Home Articles Hydrological Modelling of canal command using Remote Sensing and GIS

Hydrological Modelling of canal command using Remote Sensing and GIS

P. K. Gupta, T. Das, N. S. Raghuwanshi, R. Singh
Agricultural and Food Engg. Department,
Indian Institute of Technology, Kharagpur,
West Bengal, 721302, India

S. Dutta, S. Panigrahy
Space Applications Centre, Ahmedabad,
Gujrat – India

Introduction
The major irrigation projects in India and South Asia are reported to perform at a low overall efficiency of 30-35% (Sanmugnathan and Bolton, 1988). Total irrigated area in the country presently accounts for 92 M ha, with canal irrigation projects accounting for about 17.4 M ha of irrigated area (Ministry of Water Resources, 1997). Ensuring reliable canal releases and balancing them with the demands for water has proved difficult, with the results that crops do not receive the right quantities of water at the proper time. In the Damodar Valley Corporation (DVC) command, there is significant amount of canal water wastage due to untimely release, lack of regulatory measure at the outlet and free flooding.

Distributed Hydrological Models (DHM), (Rogers et al., 1985), provide a means of studying the interrelationships of upstream and downstream hydrological regime and of managing water and land resources. MIKE SHE (Refsgaard and Storm, 1995), is a comprehensive, distributed, and physically based modelling system capable of simulating all major hydrological processes in the land phase of hydrological cycle. A major problem in the hydrology is the inadequate field measured data to describe the hydrologic processes. Remote Sensing (RS) has been identified as a tool to produce information in spatial and temporal domain, instead of point measurement, in digital form, with high resolution. Further, RS techniques are extremely relevant as a means of estimating a number of key variables specifically in situation where DHM are required. Remote sensing techniques can produce high spatial coverage of important terms in water balance for large area, but at the cost of a rather sparse temporal resolution. Hydrological model can produce all the terms of water balance at a high temporal, but low spatial resolution (Droogers and Bastiaanssen, 2002). The use of RS data, in combination with DHM, provides new possibilities for deriving spatially distributed time series of input variables, as well as new means for calibration and validation of the hydrological model (Bastiaanssen et al., 2000, Fortin et al., 2001). The use of RS technology involves large amount of spatial data management and requires an efficient system to handle such data. Hence, Geographic Information System makes it possible to store, analyze, retrieve and manipulate data for large and complex problems.

The uniqueness of the research is perceived in the fact that work focuses on large scale hydrological modelling, where the RS data is considered to be essential for an improved understanding of the functioning of large command areas. In the present study, MIKE SHE is applied to the 6 main canal command of the DVC irrigation project for simulating the hydrological water balance taking into account remote sensing data and canal release data. The research focuses on the feasibility of integrating RS data with DHM and its subsequent impact on studies related to the hydrological regime of the command areas. The advantages of integrating RS data with DHM are depicted and discussed.

Model Description
The MIKE SHE is a comprehensive deterministic, distributed and physically based hydrologic modelling system, capable of describing the entire land phase of the hydrological cycle in a given command. The model area is discredited by two analogous horizontal-grid square networks for surface and ground water flow components. These are linked by vertical column of nodes at each grid representing the unsaturated zone. A finite difference solution of the partial differential equations, describing the processes of overland and channel flow (Saint-Venant equations), unsaturated flow (Richards’ equation) and saturated flow (Boussinesq equation for confined aquifer), is used for water movement modelling. Interception process is modelled by Jensen model (Jensen, 1983) and actual evapotranspiration is calculated by Kristensen and Jensen model (Kristensen and Jensen, 1975). Interception and Evapotranspiration Component (et)
The interception process is modelled by introducing interception storage, expressed as a function of leaf area index (Jensen, 1983). The actual evapotranspiration is calculated based on the potential evapotranspiration using the Kristensen and Jensen model (Kristensen and Jensen, 1975). The actual evapotranspiration rate, consisting of the sum of actual transpiration and soil evaporation, is further adjusted according to vegetation density and water content in the root zone. Leaf area index and root depth need to be specified as a function of time for the actual evapotranspiration calculations. The actual transpiration is calculated using the following relationship:

———————(1)
Where, Eat = actual transpiration; f1(LAI) = leaf area index function; f2(q) = soil moisture function; RDF = root distribution function; and Ep = potential evapotranspiration.

———————-(2)
where, qf = volumetric moisture content at field capacity; qw = volumetric moisture content at wilting point; q = volumetric moisture content; and C3 = empirical parameter, mm/day.

Unsaturated Zone Component (Uz)
Soil moisture distribution in the unsaturated zone is calculated by solving the one-dimensional Richards’ equation. Extraction of moisture for transpiration and soil evaporation is introduced via sink terms at the node points in the root zone. Infiltration rates are found by the upper boundary that may be either flux controlled or head controlled. The lowest node point included in the finite difference scheme depends on the pyretic surface level, and allowance is made for the unsaturated zone to disappear in cases where the pyretic surface rises to the ground surface.

Saturated Zone Component (Sz)
The ground water flow is modelled using an implicit finite difference solution of the two-dimensional non-linear Boussinesq equation for an unconfined aquifer.

General Description of Study Area
Pilot area of the 6 main canal command of Damodar Irrigation Project has been chosen as the study area, situated in the western part of West Bengal, India. Study area comprising total of 548 sq. km. The average annual rainfall in the command is about 1400 mm, spread over 75 rainy days. The relative humidity ranges from 65% to 95% whereas temperature varies from 11.4 to 41.7 degree Celsius. Paddy is the main crop in both Kharif and Rabi seasons, whereas vegetables are grown in few patches in summer season. On an average, three irrigations are provided in each cropping season (Kharif and Rabi).

Data and Methodology

Data Requirement and Collection
Since the present study involves hydrological modelling of the irrigated command, the data requirements were large. The map of the 6 main canal system showing longitudinal section, details of irrigation structures, network of canals and topographic map of its command were collected from the Irrigation and Waterways Department, Government of West Bengal. Daily flow releases at system source and daily flow rates at two locations in the main canal and at three distributaries were also collected for Kharif irrigations of 1999 from the above source. The soil survey map of the command was obtained from the National Bureau of Soil Survey and Land Use Planning, Research Station, Calcutta. The RS land use map and soil map were prepared through three dates RADARSAT SAR imagery (16th July, 1st August and 25th August 1999) and IRS IC LISS-III imagery for December 1998 respectively. The climatