pumps with shallow tubewells).

3.8.c


We considered two types of Monsoon_Rice cultivation: rainfed Monsoon_Rice when Monsoon_ Rice is not followed by any crop or is preceded by Jute, and irrigated Monsoon_Rice when it is followed by a crop which is irrigated: if there is a provision for irrigating winter or summer crop, we supposed this same facility is used to partly irrigate the Monsoon_Rice. We assessed the irrigation requirement of the crops Summer_Rice, Potato and Irrigated Monsoon_Rice in Table 12. We assumed that the requirement for Pre-monsoon_Rice is half of Summer_Rice and that of Wheat is the same as Potato or Irrigated Monsoon_Rice. Using the irrigation frequencies and crops growing period noted during eld works, we derived in Table 10 the irrigation requirement for each crop identied in the landuse map (Figure 25). Figure 28 shows the scaling of the crops with respect to irrigation requirement. After assessing irrigation requirement, we had to distribute the source of irrigation water, whether it is from diversion of the river or groundwater pumping. We used the values from DPDWB (2005). As (i) these gures for source of irrigation water are for year 2004/5, (ii) assessing areas under surface irrigation is perhaps more precise than those under groundwater and (iii) total irrigated area is expected to vary from year to year, we only considered the areas for surface sources (canal, tank, river lift), agglomerated them and assumed that the total area irrigated from surface sources did not vary since 2004/5. The spatially proportionate values calculated per sub-watersheds (Table 13) are indicative as the areas actually considered in SWAT depend on the landuse map (Figure 25) and the spatial discretisation in SWAT (cf. forthcoming section IV.1.c.). As Summer_Rice is an irrigated crop, the variation of its area provides an indication of irrigation trends in the watershed. The last 10 years agricultural statistics show a trend to increase and this change in irrigated area is supposed to be due to the variation in number of shallow tubewells. From now we only used this indicative dataset for assessing the

Table 12: Estimated irrigation per crop. Sources of data: groundwater pumping duration from Mukherji (2007) and diesel pump discharge from TERI (2007) Crop

Groundwater pumping duration (hr/ bigaa)

Diesel pump discharge (m3/hr)

Total groundwater irrigation (m3/ha)

Total groundwater irrigation (mm)

Summer_Rice

55

30

12,334

1,233

Potato

11

30

2,467

247


11

30

2,467

247

a 1 biga ~ 0.134 ha

Table 13:(2005) Indicative areas irrigated from surface sources in each sub-watershed, derived from DPDWB for year 2004/5 Sub-watershed Area (ha)

Sub-watershed Area (ha)

38

1

2

3

4

5

6

7

8

9

5,030

1,531

482

6,187

3,532

1,105

1,789

1,176

1,551

10

11

12

13

14

15

16

17

18

1,481

282

1,991

3,224

771

1,451

55

721

1,466


irrigated areas under surface water source, the remaining irrigated area assumed to be irrigated from groundwater (cf. section IV.1.c.). Table 16 summarises the gures eventually considered in SWAT. Table 13 was used with Table 10 while dening SWAT’s management tables (forthcoming sec tion IV.1.b.).

39


4

MODELLING SET UP

4.1

Initial setting of SWAT

4.1.a

Generation of the HRUs

The Hydrologic Response Units (HRUs) were generated with the ArcSWAT interface in two stages. In the rst stage, ArcSWAT intersects the GIS layers of the sub-watersheds, landuse, soil and slopeclasses. The rst GIS layers were those were presented above, in particular the landuse of Figure 25. The map of slope classes is created by ArcSWAT after the user enters the desired slope classes. Referring to the three typical topographical regions of the watershed and the histogram of Figure 4, the following 3 classes were selected: •

[0 – 3 per cent] or [0 – 1.7°],

•

[3 – 43 per cent] or [1.7 – 23.3°],

•

and greater than 43 per cent or 23.3°.

As this rst step generates a large number of geographical entities (Table 14), the second step aims to remove small units for computational efciency. The user enters a threshold for landuse, soil and slope-classes and the simplication is done per sub-watershed: if in a given sub-watershed the coverage of a unit of landuse, soil or slope-classes is less than the associated threshold (for landuse, soil or slope-classes), ArcSWAT ignores the geometrical features generated with this unit in the rst step in the considered sub-watershed. As explained by Romanowicz et al (2005), a disadvantage of this simplication is the loss of spatial information which are: •

small in extent but has a singular characteristic (i.e., Towns),

•

or dispersed in the watershed (i.e., Monsoon_Rice → [Pre-monsoon_Rice], Monsoon_Rice → [Winter Crop] → Summer_Rice).

Moreover, spatial information is lost in this second step as ArcSWAT may agglomerate spatially distinct entities which have the same combination of Sub-watershed/landuse/soil/slope-classes. One may consequently question the usefulness of this second step but Romanowicz et al (2005) and Geza and McCray (2008) report that the ability of SWAT to reproduces observed signals is not at best if all the units generated in the rst step are kept.

Table 14: HRUs generation stages. First stage Number of

Second stage Threshold

units

Landuse

Soil

Slopeclasses

970

5%

10%

10%

Exemptedlanduseclasses

of HRUs Towns(URMD) Monsoon_Rice [Pre-monsoon_Rice] Monsoon_Rice [Winter Crop] Summer_Rice →

→

→

40

Number

293


Since the landuse map spatial precision is greater than the soil map, we chose the following thresholds: •

10 per cent for soil and slope-classes,

•

5 per cent for landuse. The landuse categories Towns (VIFA), Monsoon_Rice → [Pre-monsoon_Rice] (AAAJ) and Mon-

soon_Rice → [Winter Crop] → Summer_Rice (AWJJ) were exempted from the threshold analysis. This second step generated 293 HRUs (Table 14). The number of these HRUs is greater upstream, where the elevated terrain is more contrasted compared to the plain which is more uniform.

Distribution of the cropping sequences among the agricultural HRUs Once the HRUs have been generated, a fundamental task is to set the management practices of the two agricultural landuse units. As presented above in section III.7.c., the two agricultural units •

Monsoon_Rice → [Pre-monsoon_Rice] (AAAJ),

•

Monsoon_Rice → [Winter Crop] → [Jute or Pre-monsoon_Rice] (AWJJ),

have to be differentiated with respect to the various cropping sequences occurring in the watershed. For this purpose, the cropping sequences were distributed among the HRUs pertaining to the landuse category AAAJ or AWJJ so as to reproduce as much as possible the percentages of Table 11. While doing so: •

cropping sequences Irrigated Monsoon_Rice – Pre-monsoon_Rice and Monsoon_Rice → Winter Crop → Jute were distributed in priority within the unit AAAJ and AWJJ respectively to concur with the landuse map,

•

attention was given to the value of the HRU slope so that Rainfed Monsoon_Rice would rather be grown on uneven HRU while the other cropping sequences would be cultivated on at HRUs.

The cropping sequence of the category Monsoon_Rice –> [Winter Crop] –> Summer_Rice (AWBJ) was chosen to be Monsoon_Rice followed by Summer_Rice. The distribution of all the landuse categories presented in section III.6.d. among the HRUs and the management operations are summarised in Table 15: this is the input landuse information considered by SWAT after pre-processing of ArcSWAT’s interface. The landuse information read as input by ArcSWAT (Table 7) is slightly different from the output of the pre-processing (Table 15), which is due to the HRUs generation routine.

41


Table 15: Landuse distribution considered in SWAT after pre-processing by ArcSWAT, with respect to the discretisation in HRUs, and management operations for each category. Landuse category

Area (km²)

Management operations (table mgt2) Date harIrrigaDate planting vesting tion or or Beginning End growgrowing ing season

(%)

Fertilisation

1,442

24.9

season 1stMarch

Tea or Light Forest (TEAB)

438

7.6

1stMarch

30thNovember

Auto-irri- Auto-fergation tilisation

Small Trees or Shrubland or Settlement (FRMJ)

805

13.9

1stJanuary

31thDecember

None


1,002

17.3

As per Table 10

As per Table 10

As per Auto-ferTable 10 tilisation

Irrigated Monsoon_ Rice – Pre-monsoon_ Rice

173

3.0

AsperTable 10

As per Table 10


Irrigated Monsoon_ Rice – Wheat

53

0.9

AsperTable 10

As per Table 10


Irrigated Monsoon_ Rice – Potato – Jute

825

14.2

AsperTable 10

As per Table 10


Irrigated Monsoon_ Rice – Summer_Rice

342

5.9

AsperTable 10

As per Table 10


Forest(FRSJ)

Village or Fallow (VIFA) Town(URMD)

232 8

4.0 0.1

30thNovember

None

None

Modelled as bare soil Modelledasurbanzones

Water (WATR)

475

8.2

Modelled as Water

Total

5,795

100.0

-

4.1.c

None

Denition of the cropping sequences in SWAT’s management table

Once the cropping sequences have been distributed among all the agricultural HRUs, their schedule in planting, irrigation, fertilisation and harvest were detailed in the management table (table mgt2), following what was summarised in Table 10. The PHU values compiled in Table A.2 were entered for each crop. For trees and tea, as (i) temperature measurements were only available at two stations (Cooch Behar and Jalpaiguri) located in the plain and (ii) Forests and Tea plantations located in hilly / mountainous zones experience low temperatures during winter, these two perennial categories were set to hibernate during months with lowest minimum temperatures (Figure 11), i.e., from beginning of December to end of February. The water source for irrigation (i.e., river or groundwater) tried to reproduce as much as possible the indicative Table 13 for river source while the remaining area was under groundwater irrigation (table mgt1). The Table 16 summarises the consequent irrigated areas and amount per sub-watershed. These gures were partly derived from the landuse map, hence they are representative of the year 2008. 42


Table 16: Estimation of irrigation areas and amount for 2008, with respect to the discretisation in HRUs. Subwatershed

(ha)

Fromsurface (mm/ (% of year) total)

Fromgroundwater (ha) (mm/ (% of year) total)

Total (ha) (mm/ year)

1

3,364

15

100

0

0

0

3,364

15

2

1,098

22

100

0

0

0

1,098

22

3 4

1,170 6,187

146 36

100 21

0 12,786

0 120

0 72

1,170 18,973

146 167

5

3,532

64

40

5,456

96

60

8,988

160

6

1,105

72

23

4,926

244

77

6,031

315

7

3,201

73

32

6,430

154

68

9,631

226

8

1,176

93

29

6,415

230

71

7,591

323

9

1,551

52

34

6,313

100

66

7,864

152

10

1,481

61

43

4,935

80

57

6,416

141

11

282

34

14

1,385

204

86

1,667

239

12

1,991

41

12

13,719

293

88

15,710

334

13

3,224

75

27

10,274

199

73

13,498

274

14

771

0

0

10,629

475

100

11,400

475

15

1,451

0

0

4,970

141

100

6,421

141

16

55

0

0

536

280

100

591

280

17

721

0

0

10,255

623

100

10,976

623

18

1,466

80

45

5,569

99

55

7,035

179

Watershed total / weighted equivalent

33,826

42

22

104,598

145

78

138,424

187

Surface water irrigation is predominant in upstream sub-watersheds and the contrary for groundwater irrigation. Irrigated areas and amount is higher from groundwater source than river. With SWAT’s spatial representation, i.e., with respect to the discretisation with HRUs, the irrigated areas and amount representative of the whole watershed are 138,424 ha and 187 mm / year respectively. As much as 78 per cent of the irrigation (145 mm / year) comes from groundwater and the remaining 22 per cent (42 mm / year) are obtained by river diversion. This amount is very small when compared to the rainfall (Figure 10) and the terms of the water budget, as shown in de Condappa et al. (2011). The sub-watershed 17 has a very high amount of groundwater irrigation as this sub-watershed is located downstream on the west border of the watershed (Figure 6 and Figure 25), where lies most of the area under Irrigated Monsoon_Rice – Summer_Rice. It will be explained in section IV.2.a. that the calibration period was 1998 – 2008. During this period the amount of irrigation was generally increasing, reaching the values compiled in Table 16. Viewing the variation in Summer_Rice area during period 1998 – 2008, which is a crop totally irrigated (Figure 43


28), we assumed that (i) in 1998 the amount of irrigation was 50 per cent of the quantum in 2008 and (ii) irrigation increased linearly between the two dates. This variation was entered in the groundwater model. Considering an inter-annual increase of irrigation is however cumbersome in SWAT’s management table (mgt2), hence we assumed that the irrigation was constant during the calibration period and equal to 75 per cent of values compiled in Table 16. For tea, auto-irrigation was chosen. Without specic data on fertilisation, auto-fertilisation was applied to all the crops and Tea. The correct simulation with SWAT of a paddy eld with impounded water requires that the con cerned HRU is declared as being a pothole (S. L. Neitsch et al. 2005). Unfortunately, only one HRU per sub-watershed can be declared as being a pothole in the version SWAT 2009 used in this work. This would mean that only one HRU per sub-watershed can correctly model rice cultivation, which would not be acceptable in our case as all the agricultural units have a rice cultivation. Hence, we did not declare any pothole, which entailed that rice elds were not simulated as being impounded. This is a limitation of this modelling work as estimations of rice yields and water budget terms (evapotranspiration, groundwater recharge, surface and sub-surface runoff) may be incorrect in agricultural HRUs.

4.1.d

Initial values

To initiate the calibration, initial values were entered for parameters in SWAT which were not determined in preceding sections (Table 17). The values of CN2 were chosen as advised by Neitsch et al. (2010) with respect of the soil hydrologic group (Table A.1, Annex) and the landuse category. The groundwater parameters were guessed considering typical characteristics of alluvial shallow aquifer. Finally, the hydraulic conductivity of the main and tributary channels were taken equal to 0 mm/h, as suggested by Neitsch et al. (2010) for perennial rivers. Regarding the groundwater model, the specic yield Sy is required but values measured in the Jaldhaka basin were not available. Instead, we followed the range of values mentioned by Chatterjee and Purohit (2009) for whole India. Initial value of specic yield was about 0.15 in the in the alluvial system of the plains and less than 0.1 in mountainous sub-watersheds.

Table 17: Initial values for undetermined SWAT’s parameters. Category Management

Parameter CN2

Groundwater

SHALLST

3,000 mm

DEEPST

3,000 mm

GW_DELAY ALPHA_BF GWQMIN GW_REVAP REVAPMN RCHRG_DP Channels

CH_K(1) CH_K(2)

44

Initiavl alue AsadvisedbyNeitschetal.(2010)

10 days 0.5 2,500 mm 0.1 300 0 mm/h 0 mm/h 0


Most of the parameters of Table 17 were adjusted during the calibration.

4.2

Calibration of the groundwater model and SWAT

4.2.a

Method

It is reminded that the modelling set-up is illustrated on Figure 2. The general calibration strategy was to reproduce (1) the evapotranspiration (section IV.2.b.), (2) the measured groundwater levels with the groundwater model so as to assess the groundwater recharge and subsequently reproduce this signal with SWAT (section IV.2.c.), (3) the same as step 2 with groundwater baseow (section IV.2.d.) and (4) the streamow (section IV.2.e.). A last stage, which was not part of the calibration process, attempted to validate the crop yields simulated by SWAT (section IV.2.f.). The procedure followed these steps but as calibrating on one variable may change the result of the previously calibrated variable (e.g., tuning the soil parameters to calibrate the groundwater recharge may change the evapotranspiration as compared to the values calculated in the previous calibration step), the procedure was completed several time until variation in all the four uxes (evapotranspiration, groundwater recharge and baseow and streamow) was negligible. In particular, any transformations in steps 1 to 3 impacted the streamow at the outlet of the watershed (Kurigram gauge station) as this ow is an integrated signal of the whole watershed. A total of 100 manual calibration runs were necessary and following sections will only present results of the initial run, using parameters dened above in section IV.1.d., and of the nal calibration (calibration run n°100). The Table A.4 (Annex) summarises the calibration steps. The modelling period of each calibration run was imposed by the availability of observed streamow, i.e., 1998 to 2008. We calibrated on an average monthly time step, except for the streamows as time-series of daily observed ow was available. To stabilise the water budget and avoid effects of initial conditions, each calibration run was initiated 4 years earlier, i.e., from 1994 to 2008 (climate data were available since 1988); the outputs of period 1994 to 1997 were ignored. The irrigation, as explained in section IV.1.c. above, was equal to 75 per cent of values compiled in Table 16 during this calibration phase. The groundwater model was ran on the 33 wells from SWID within the watershed and the outputs were then extrapolated to the sub-watersheds.

4.2.b

Evapotranspiration – Average monthly time step

We started to check that the calculated reference evapotranspiration as dened by Allen et al. (1998) and referred to as ET0 [L/T], was satisfactory. SWAT can calculate ET0 with three formulas: PriestleyTaylor, Penman-Monteith or Hargreaves. The three options were selected and their outputs were compared with values reported in Kundu and Soppe (2002) and Raghuwanshi et al. (2007) (Figure 29 and Table 18). First it can be noticed that Kundu and Soppe (2002) and Raghuwanshi et al. (2007) provide similar trends. Hargreaves over-estimated ET0, Priestley-Taylor on the contrary under-estimated it. Although solar radiation data from the region was not provided as input climatic data, Penman-Monteith formula yielded the closest values to the two references and was therefore selected to calculate ET0. We then examined values calculated for the actual evapotranspiration, referred to as ETa [L/T]. More precisely, we inspected for each vegetative landuse category the values taken by the ratio and checked that its monthly average value took sensible values (Figure 30 and Table 19). The initial values for the landuse category Forest (FRSJ), an important category for the watershed, were judged to be too small, hence we aimed at increasing the values taken by this ratio by tuning some soils parameters. 45


180

Kundu and Soppe (2002) Raghuwanshi et al. (2007) SWAT PenmanMonteith SWAT Hargreaves SWAT PriestleyTaylor

160 140 120

) th n o 100 m / m m ( 80 0 T E 60

40 20 0

1

2

3

4

5

6

7

8

9

101

11

2

Month

Figure 29: Average monthly reference evapotranspiration calculated from difference sources

Table 18: Average annual reference evapotraspiration calculated from different sources. Source

Kundu and Soppe (2002) Raghuwanshietal.(2007)

ET(0 mm/year)

1,360 1,284

SWAT Penman-Monteith SWAT Hargreaves SWAT Priestley-Taylor

1,296 1,476 1,163

With the aim of increasing ETa, sensitivity analysis showed that ETa calculated by SWAT is mainly sensitive to soils’ AWC, ESCO, EPCO and to a less extent soil thickness (Sol_Z). ETa increases with the parameter EPCO but this parameter was already equal to 1, its maximum value in the initial run. Therefore in this calibration we mainly increased the AWC and Sol_Z (Table A.4, Annex), in coordination with next calibration step (section IV.2.c.) as these two soil parameters also inuence on SWAT calculation of groundwater recharge. It was not required to tune ESCO parameter.

4.2.c

Groundwater recharge – Average monthly time step

This stage saw an interaction between the groundwater model and SWAT so as to model the groundwater recharge and baseow. Indeed, the groundwater model interpreted the groundwater levels, rainfall and streamows during the low ow season to provide an indication of the magnitude of the recharge and baseow. At the same time SWAT gave an indication of the spatial and temporal variation of this recharge, with respect to properties of the overlaying soil cover. The aim was that both models converge to similar outputs.

46


FRSJ Aman Aman – Winter crop – Jute

FRMJ Aman – Aus Aman – Boro

TEAB Aman – Wheat

FRSJ Aman Aman – Winter crop – Jute

1.0

1.0

0.9

0.9

0.8

0.8

0.7

0.7

0 0.6 T E / 0.5 a T 0.4

0 0.6 T E / 0.5 a T 0.4

E

E

0.3

0.3

0.2

0.2

0.1 0.0

FRMJ Aman – Aus Aman – Boro

TEAB Aman – Wheat

5

8

0.1 1

2

3

4

5

6

7

8

9

101

11

2

0.0

1

2

3

4

Month

6

7

9

101

11

2

Month

Figure 30: Calibration with respect to the actual evapotranspiration ETa. Monthly value of the different landuse vegetation categories (average over the calibration period, 1998 – 2008). Left: initial run. Right: after calibration (calibration run n°100). Aman: monsoon rice, Boro: summer rice, Aus: premonsoon rice.

Table 19: Calibration with respect to the evapotranspiration ETa. Annual values of the ratio ETa / ET0 for the different landuse vegetation categories (average over the calibration period, 1998 – 2008). Aman: monsoon rice, Boro: summer rice, Aus: pre-monsoon rice. Landuse category

FRSJ

FRMJ

TEAB Aman Aman – Aus

0.69

0.56

Aman – Wheat

Aman – Potato –

Aman Whole – Boro water-

Jute Initialrun Calibrated (calibration run n°100)

0.51 0.66

0.72

0.54

0.55 0.52

0.58 0.57

0.66 0.67

0.78 0.76

shed 0.70 0.66

0.61 0.64

The groundwater model was ran on the 33 wells from SWID located within the Jaldhaka watershed. It is reminded that the groundwater draft Dnet (cf. section II.2.) for irrigation increased linearly from 50 per cent to 100 per cent of the quantum of 2008 between the years 1998 to 2008 (cf. section III.8.c.). At each observation well, the groundwater model utilised the Kalman lter (Kalman 1960) to t the Eq. (4) and (8) to the observed groundwater levels, using the rainfall P, the groundwater draft Dnet and the irrigation I which are known (Dnet and I are different if irrigation is also provided by a river source, in addition to groundwater). This estimated the total recharge RG, the baseow B and the groundwater level h at each of the 33 wells. The simulated levels compared with measurements are shown on Figure 31 for wells located in different topographical positions. It is noteworthy to remind that the groundwater model was required to reproduce the observed groundwater levels, which was not possible with the version of SWAT used in this work (SWAT 2009). At well D-12, the model simulated groundwater level surfacing at some time step. Simulations suggest that the measurements may have missed moments when groundwater levels were the deepest or the shallowest.

The recharge representative of each sub-watershed was then calculated by averaging the values of wells within the sub-watershed. Since the observed groundwater levels were not available at a monthly time step (cf. section III.5.b.) and the groundwater model used in this work is lumped, i.e., a simpler approach than the semi-distributed of SWAT, we aimed at reproducing with SWAT 47


Date 01/96 182 180

10/98

Date

07/01

04/04

01/07

10/98

04/04

01/07 Soil level

100

) m ( l e v le c i tr e m o z ie P

Observed

98 96 Simulated 94

Observed

92 90 88

168

86

166

Inpiedmont(wellD-10)

Inpiedmont(wellD-12)

Date 01/96 71

10/98

07/01

Date 04/04

01/07 Soil level

69

) (m l 67 e v le 65 c ir t e 63 m o z 61 ie P

07/01

Soil level Simulated

) 178 (m l e v 176 e l c rit 174 e m172 o z ie P 170

01/96 102

Simulated Observed

01/96 36

) 34 (m l e 32 v le 30 c ir t e 28 m o z ie 26 P

59

24

57

22

55

10/98

07/01

04/04

01/07

Soil level

Simulated Observed

20

Inplain(wellP-25)

Inplain(wellPTC-9)

Figure 31: Piezometric levels simulated at a monthly time-step by the groundwater model vs. observations The wells are located on Figure 16

the simulated groundwater recharge at an average monthly time step over the calibration period (1998 – 2008), instead of a monthly time step. Eventually the groundwater model enabled to translate the variation in groundwater levels into recharge, which is reproducible with SWAT. The groundwater recharge of the shallow aquifer calculated by SWAT (GW_RCHG) was sensible to GW_DELAYS, the surface curve number CN2, Sol_Z, to a lesser extent AWC and Sol_K where slope is important. Initially the recharge calculated by SWAT was too small in mountainous sub-watersheds while too high in the plains as compared to the estimation from the groundwater model. Overall SWAT’s recharge was too hight (Figure 32). We increased CN2 to reduce inltra tion at the soil surface, decreased Sol_K in mountainous region to reduce sub-surface later ow and controlled the increase in Sol_Z in the previous calibration step (section IV.2.b.). We also xed GW_DELAYS (the time that percolation out of the soil cover becomes shallow groundwa ter recharge) equal to 0 as the groundwater levels are shallow.

48


300 ) th n 250 o m / m 200 m ( e g r a 150 h c e r r fe i 100 u q a w o ll 50 a h S

0

300 Groundwater model SWAT

Average annual recharge: * Groundwater model: 409 mm / year * SWAT: 967 mm / year

1

2

3

4

5

6

7

8

9

10

11

) th n o 250 m / m 200 m ( e rg a 150 h c re r e fi 100 u q a w o ll 50 a h S

12

0

Average annual recharge: * Groundwater model: 569 mm / year * SWAT: 571 mm / year

1

2

3

Month

4

5

6

Groundwater model SWAT

7

8

9

10

11

12

Month

Figure 32: Calibration with respect to the recharge of the shallow aquifer (GW_RCHG), average for the Jaldhaka watershed over the calibration period (1998 – 2008) Left: initial run. Right: after calibration (calibration run n°100)

The calibrated shallow aquifer recharge (GW_RCHG) is plotted on Figure 32. Note that output of the groundwater model also changed during the numerous calibration runs as this is an interactive and iterative process. The annual sum calculated by SWAT is very close to the value simulated by the groundwater model (569 mm/year vs. 571 mm/year) but the monthly gures are different. This is due to the different approach that both models follow to compute the recharge from rainfall and irrigation: • the groundwater model uses a linear relationship, • while SWAT calculates the recharge with non-linear rules function of soils properties. Therefore we focused on reproducing the annual amount and optionally match as much as possible the monthly trends. Calibrated values of CN2 are very high. Its maximum value is 95 for the agricultural unit Monsoon_Rice – Summer_Rice: in the absence of any pothole, this value was purposely chosen to simulate the rather impermeable soils of this agricultural practice. Others agricultural units in soils of the plain were given the value 90, which is also a high value, which is consistent with the low conductivity reported by Kundu and Soppe (2002) for soils of the region and the fact that values reported in Neitsch et al. (2010) are appropriate for a 5 per cent (a little less than 3°) slope, while the slope is mostly less than 1° in the plain (Figure 4).

4.2.d

Groundwater baseow – Average monthly time step

Similarly to the groundwater recharge, this calibration step aimed at converging the simulations by both models of the groundwater baseow. Using the simulations of the groundwater model, the baseow representative of each sub-watershed was calculated by averaging the values of wells within the sub-watershed. Eventually the groundwater model enabled to translate the varia tion in groundwater levels into baseow that can be in turn reproduced with SWAT. As was the case for the recharge, the calibration in SWAT was also realised on an average monthly time step over the calibration period (1998 – 2008), instead of a monthly time step. The groundwater baseow of the shallow aquifer calculated by SWAT (GW_Q) was sensible to the 49


baseow recession constant (ALPHA_BF), the deep aquifer percolation fraction (RCHRG_DP) and the shallow aquifer threshold for baseow (GWQMIN). The observed streamow at the out let of the watershed (Kurigram gauge, cf. Figure 7) and the groundwater model demonstrate that the groundwater baseow is buffered all along the year and in particular baseow occurs during the relative dry season. This buffered groundwater baseow was absent in the rst run (Figure 33) and the solution was to decrease drastically ALPHA_BF and play with GWQMIN. This was in coordination with next calibration step (section IV.2.e.) as these two groundwater parameters also inuence on SWAT calculation of streamow. For simplication purpose, RCHRG_DP was taken equal to 0. Indeed, processes of the deep aquifer in such alluvial context were assumed to be governed by lateral transfers between regions outside of the Jaldhaka watershed.

) 90 h t n o 80 m / 70 m m ( r 60 fe i u 50 q a w o 40 ll a h s 30 m o rf 20 w o fl 10 e s a B 0


Average annual baseflow: * Groundwater model: 427 mm / year * SWAT: 313 mm / year

1

2

3

4

5

6

7

8

9

10

11

12

) 90 h t n o 80 m / 70 m m ( r 60 e fi u q 50 a w o 40 ll a h s 30 m ro f 20 w lfo 10 e s a B 0

Average annual baseflow: * Groundwater model: 421 mm / year * SWAT: 418 mm / year

1

Month

2

3

4

5

6


7

8

9

10

11

12

Month

Figure 33: Calibration with respect to the shallow groundwater baseflow (GW_Q), average for the Jaldhaka watershed over the calibration period (1998 – 2008) Left: initial run. Right: after calibration (calibration run n°100)

In the nal calibrated run, the matching is greatly improved (Figure 33). Note that output of the groundwater model again changed during the numerous calibration runs. The nal value of ALPHA_BF (0.002, cf Table A.4, Annex) is extremely small with respect to values presented in Neitsch et al. (2010), hence this watershed setting of Jaldhaka may be an extreme case reaching the limitation of SWAT modelling. Having this in mind and similarly to the case of groundwater recharge, we focused on reproducing the annual amount and optionally match as much as possible the monthly trends. The calibrated value of shallow groundwater baseow is less than the recharge: this is logical since the modelling accounted for irrigation groundwater pumping.

4.2.e Streamows – Daily time step The critical step of the calibration was to reproduce the observed streamows at the stations Taluk-Simulbari and Kurigram in Bangaldesh. As it could be expected, there was a gap between modelled and observed values in the initial run (Figure 34). More preciselyi)( simulated ows were too low during the low season, as the groundwater baseow was greatly underestimated during this period (cf. section IV.2.d.) and ii() the signal of the streamow was too sharp hence it was required to delay the ows, i.e., the runoff. 50


1,600

Simulated Observed 1,400 5,000

Simulated 4,500

1,200

Observed

4,000

) /s 1,000 3 (m

3,500

) /s 3,000 3 (m

w 800 lo f m a e 600 rt S

w 2,500 lfo m 2,000 a e rt S 1,500

400

1,000

200 500 0 06/98

0 10/99

03/01

07/02

12/03

04/05

08/06

01/08

1

2

3

4

5

6

7

8

9 1 0 11 12

Month

Figure 34: Streamflow simulated (FLOW_OUT) at Kurigram in the initial run over the calibration period (1998 – 2008) Left: daily time-series. Right: monthly averages

Since the streamow at Kurigram and Taluk-Simulbari is an integrated signal representative of the areas upstream, the simulated streamow (FLOW_OUT) was sensitive to all the parameters modied in the previous calibration steps (sections IV.2.b. to IV.2.d.) plus the values of Man ning’s roughness coefcient for the channels (CH_N(1) and CH_N(2)) and the surface lag coef cient (SURLAG). Only CH_N(1) and CH_N(2) were modied in this section to introduce lag in the runoff and it was not required to tune SURLAG. To quantitatively assess the quality of the calibration, we used four metric indicators. The rst checked that there is no bias in the modelling by making sure that the magnitude of the simulations over the calibration period is close to the values measured. For this purpose we calculated the following bias indicatorM [-]: (10)

with QSWATi [L3/T] and QObsi [L3/T] respectively the simulated and observed on day i; the sum is calculated over the calibration period (1998 – 2008). The second indicator was the Nash and Sutcliffe (1970) efciency NS [-]:

(11)

with

the average of QObsi over the calibration period (1998 – 2008).

The third indicator was a modied version of NS to emphasise on low ows NSlow [-] (Krause, Boyle, and Bäse 2005):

51


(12)

The forth and last indicator was a modied version of NS to emphasise on high ows Nshigh [-]:

(13)

.

The aim of the calibration was quantitatively to have all the 4 indicators closest to 1 and qualitatively to improve the visual t. We paid greater attention to values of NSlow, in particular at TalukSimulbari, as measurement of high ows are suspected to be less precise than low ows and number of readings is more at Taluk-Simulbari (section III.2.b.). To achieve this, we increased CH_N(1) and CH_N(2) to create roughness, hence lag, along respectively the tributaries and the main channel. The nal value of CH_N(1) and CH_N(2) (respectively 0.5 and 0.3, cf Table A.4, Annex) are high (S. Neitsch et al. 2010) and can be explained by the particular conguration of the watershed middle and downstream which is extremely at with various depressions. Values of the 4 indicators at the initial run and nal calibration (calibration run n°100) are showed in Table 20 and the calibrated hydrographs are shown on Figure 35. Unfortunately values of the 4 indicators are not available for Taluk-Simulbari for the initial run. The value of M for Taluk-Simulbari is very close to 1, which shows that there is no bias for the simulation down to this station. The values of NS, NShigh and NSlow are lower for Taluk-Simulbari than for Kurigram which is due to the much greater number of available observations at Taluk-Simulbari. Overall values of these three parameters are judged to be satisfactory at both gauge stations. The average ows during the low season are apparently slightly over-estimated at both stations but this nal calibration conguration yielded the values ofM, NS, NShigh and NSlow closest to 1. Table 20: Values of the calibration indicators dened by Eq. (10) to (13).

Station

Sub-water- Number shed of observations

TalukSimulbari

16

878

Kurigram

18

197

Initialrun

Calibrated (calibration run n°100)

M

NS

NSlow

NShigh

M

NS

NSlow

NShigh

NA

NA

NA

NA

1.02

0.74

0.77

0.75

Incalculable

0.48

0.95

0.78

0.82

0.77

0.80 0.28

The gap between simulation and observation is greater for high ows, in particular at Kurigram where M takes the low value of 0.95. This is acceptable referring to section III.2.c.: • the greater the ow, the greater the error in measurement; • oods from other river systems may invade the most downstream part of the watershed hence create ow-peaks or higher magnitude; this cannot be modelled with the current modelling set-up.

52


4,000

4,000

Simulated 3,500

Simulated Observed

Taluk-Simulbari 3,500

Observed

3,000

3,000

) s / 3 2,500 m ( w 2,000 o lf m a 1,500 e rt S 1,000

) /s 3 2,500 (m

500

500

Kurigram

w 2,000 o lf m a 1,500 e rt S

1,000

0 06/98 10/99 03/01 07/02 12/03 04/05 08/06 01/08

0 06/98 10/99 03/01 07/02 12/03 04/05 08/06 01/08

1,600

1,600

Simulated 1,400

Taluk-Simulbari 1,400

Observed

1,200

1,200

) /s 3 1,000 m ( w 800 lfo m a 600 rte S 400

) /s 3 1,000 (m w 800 o fl m a 600 e tr S

200

200

0

Kurigram

Simulated Observed

400

0 1 2 3 4 5 6 7 8 9 1 0 11 12

1 2 3 4 5 6 7 8 9 1 0 11 12

Month

Month

Figure 35: Streamflow simulated (FLOW_OUT) in the final calibration (calibration run n°100) over the calibration period (1998 – 2008). Top: daily time-series. Bottom: monthly averages.

4.2.f

Crop yields – Annual time step

This stage did not involve any calibration. We attempted instead to validate the dry crop yields simulated by SWAT in each agricultural HRU by comparing them to the administrative agricultural statistics. The average yields (period 1998 – 2008) simulated for each cropping sequence dened in Table 15 and for the whole watershed are placed in Table 21; the average yield of the Monsoon_Rice from all the cropping sequences is mentioned as well.

Table 21: Watershed-average dry crop yields simulated by SWAT in the final calibration (calibration run n°100) over the calibration period (1998 – 2008). Aman: monsoon rice, Boro: summer rice, Aus: pre-monsoon rice.

Cropping sequence

Aman

Aman – Aman – Aman – Potato – Aman – All Aus Wheat Jute Boro Aman Aman Aus Aman Wheat Aman Potato Jute Aman Boro Aman

Dryyield(T/ha)

2.7

2.2

2.0

2.2

3.6

1.8

3.5

2.6

2.6

2.7

2.3

53


When compared with the agricultural statistics (Table 9), those values do not match well for at least four reasons. First, the crop parameters presented in section III.7.c. for Rice, Wheat and Potato are not specic for the region of the Jaldhaka watershed. Moreover, Jute was not present in SWAT’s database and no agronomic information was available on this crop, hence its characteristics were arbitrarily derived from SWAT’s generic agricultural class. Second, yields can be calculated in different ways. SWAT estimates the dry yield while the agricultural statistics mention: • for rice, the clean yield, which does not match with SWAT’s results, and the dry yield, to which SWAT’s estimation are closer, • for potato, the wet yields as potato contains a signicant amount of water after harvest; if the water content is assumed to be about 80 per cent, hence the dry matter is 1/5th of the total potato mass, SWAT’s yield is closer to agricultural statistics.

Third, the management practices in SWAT considered auto-fertilisation, which results in greater application of fertilisers as compared to reality. This may entail an over-estimation of the yields. Fourth, as was emphasised in section IV.1.c., a correct simulation of paddy elds by SWAT would require to use thepothole characteristic, which was not possible in this application in the Jaldhaka watershed. Hence, in SWAT rice elds were simulated as regular elds which are not impounded. This should obviously lead to wrong yield calculations, in particular for the Summer_Rice and Pre-Monsoon_Rice which are irrigated, hence impounded. This limitation may not impact as much the yield simulation of rainfed Monsoon_Rice. All these reasons imply that the yields simulated by SWAT are not reliable, although calculated values are close to the administrative statistics for clean rice and potato. Analyses of modelling output should not utilise the crop yields simulated by SWAT.

4.2.g

Simplications and limitations of the modelling

The Table 22 summarises the simplications and limitations of the modelling. As the dataset was limited (no monthly groundwater levels, only the streamow was on a daily time step and this only for 2 gauges) and the model was not validated with an independent dataset, the current version is only valid to show monthly or annual average trends at the scale of the whole watershed. In particular, it should not be used to analyse outputs of individual month, year or sub-watershed. Application of this modelling is reported in de Condappa et al. (2011).

54


Table 22: Simplifications and limitations of the modelling. Category

Limitation

Why?

Consequence

Ponding of paddy field should be modelled in SWAT with a pothole

Ignored

Only one HRU per sub-watershed can be a pothole in SWAT 2009

The hydrological functioning of a paddy field is not modelled correctly, which could affect calculations of: rice yield, water budget terms around the paddy field (evapotranspiration, groundwater recharge, surface and sub-surface runoff).

Domestic & industrial water consumption

Ignored

Nodata

Deep aquifer

Ignored

Deep aquifer processes in such alluvial context were assumed to be governed by lateral transfers between regions outside of the Jaldhaka watershed.

Should not affect the hydrological modelling of the Jaldhaka river network.

Possible bad representation in SWAT of the crops and agricultural practices in the Jaldhaka watershed.

Analysis of the current and scenario contexts ignored SWAT's simulations for crop yields.

Should be negligible as main rain occurs during the warm period.

Crop yield not simulated correctly

Snow accumulation and melting upstream

Ignored

No data on snow accumulation and lapse in temperature

Groundwater pumping for irrigation

Considered constant in SWAT

Cumbersome to enter a varying irrigation in management table (mgt2)

Streamflow

Available only at 2 locations downstream

Data restriction in India

Shouldnotbeanissueasthereis no major city nor industries along the Jaldhaka. Moreover these demands usually consume little water as compared to irrigation.

Not analyses are possible at subwatershed scale but only for the whole watershed.

55


5 5.1

DISCUSSION … on the input dataset

Input data is a critical requirement for a successful modelling and are of two sorts: ( i) data characterising the studied region and (ii) the forcing data to calibrate and validate the model. A fair dataset was gathered in this work and we tried to complement gaps. This enabled a satisfactory modelling of average trends at the watershed scale. There were nonetheless some shortcomings. In term of data characterising the studied region, climatic data, and in particular rainfall, is the primary input data. In the Jaldhaka watershed, the daily rainfall was the most varying variable within a month, followed by wind while the humidity was less variable and the temperature almost stable monthly-wise (Figures 10 and 11). Hence the priority to capture the climatic characteristics of a region is to gather rainfall and wind data at a time and space resolution as ne as possible, while humidity and temperatures could be compiled at a coarser scale. In this work, we collected daily climatic data at two local stations and complemented with a gridded daily rainfall dataset to include a spatial representation of the variability of rainfall. To characterise succinctly watershed, SWAT generatesHydrologic Response Units (HRUs) from the topographical, soil and landuse information. Obtaining topographical data, ie., the Digital Elevation Model (DEM), is usually not an issue for large watersheds / basins, as it was the case in this work with the Shuttle Radar Topography Mission(SRTM). Soil information was the second information required to create the HRUs and describe the soil hydrological processes and vegetation growth cycles. Qualitative information is usually accessible through international database, such as the FAO Digital Soil Map of the World or the World Reference Base, or national agencies, as was the case in this work with the soil map of the Indian National Bureau of Soil Survey and Landuse Planning. A agro-hydrological model like SWAT however requires quantitative soil data usually not as available, such as the thickness of the different horizons, the soil texture (e.g., percentage of clay, silt, sand), and the soil structure (e.g., soil water retention parameter such as the Available Water Capacity in SWAT – AWC -, the soil conductivity parameter such as the empirical curve number CN2 in SWAT) in these horizons. In our case, we did not have specic local data for these quantitative soil parameters and we used instead the free Harmonised World Soil Database. This international database was not matching with the local soil knowledge and literature hence we had to adjust and complement to / with the local knowledge. During SWAT’s calibration, we found that the most sensitive soil parameters were the AWC, the soil thickness and the curve number CN2. One may question the relevance of modifying these three last soil parameters as they should be estimated based on local data, which we try to attain before the calibration. Modifying the soil thickness could indeed be questioned as it is informed by the soil map, although in an approximative manner. However, the AWC and the curve number are much dependent of the soil structure, i.e., the arrangement of the soil particles in the soil medium, which is extremely variable in space and time, thus these two parameters can be considered as calibration parameters. The landuse is the third layer necessary to generate the HRUs. In addition to describing the land occupation, it should inform as much as possible on the cropping patterns in agricultural lands. Our case was ideal as we contributed to the generation of a ne resolution landuse map in the context the AgWater Solutions Project in the Jaldhaka watershed. In case no specic landuse map is accessible, an alternative could be to gather landuse map from international database and try to use available tools, such as Google Earth, to attempt to improve the landuse map. In this work without reservoirs in the watershed and in addition to generate HRUs, crop information and management practices are necessary to describe appropriately the cropping sequences. Crop data pertain to agronomic properties for simulating the growth of vegetation and the crop yields. We lacked the specic characteristics for the crops grown in the Jaldhaka and approximated them with values 56


present in the ArcSWAT crop database. This was one of the causes of our failure to simulate correctly crop yields in the Jaldhaka basin. Finally, knowledge of management practices, such as irrigation and fertilisation applications, are essential as well. Though we collected typical irrigation time-schedules during the groundtruthing of the landuse map developed in AgWater Solution project, we had no systematic and continuous information on the current and past irrigation amount in the watershed. Alternatively, we approximated fairly well these irrigations using the precise landuse map and taking the trivial assumption that farmers growing an irrigated crop (e.g., Summer_Rice) have the mean to irrigate. Subsequently and in a further approximation, we chose for the irrigation amount values reported in the literature for West Bengal regions. Regarding the fertilisation, we possessed no data specic to the Jaldhaka watershed map and instead used the auto-fertilisation function of SWAT. This was not satisfactory and surely contributed to a wrong simulation of crop yields. With respect to forcing data, the two variables considered here were streamows and groundwater levels. Availability of the measured streamows at the outlet of the watershed was critical to calibrate SWAT, which guaranteed a reproduction of the integrated hydrological signature of the watershed. Moreover, the streamows being highly variable within a month, it is advisable to collect daily measurements, which was feasible in this work. We lacked however measurements at intermediary location in the watershed and therefore we could not simulate sub-watershed hydrological processes. The groundwater data was particularly required in the Jaldhaka watershed as groundwater is the predominant source for irrigation and it was therefore important to model it appropriately. In India groundwater levels are more easily obtainable than streamows and, in our case and thanks to the State Water Investigation Directorateof West Bengal, a good dataset of groundwater levels was available free of cost. Although this dataset was limited as it was not monthly, which means that we might have missed the months when groundwater levels are the deepest or the shallowest, it enabled the simulation of groundwater levels uctuation. We missed in our dataset measured values of the specic yield in the Jaldhaka watershed, that we eventually approximated with range of values for whole India.

5.2

… on the model set up

One characteristic of this work is to have associated two different models: SWAT and the groundwater model developed by Tomer et al. (2010). The groundwater model enabled the interpretation of the measured groundwater levels, which was impossible with the version of SWAT used here (SWAT 2009, version 433). Subsequently, the groundwater model translated the variation in groundwater levels into uctuation of groundwater recharge. SWAT also contributed to the determination of the recharge by improving its spatial and temporal variation, with respect to properties of the overlaying soil cover. Eventually both models interacted to simulate correctly the recharge with respect to the rainfall and the groundwater levels. An additional benet of using the groundwater model was to guide SWAT in reproducing the buff ered groundwater baseow, which is essential to simulate the perennial nature of the Jaldhaka river. Indeed, the rst calibration runs with SWAT failed critically to model the baseow during the dry season (Figure 33). The calculations of the groundwater model motivated the reduction of SWAT’s groundwater parameter ALPHA_BF to an extremely small value, which improved drastically the simulation of the streamows. Since the nal value of ALPHA_BF is smaller than the range mentioned by Neitsch et al. (2010), this case is extreme in term of SWAT’s groundwater modelling and the groundwater model was critical to guide towards this singular setting. Application of this modelling is reported in de Condappa et al. (2011).

57


6

CONCLUSION

This paper contributed to the understanding of potential for development of Agricultural Water Management (AWM) in the watershed of the Jaldhaka river, a tributary to the Brahmaputra river, located in Bhutan, India and Bangladesh. An application of the Soil Water Assessment Tool (SWAT) and of the lumped groundwater model of Tomer et al. (2010) was developed as a tool to study AWM development scenarios in the Jaldhaka watershed. The rst stage of this work was to collect data / information to characterise the natural and agricultural contexts of the Jaldhaka watershed. The watershed has a contrasted topography, with mountains upstream and large plains downstream. It experiences high rainfall with a monsoonal pattern and an average of 3,300 mm/year. The river ow is perennial with recurrent occurrence of ood events during the monsoon. The aquifers are alluvial in the region and the groundwater levels are shallow and stable in the watershed. This study contributed to the development of a precise landuse map which identies in particular the different cropping sequences in agricultural lands. Agricultural statistics were gathered at administrative levels and the irrigation in the watershed was found to be predominantly from groundwater, with diesel pumps, and to irrigate rice during summer and potato during winter. A fairly large dataset was gathered and we tried to complement gaps. There were nonetheless some shortcomings. In term of data characterising the studied region, climatic data, and in particular rainfall, are the most required input data. In the Jaldhaka watershed, the daily rainfall and wind were the critical climate data as they were the most varying variable within a month. Soil information is usually available in a qualitative form but quantitative information necessary for agro-hydrological modelling is rarer and in our case we combined local soil knowledge and literature with an international database. The landuse map describes the land occupation and should inform as much as possible on the cropping pattern in agricultural lands. Our case was ideal as we contributed to the generation of a ne resolution landuse map in the context the AgWater Solutions Project in the Jaldhaka watershed. Precise knowledge of irrigation patterns is required to account for anthropogenic water uses and simulate correctly crop cultivations. Though we had no systematic and continuous information on the current and past irrigation in the watershed, we approximated it fairly well using the precise landuse map and irrigation amount values reported in the literature for West Bengal regions. We lacked specic agronomic information on crops and agricultural practices in the Jaldhaka watershed and consequently failed to reproduce correctly crop yields. With respect to calibration data, we used streamows and groundwater levels. Availability of the measured streamow at the outlet of the watershed was critical to calibrate SWAT, which guaranteed a reproduction of the integrated hydrological signature of the watershed but we lacked however measurement at intermediary locations in the watershed and therefore we could not simulate sub-watershed hydrological processes. The groundwater data was particularly required in the Jaldhaka watershed as groundwater is the predominant source for irrigation and it was important to model it appropriately. We gathered a good dataset of groundwater levels and were able to simulate uctuation of groundwater levels. A characteristic feature of this work was to have associated in an interactive manner the model SWAT with the groundwater model of Tomer et al. (2010). The last enabled the interpretation of the measured groundwater levels, which was impossible with SWAT and which was particularly important in the context of the Jaldhaka watershed. It also guided SWAT in reproducing correctly the buffered groundwater baseow, which is critical to simulate the perennial nature of the Jaldhaka river. At the same time, SWAT interpreted the measured streamows and improved the spatial and temporal description of the groundwater recharge. Eventually both models interacted to convergence to a satisfactory simulation of hydrological processes in the Jaldhaka watershed. However, the model set-up failed to reproduce adequately the crop yields. 58


This modelling framework was applied in an accompanying report (de Condappa et al. 2011) to study the current state of the hydrology in the Jaldhaka watershed and the impacts of two types of AWM scenarios.

59


ACKNOWLEDGEMENTS This work was supported by the AgWater Solutions project, funded by the Bill and Melinda Gates Foundation. We are also thankful to the following persons for their contribution (by alphabetical order): • Nyayapati Aakanksh, International Water Management Institute, for supporting request of streamow data. • Badrul Alam, International Development Enterprises - Bangladesh, for supporting request of streamow data. • U. S. Aich, Directorate of Agriculture (Kolkata), for providing agricultural statistics on potatoes. • Saswati Bandyopadhyay, State Water Investigation Directorate (West Bengal), for providing groundwater data. • Gopal Barma, Assistant Director of Agriculture (Administration, Mathabhanga), for providing agricultural statistics. • Salim Bhuiyan, Bangladesh Water Development Board, for providing streamow data. • Suman Biswas, International Development Enterprises India, for support during eld works. • S. Biswas, Agricultural University of Cooch Behar, for providing information on soils. • P. K. Biswas, Assistant Agricultural Meteorologist (Jalpaiguri), for providing a copy of Kundu and Soppe (2002). • Aniruddha Brahmachari, International Development Enterprises India, for support during eld works and data collection. • Annemarieke de Bruin, Stockholm Environment Institute, for sharing results of the Participatory GIS in the Jaldhaka watershed. • Xueliang Cai, International Water Management Institute, for contributing to the development of the landuse map. • Howard Cambridge, Stockholm Environment Institute, for advices onSWAT. • D. Dutta, Indian Space Research Organisation, for general advices. • Sylvain Ferrant, Indo-French Center for Groundwater Research, for numerous advices on SWAT. • Charlotte de Fraiture, International Water Management Institute, for initial advices. • Victor Kongo, Stockholm Environment Institute, for advices on SWAT. • Monique Mikhail, Stockholm Environment Institute, for sharing results of the Participatory GIS in the Jaldhaka watershed.

60


• K. K. Mondal, Director of the Bureau of Applied Economics & Statistics (Kolkata), for providing agricultural statistics. • Aditi Mukherji, International Water Management Institute, for numerous advices on groundwater and data collection. • Rajiv Pradhan, Director of International Development Enterprises - Bangladesh, for supporting request of streamow data. • Mala Ranawake, International Water Management Institute, for supporting data requests. • Adam Regis, Stockholm Environment Institute, for administrative works. • Bhaskar Roy, Assistant Director of Agriculture (Cooch Behar), for providing general information. • Bharat Sharma, International Water Management Institute, for supporting request of streamow data. • Raghuwanshi Narendra Singh, Indian Institute of Technology Kharagpur, for providing data of Raghuwanshi et al. (2007). • A. K. Sinha, Agricultural University of Cooch Behar, for providing information on soils. • Jean Philippe Venot, International Water Management Institute, for support on ArcSWAT interface. • Hua Xie, International Food Policy Research Institute, for advices on SWAT.

61


. g n i n n a l P se u d n a L d n a y e rv u S il o S f o u a re u B l a n o it a N n a i d In e h t

X E N N A 62

m o fr p a m il so l a n i irg o e th m o rf ta a d : e g n a r o t h ig L . e trs e m a r a p li o S : 1 . A le b a T

ta a D li o S ld r o W d e si n o rm H a e th m ro f a t a d : e u l b t h g Li r. a h e B h c o o C r, e icf f O l ra u lt ciu r g A la p ci n ri P e h t m ro f n o it a rm o f n i th i w ta a d p a m li so g in ss o cr y b d ve ri e d rs e t e m ra a p y: e r g t h g Li

8 2 0 W 6 2 0 W

u ll a t n e c e R

ts n e u q a v st p e u q a l u l ift s U

5 2 0 W

ic u q A

8 1 0 W

ts n e h tr ic o t p y sU T

0 1 0 W

u fli ts U

st n e v

l iva u ll a

t n ce e R

l a vi u ll a

llu a t n e c e R st n e v

e ivt c A

8 0 0 W 7 0 0 W

ci t t n - sp ve tro e r lu E F u ch

t n o m d e i P t n o m d ie P t n o m d ie P

st n e u q la

rto ys D st p re h c o rt

ts p re h c

4 0 0 W

icr b m U ys D ci p yT

3 0 0 W

ci r b m U

o rt sy D

st p re ch

sll i H

2 0 0 W

ic yp T

-r o d U

st n e th

llis H

1 0 0 W

ic th i L

iv R y m o n xo a T A D S U

-r o d

sll i H

ts n e h U t

in la p l iva

t n ce e R

ci u q A p a H ci p yT

6 0 0 W

.T A W S in d e r e t n e y ll a tu n ve e ta a d : w o ll ye t h g Li . se a b

u l F ic p yT p a H irc e A

lsil H ld Ja

n ito si o P

in la p l ia v l a vi lu l a i a r e T ( n i a l p n i la p n i la p e id s d n a e isd d n a e d si d n a e d is d n a a k a h

t n e c re st o (m ts o m ( in la p n i la p t n e c re ts o m (

) lsi o s

, lte n e g ry e V

t n e c e r

si)l so

n e g ry e V

t n e c e r

-n e g ry e V

st o m (

) ils o s

t n e c e r n o n o , e lt

l a iv lu l a u ll a t n e c e r

t n e c e r n o

t n e c e r s in la p l ia v llu a

n o , e tl

e ivt c a

n o , le t

p e e d ry e V

s n i a l p s in la p l ia v l a vi u ll a

p e e d ry e V s n i a l p

e p d ry e V

s n i a l p

p e e d ry e V p e e d yr e V

) lsi so

, e tl n e g ry e V n e g yr e V lte n e g ry e V n e g yr e V

i ra e T (

)s il so

e lt n e G

s e p lso

n w ro (b

st re o f

)s il o s

p e e t S

e p lo s

s e p lso

n w ro (b

ts re fo

s)li o s

p e te S

e p lo s

s e p o sl

n w o r b (

st e r o f

)s il so

p e te S

e p o sl

p e e D te a r e d o M

p e te s

w llo a h S

n i a l p

s)li o s d o o l (F

n i a l p

) lsi so

) lis o s i a r e (T

s e p lo s r e vi R

n w o r b (

st e r o f

)s li o s

yr e V

n o tii s o p e p lo S

l a vi lu l a

p e e d ry e V p e e d yr e V p e e d ry e V te a r e d o M

tle

e p lo s

r o o P

h t p e D

r o o P t c fe r e p m I

r o o P -r e d o M

ll e w yl te a

te a r e d o M

-

-

-

yt l si e n i F y m a o l e n iF

-

te a r e d o M

e rs a o C

y m a o l

e sr a o C

y m a lo

se r a o C

y m a lo y m a lo

-

-

te ra e d o M

te a r e d

-

sre a o C

o M e t ra e d o M te a r e d o M e t ra e d o M

-

y e m in lo F a

-

se r a o C

y m a lo

-

yll e v ra G

y m a o l

-

y e m in a F lo

-

se r a o C

E

re e v e S

-

yll y ve m ra a o G l

e g a in ra D

n io s ro E

d o lo F

re u xt e T

r o o P tc fe r e p Im tc e rf e p Im w o ll a h s

ll e W

w lo l a sh

ll e W vie ss ce x E e v sis e cx

re ve e S

y m a lo


8 2 0 W

m a o L tli S

0 5 8 3

: 3 3 : 4

2 2 : 5 4

0 0 1

5 1 . 0

.4 1

7 . 3

: 7 :3 5

8 :2 5 3

.3 1

7 . 0

6 2 0 W

m a o L

0 5 8 3

: 4 3 : 9

3 2 : 3 4

0 0 1

5 .1 0

.4 1

.8 0

: 8 2 : 7

4 3 : 8 3

.3 1

.3 0

5 2 0 W

y d n a S

0 5 8 3

: 9 4 : 4

9 :1 2 3

0 0 1

5 .1 0

.4 1

.1 1

: 2 5 : 8

0 :2 8 2

.4 1

.4 0

8 1 0 W

m a o L y d n a S

0 5 8 3

: 2 :4 9

2 :2 6 3

0 0 1

5 .1 0

.4 1

.0 1

: 0 4 : 2 1

5 2 : 5 3

.4 1

.4 0

0 1 0 W

y d n a S

m a o L

3 0 7 3

: 7 :3 4

3 :2 0 4

0 0 1

5 .1 0

.4 1

.1 1

: 6 :3 5

8 :2 6 3

.3 1

.4 0

8 0 0 W

y d n a S

m a o L

3 8 6 3

: 2 :4 9

2 :2 6 3

0 0 1

5 1 . 0

4 . 1

0 . 1

: 0 4 : 2 1

5 2 : 5 3

4 . 1

4 . 0

7 0 0 W

m a Lo

0 5 8 3

: 4 3 : 9

3 2 : 3 4

0 0 1

5 1 . 0

4 . 1

8 . 0

: 8 2 : 7

4 :3 8 3

3 . 1

3 . 0

6 0 0 W

y d m n a a o S L

9 4 8 3

: 9 3 : 4

0 :2 1 4

0 0 1

5 .1 0

.4 1

.9 0

: 1 4 : 8

1 :2 8 3

.4 1

.4 0

4 0 0 W

y d n a S

2 6 6 3

: 1 4 : 0 1

0 2 : 9 3

0 0 1

0 1 . 0

4 . 1

4 . 1

: 5 4 : 0 2

0 2 : 5 3

4 . 1

5 . 0

3 0 0 W

m a Lo

2 6 6 3

: 4 :4 1 1

3 :2 3 3

0 0 1

5 .1 0

.4 1

.0 1

: 5 :3 6 2

1 :4 4 2

.3 1

.4 0

2 0 0 W

y d n a S

m a o L

2 6 6 3

: 2 :4 0 2

1 :2 7 3

0 0 1

5 .1 0

.4 1

.4 1

: 4 4 : 5 3

1 :2 5 3

.4 1

.6 0

1 0 0

y d n a S y d n a S

m a o L

7 1 7 3 3 0 7 3

: 3 4 : 6 2 : 8 :7 5

3 2 : 4 3

0 1

1 .0 0

4 . 1

4 . 1

-

-

-

:9 3 1

0 0 1

5 .1 0 ) m m / m m ( C W A

.6 1 lk u b yr d il so p o T

.6 0

.4 1 lk u b yr d il so b u S

.3 0

W iv R

re txu e t d e vi r e D

m a o L

m a o L

m a o L

D S W H d e h tc a M

its n u

e r txu e t il o s p o T

t) n ce r e p (

: d n a S :l e v ra G

y la C : lit S

) m c( h t p e d li o S

) cm / (g y its n e d

3

il o s p o T

t n te n o c ci n a g r o

t) n e c r e p (

e r u xt te li so b u S

)t n ce r e p (

: d n a S :l e v ra G

y la :C itl S

) m /c g ( itys n e d

3

il so b u S

t n te n o c ci n a g r o

t) n e c r e p (

63


8 2 0 W

: 3 :2 4

2 :2 5 5

: 8 :3 5

7 :1 5 4

C

2

0 5

0 0 2

0 0 3

0 3 . 0

5 2 . 0

5 1

0 3

6 2 0 W

: 4 3 : 9

3 :2 3 4

: 9 4 : 7

8 :1 3 3

C

2

0 5

0 0 2

0 0 3

5 .2 0

0 .2 0

5 1

0 3

5 2 0 W

: 9 :9 :5 2 4 3

: 4 :4 7 : 2 8 2

B

2

0 5

0 0 2

0 0 3

0 .2 0

5 .1 0

5 2

0 5

8 1 0 W

: 2 5 : 9

2 1 : 6 3

: 7 6 : :7 2 6 1 2

B

2

0 5

0 0 2

0 0 3

0 .2 0

5 .1 0

0 5

0 0 1

0 1 0 W

: 7 :4 4

3 :1 0 4

: 2 :8 6 : 0 5 3

B

2

0 5

0 0 2

0 0 3

0 .2 0

5 .1 0

0 5

0 0 1

8 0 0 W

: 2 5 : 9

2 1 : 6 3

: 7 :6 2 1

:7 6 2

B

2

0 5

0 0 2

0 0 3

0 2 . 0

5 1 . 0

0 5

0 0 1

7 0 0 W

: 4 3 : 9

3 :2 3 4

: 9 4 : 7

8 :1 3

C

2

0 5

0 0 2

0 0 3

5 2 . 0

0 2 . 0

5 2

0 5

6 0 0 W

: 9 :4 4

0 1 : 1 4

: 4 5 : 6 : 1 8 3

B

2

0 5

0 0 2

0 0 3

0 .2 0

5 .1 0

0 5

0 0 1

4 0 0 W

: 1 :6 0 1

0 :1 9 2

: 6 :7 0 2

:5 9 1

A

2

0 3

0 7

0 0 3

5 1 . 0

0 1 . 0

0 0 1

0 5 1

3 0 0 W

: 4 4 : 1 1

3 2 : 3 3

: 9 :5 6 2

8 :1 3 2

B

2

0 4

0 0 1

0 0 3

0 .2 0

5 .1 0

4 8

0 0 1

2 0 0 W

: 2 :6 0 2

1 :1 7 2

: 7 7 : 5 3

6 : 7 1

A

2

0 3

0 7

0 0 3

5 1 . 0

0 1 . 0

0 0 1

0 5 1

1 0 0

: 8 6 : 6 2 : 8 :7 5

1

0 5

-

0 0 3

0 1 . 0

-

0 0 2

-

1

0 0 1

-

0 0 3

8 .0 0

-

0 5 2

-

rs ye a l f o ° N

r ye a l ss e n k ic h T

r ye a l ss e n k ic h T

) m c( X M Z L_ O S

1 r ye la C W A

2 r ye la C W A

1 r e y la K L_ O S

2 r e y la K L_ O S

W iv R

p to d e tc rre o C

64

3 1 : 9 1

-

:9 3 1

e r u xt te li o s

A

-

)t n e c r e p (

: d n a S l: e v a r G

y a l C tl:i S

b su d e tc rre o C

A

e r u xt te li o s

)t n e c r e p (

: d n a S l: e v a r G

y a l C tl:i S

ic g o l o r d y H

p u ro g

) m c( 1

) m c( 2

) h / m m (

) /h m m (


0 2 2 , 2 : e ic R _ n o o s n o M

) C (° U -

H P

. rse et m raa p p rco / noi at etg ve T A W S : .2 A el b aT

°C 5 2

e lu a v l a n i m ig r 0 O 1

C ° 0 3

r te e m ra a P

T P O _ T

m ro f d e t a e r C

-

X M T H C

re v -E st e r o F ( E S R F

) n e re g

3

.7 0

m 1

2

0

m 2

I LA B

N I M I_ A L A

X M D R

) h s u r B e g n a (R B G N R

4

m 0 1

m 2

°C 0 3

0 3 3 , 1

5

m .5 3

m 5 1

°C 5 2

I LA B

X M D R

X M T H C

T P O _ T

-

rm a W

-

n o s a se ld o C

) e ic R ( E C I R

J S R F

0 4 5 , 1

5 5 6 ,1

n o s a se

l a u n n a

°C 5 2

l a u n n a

C D I

-

°C 8

5

J S R F

B A E T

J M R F

W A J, A A A

n io t ta e g e V

s e e tr e g r La

a e T

m iu s d e e e M tr

e ic R

J B W A

m 1

rm a W

l a u n n a

C ° 5 2

C ° 5

n o s a se d l o C

l a u n n a

C ° 8 1

C ° 0

T P O _ T

E S A B _ T

°C 2 2

°C 7

3

m 2

T P O _ T

E S A B _ T

I LA B

X M D R

) o t ta o P ( A T O P

l a r u lt u ci r g (A L R G A

)c ir e n e -G d n a L

d n a JJ

e s u d n la T A it W n S u

p ro c /

0 4 5 , 1 : e ic R _ r e m m u S

n o s a se

d e fii d m o 5 M 1

s n o it a ci fi d o M

-

0 4 ,5 1 : e ci R _ n o o s n o m e r P

C D I

)t a e h W g n ir p S ( T H W S

JJ W A

JJ W A

JJ W A d n a J A A A

to a t o P

e t u J

t a e h W

65


)l % % a 3 5 tt o ll f e o a w e e r % b A ( u 0 0 T 6 ) 6 p a 1 3 e h ( e D 4 9 . o N )l % % 6 a t 3 1 1 to f o fti a e ( L rA % r 0 0 e 0 ) 4 iv a 6 2 , R 1 (h

, d e h sr e t a w a k a h ld Ja e h t g n i in a t n o c sk c lo b e ivt a rt si in m d a r e p n o ti a rig ir f o s e cr u o S : 3 . A le b a T 66

. o N

l a n a C

) a h (

5 / 4 0 0 2 r a e y

) 5 0 0 2 ( B W D P D : a t a d f o e rc u o S

t c tri is D

k c o l B

% % % 3 2 6 1

% % % % % % % % % 6 5 0 1 9 9 6 6 9 2 2 1 3 5 1

0 4 2

0 0 0 2 8 0 8 4

0 3 0 8 2 0 5 0 4 4 3 9 0 1 3 3 ,4 5 ,7 4 ,8 4 2 ,2 8 5 1 3

6

2 2 0 1

3 9 2 8 0 2 5 0 2 1 1 2 1 1 1 3 1

8 4 ,9 2 2

% % % % % % % % % % % % % % % % 2 0 6 0 6 4 0 2 7 5 8 3 2 1 2 6 1 1 1 1 2 2 1 1 4 1 1

0 0 0 0 0 0 0 4 2 5 8 0 2 8 6 5 2 4 8 0 0 8 4 4 7 6 6 3 3 1 2 5 8 2 3 9 8 , 2 2 8 , 7 2 , 3 3 5 4 3 3 , 1 8 1 9 1

6 0 9 8 3 3 6 4 0 3 3 3 2 2 0 4 3 9 1 4 2 1 3 5 1 1 3 4 2 3 3 2 1 2 3 4

l) % 6 % 4 ta o t f a o e k rA % ( n a 0 0 T ) 0 0 a 3 3 (h . o N )l a t to f a o re % ( A

% 4

% 2 % 2 % 2 % 4 % 5 % 4 % 2 % 6 % % 0 % 3 % 3 % 1 % 3 % 2 % 3 3

0 0 0 0 0 0 0 5 0 5 0 8 4 0 5 7 0 5 0 0 0 0 0 9 9 3 5 1 6 2 5 7 2 1 1 3 1 1 3 6 1 3 1 3 6 , 3

2 2 1 1 1 1 3 1 1 0 8 2 3 0 6 0 7 9 1 4 1 2 3 2 6 1 % % % % % % % % % % % % % % % % % % 9 0 4 8 5 9 3 0 1 1 1 0 4 2 2 1 2 7 3 4 4 5 5 4 2 5 6 1

0 0 0 , 2

riu ig a lp a J

I r a u rd u p il A

0 2 0 , 3

riu ig a lp a J

ri u g p u h D

0 0 ,0 3

riu g i a lp a J

a t a k la a F

0 5 ,3 1

riu g i a lp a J

i n i h c l a K

0 4 ,5 3

riu g i a lp a J

0 4 ,2 2

riu g i a lp a J

t a h ir a d l a a M M

0 0 ,6 1

riu g i a lp a J

ri u g a n y a M

0 0 ,0 1

riu g i a lp a J

il ia t e M

0 0 0 0 0 0 0 0 0 0 0 0 5 4 0 4 0 5 3 6 2 3 2 2 1 , ,4 2 1 1 0 2

riu g i a lp a J

r a h e B h c o o C

a t a k a r g a N

I r a h e B h c o o C

r a h e B h c o o C

I a t a h n i D

r a h e B h c o o C

r a h e B h c o o C

r a h e B h c o o C

r a h e B h c o o C

r a h e B h c o o C

r a h e B h c o o C

II a t a h n i D

I a g n a h b a h t a M

II a g n a h b a h t a M

j n a ig l h k e M

i h c u i lk a ti a ti S S

l a t o T


l ta o T

l) a t to f o a e ( r % A

% 0 0 1

% 0 0 1

4 9 1 2 ) ,1 ,5 a 5 7 h (

. o N

6 9 ,6 1

3 1 ,5 2

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

% 0 0 1

1 8 ,7 6

3 1 ,3 2

8 9 ,3 6

7 6 5 , 4

0 3 8 , 6

0 2 0 , 2

5 9 2 , 2

2 9 1 , 1 2

6 6 7 , 1 1

5 1 5 , 8

9 2 7 , 4

5 4 7 , 9

0 6 3 , 4

9 4 7 , 3

6 0 6 , 4 1

5 5 1 3 4 1 1 4 ,5 2 1 6 , 2 5

6 7 ,2 5

5 0 2 , 4

5 2 ,5 4

2 9 ,3 5

9 7 ,7 4

0 2 ,5 1

3 7 ,1 3

9 0 5 , 2 2 1 2 0 ,2 6 4

2 4 5 8 3 0 7 9 ,7 3 3 ,1 2 1

)l % % % 3 4 a t 2 3 2 2 o t f a o e r % rs A ( e 6 5 9 h 1 4 ) 2 t a ,6 ,7 ,6 O h ( 1 1 1 . o N )l ta to f a o r % ll e e A ( w g ) u a D h ( . o N

0 8 3 , 1

0 0 9 , 1

k c lo B

3 0 3 8 0 0 2 3 4 4 4 1 2 9 3 9 0 9 1 8 0 4 9 5 0 4 7 6 1 6 4 ,2 6 8 , 4 5 3 6 4 5 3 4 1 2 ,3 1 1 3 1

0 5 0 0 0 0 5 7 7 8 7 2 8 8 6 3 9 1 3 6 3 4 4 1 9 2 3 9 9 1 8 0 7 9 2 0 1 1 , 1 , , 9 2 1

% % % % % % % % % % % % % % % % % % 4 3 5 8 5 5 3 8 2 1 4 1 9 1 6 6 1 5 3 1

0 0 0 0 0 0 0 0 5 9 8 5 6 0 0 7 8 8 0 9 4 8 0 5 8 6 5 0 1 1 7 3 9 3 0 3 2 1 3 1 3 2 1 1 2 4 1 8 1 5 , 2 1 ,5 1 5 0 0 0 0 0 0 0 0 5 1 3 8 1 5 3 1 1 8 0 9 4 8 0 5 8 6 5 1 2 0 5 6 2 5 4 2 2 1 3 1 3 2 1 1 1 1 6 ,4 ,3 , 1 3 7

)l % % ta 4 0 1 to ll f e o w a e e % b r ( u A 8 4 T 4 ) 8 a 1 7 w h o ( ll a h 4 2 S . 9 7 3 o N

t ic tr s i D

% % % % % % % % % % % % % % % 1 9 5 7 4 2 2 6 5 1 4 1 4 1 1 2 1 1 2 2 2 1 1 1

i rg u i a p l a J I r a u d r u p li A

i r u g i a p l a J

ir u g p u h D

% 8

% % % 1 7 1 1

% % % % % % % % % 7 2 7 5 7 7 7 5 0 2 7 5 4 4 2 6 3 3

2 3 5

8 4 2 5 0 3 3 7

2 6 7 , 5

0 0 5 , 8

6 7 ,8 4

3 3 ,1 2

1 8 ,5 4

3 9 ,1 1

7 2 ,5 2

8 3 ,1 5

8 6 ,2 7 3

6 6 2

9 2 6 2 5 6 1 3

9 5 4 , 5

6 4 9 , 4

2 2 ,0 4

4 9 ,6 2

3 8 ,1 5

0 0 0 , 1

2 1 4 , 1

6 5 0 , 3

1 5 0 , 9 2

i rg u i a p l a J

r a h e B h c o o C

r a h e B h c o o C

r a h e B h c o o C

r a h e B h c o o C

r a h e B h c o o C

r a h e B h c o o C

r a h e B h c o o C

r a h e B h c o o C

ta a k ra g a N

I r a h e B h c o o C

II a t a h in D

I a g n a h b a h t a M

II a g n a h b a th a M

j i n h a c ig u l h i k l k a a e it ti M S S

i rg u i a p l a J

ta a k a l a F

i rg u i a p l a J

i rg u i a p l a J

i rg u i a p l a J

i rg u i a p l a J

i n i h c l a K

ir t a u g ih r a a n d l y a a a M M M

i rg u i a p l a J

il a ti e M

I a t a h in D

l ta o T

67


.s p e st n tio ra ib l ca T A W S : 4 . A le b a T 68

s e lu a v l a in F

5 1 . 0 , 5 2 . :s0 il o s ll a ,J S R F

? y h W

a T E e s a rce In

e g n a h C

d e li p p a

r te e m a r a p n o ti a r ib l a C t u p t u O

n o c e li f

la u im S

5 .2 0 , 0 .3 s:0 li so ll a ,J M R F

, 5 2 . 0 : C & B p u o r g isl so , its n u l a r tu l ciu 5 r 1 g . A 0

2 x : A p u ro g sli so ,J S R F

e s a e rc In

e s a e rc In

r te e m a r a P

C W A

Z _ l o S

le b a T

l so

l o s

d e r e id s

.5 1 x : A p o ru g lsi o s J, M R F

l_ o S : C & B p u o r g sil so , its n u l a r tu l u icr g A

7 3 : 7 A 7 m m 0 0 0 3 = 2 Z

p u ro g sli so ,J S R F

:C & B lsi so ,J S R F

e s a e rc e D

e rg a ch re

9 3 : A p o ru g lsi o s J, M R F

0 8 : C & B p o ru g lsi o s J, M R F

2 8 : C & B p o ru g sil so , B A E T

3 6 : A p u o r g sil so , its n u l a r tu l u icr g A

0 9 : C & B p u o r g sil so , its n u l a r tu l u icr g A

, e ci R r_ e m m u S – e ci R _ n o so n o M d e t a ig r Ir

5 9 s:il o s ll a

3 9 : lsi o s ll a , A F I V

e p o sl h it w s ie r a V

1 . 0 x : % 3 > e p lo s r ve e r e h W

sy a d 0

m m 0 5 ,1 2

2 0 .0 0

e h t r e ff u B

0

.3 0

e th r e ff u B

w o lf e s a b

.5 0

ff o n u r

n o it p O

se a re c e D

se a re c e D

e s a re c e D

e s a re c e D

e s a re c e D

e s a rce In

e s a rce In

2 N C

2 N C

K _ l o S

Y A L E D _ W G

F B _ A H P L A

IN M Q W G

P D _ G R H C R

) 1 ( N _ H C

) 2 ( N _ H C

t1 g m

t1 g m

l o s

w g

w g

w g

w g

b u s

e rt

e s a e rc In

.t u tp u u r o h

.t u tp u u r o h

.t u tp u u r o h

.t u tp h u o rc

e im t n p e o ti ts

e g a r e v A

ly th n o m

e g a r e v A

ly th n o m

e g a r e v A

yl th n o m

yil a D

le b a ir a V

d e s ly a n a

n a tr o p va E

n o ti a ir sp

w llo a h S

d n u o r g

w llo a h S

a w d n u ro g

-i l a C

n io t p ra e b st

1

2

r te a w

e rg a h c re

3

e s a b w r o te fl

m a e tr S

4

s w o lf


st n a sit ss a d l e fi e th y b d e lli f ) 5 3 T G e its ( m r o f g n i th rtu d n u ro g

). 0 1 0 (2 e am th f ra o h e l S p nd m ai a x a E C : m 1 . or A fd e r te u pa g d iF A

69


REFERENCES Allen, RG, L S Pereira, D Raes, and M Smith. 1998. Crop evapotranspiration. Rome, Italy: Food and Agriculture Organisation. Arnold, J G, PM Allen, and G Bernhardt. 1993. “A comprehensive surface-groundwater ow model.” Journal of Hydrology 142:47-69. Arnold, J G, R Srinivasan, R S Muttiah, and J R Williams. 1998. “Large area hydrologic modeling and assessment: Part I. Model development.” Journal of the American Water Resources Association 34:73-89. Arnold, J G, J R Williams, and D R Maidment. 1995. “Continuous-time water and sediment-routing model for large basins.” Journal of Hydraulic Engineering 121:171-183. de Bruin, A, M Mikhail, A Brahmachari, and J Barron. 2010. Baseline assessment of current livelihood strategies in Jaldhaka watershed, West Bengal, India . York, UK: Stockholm Environment Institute. Cai, X L, and B R Sharma. 2010. “Integrating remote sensing, census and weather data for an assessment of rice yield, water consumption and water productivity in the Indo-Gangetic river basin.” Agricultural Water Management 97(2):309-316. CGWB. 2009. Ground water information booklet, Koch Bihar district, West Bangal. New Delhi, India: Central Ground Water Board Retrieved November 15, 2010 (http://cgwb.gov.in/District_Prole/WestBangal/Koch_Behar.pdf). Chatterjee, R, and R R Purohit. 2009. “Estimation of replenishable groundwater resources of India and their status of utilization.” Current Science96:1581-1591. de Condappa, D, J Barron, S K Tomer, and M Sekhar. 2011.Application of SWAT and a groundwater model for impact assessment of agricultural water management interventions in Jald haka watershed: Current state of water resources and scenarios of agricultural development. York, UK: Stockholm Environment Institute. DPDWB. 2005. Districts statistical handbook. Kolkata, India: Development & Planning Department - West Bengal Retrieved June 3, 2010 (http://www.wbplan.gov.in). FAO/IIASA/ISRIC/ISS-CAS/JRC. 2009. Harmonized World Soil Database (version 1.1). FAO, Rome, Italy and IIASA, Laxenburg, Austria. Geza, M, and J E McCray. 2008. “Effects of soil data resolution on SWAT model stream ow and water quality predictions.” Journal of Environmental Management 88:393-406. HWSD. 2009. “Harmonized World Soil Database.” Retrieved November 3, 2010 (http://www.iiasa. ac.at/Research/LUC/External-World-soil-database/HTML/index.html?sb=1). Jarvis, A, H I Reuter, A Nelson, and E Guevara. 2008. Hole-lled SRTM for the globe Version 4. Retrieved March 10, 2009 (http://srtm.csi.cgiar.org). 70


Kalman, R E. 1960. “A new approach to linear ltering and prediction problems.” Transaction of the ASME—Journal of Basic Engineering 35-45. Krause, P, D P Boyle, and F Bäse. 2005. “Comparison of different efciency criteria for hydrological modelassessment.” Advances in Geosciences 5:89-97. Kundu, N, and G. N. A. Soppe. 2002. Water resources assessment. Terai region of West Bengal.New Delhi, India: Jawahar Publishers & Distributors. Leduc, C, J Bromley, and P Schroeter. 1997. “Water table uctuation and recharge in semi-arid -cli mate: some results of the HAPEX-Sahel hydrodynamic survey (Niger).” Journal of Hydrology 188-189:123-138. Maréchal, J C, B Dewandel, S Ahmed, L Galeazzi, and F K Zaidi. 2006. “Combined estimation of specic yield and natural recharge in a semi-arid groundwater basin with irrigated agriculture.” Journal of Hydrology 329:281-293. Moon, S.-K., N C Woo, and K S Lee. 2004. “Statistical analysis of hydrographs and water-table uctuation to estimate groundwater recharge.” Journal of Hydrology 292:198-209. Mukherji, A. 2007. “Political economy of groundwater markets in West Bengal, India: evolution, extent and impacts.” PhD thesis, United Kingdom: University of Cambridge. Nash, J E, and J V Sutcliffe. 1970. “River ow forecasting through conceptual models part I – A discussion of principles.” Journal of Hydrology 10:282-290. Neitsch, S, J G Arnold, J R Kiniry, R Srinivasan, and J R Williams. 2010. Soil and Water Assessment Tool Input/Ouput le documentation. Version 2009.College Station, Texas, USA: Texas Water Resources Institute. Neitsch, SL, J G Arnold, J R Kiniry, and J R Williams. 2005. Soil and Water Assessment Tool, theoreti cal socumentation. Version 2005.Temple, USA: Agricultural Research Service. Park, E, and J C Parker. 2008. “A simple model for water table uctuations in response to precipita tion.” Journal of Hydrology 356:344-349. Raghuwanshi, N S, N S Singh, and A Bandyopadhyay. 2007. Development and application of a decision support system for estimating reference crop evapotranspiration for different agro-cli-

matic regions. New Delhi, India: Final report submitted to ICAR. Rajeevan, M, and J Bhate. 2008. A high resolution daily gridded rainfall data set (1971-2005) for mesoscale meteorological studies. Pune, India: National Climate Centre. Rockström, Johan. 2003. “Water for food and nature in drought–prone tropics: vapour shift in rain–fed agriculture.” Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences 358(1440):1997 -2009. Retrieved (http://rstb.royalsocietypublishing.org/content/358/1440/1997.abstract). 71


Romanowicz, A A, M Vanclooster, M Rounsevell, and I La Junesse. 2005. “Sensitivity of the SWAT model to the soil and land use data parametrisation: a case study in the Thyle catchment, Belgium.” Ecological Modelling 187:27-39. Ruud, N, T Harter, and A Naugle. 2004. “Estimation of groundwater pumping as closure to the water balance of a semi-arid, irrigated agricultural basin.” Journal of Hydrology 297:51-73. Shamsudduha, M, R E Chandler, R G Taylor, and KM Ahmed. 2009. “Recent trends in groundwater levels in a highly seasonal hydrological system: the Ganges-Brahmaputra-Meghna Delta.” Hydrology and Earth System Sciences 13:2373-2385. Simon, D. 2006. Optimal state estimation. Kalman, Hinnity and nonlinear approaches . Hoboken, USA: Wiley-Interscience. Srinivasan, R, and J G Arnold. 1994. “Integration of a basin-scale water quality model with GIS.” Water Resources Bulletin30:453-462. TERI. 2007. Socio-economic-techno-environmental assessment of IDEI products. Treadle Pump. New Delhi, India: The Energy and Resources Institute. Tomer, S K, Y Javeed, and M Sekhar. 2010. “An alternative approach for estimation of groundwater recharge and discharge in hard rock aquifer.” inProceedings of Workshop “Ground Water Resources Estimation.”New Delhi, India.

72

SEI - Africa Institute of Resource Assessment University of Dar es Salaam P.O. Box 35097, Dar es Salaam Tanzania Tel: +255-(0)766079061 SEI - Asia 15th Floor Witthyakit Building 254 Chulalongkorn University Chulalongkorn Soi 64 Phyathai Road Pathumwan Bangkok 10330 Thailand Tel: +(66) 22514415 SEI - Oxford Suite 193 266 Banbury Road, Oxford, OX2 7DL UK Tel: +44 1865 426316 SEI - Stockholm Kräftriket 2B SE -106 91 Stockholm Sweden Tel: +46 8 674 7070

SEI - Tallinn Lai 34, Box 160 EE -10502, Tallinn Estonia Tel: +372 6 276 100 SEI - U.S. 11 Curtis Avenue Somerville, MA 02144 USA Tel: +1 617 627-3786

SEI - York University of York Heslington York YO10 5DD UK Tel: +44 1904 43 2897

The Stockholm Environment Institute SEI is an independent, international research institute. It has been engaged in environment and development issues at local, national, regional and global policy levels for more than a quarter of a century. SEI supports decision making for sustainable development by bridging science and policy.

sei-international.org

Application of Swat

Recommend Documents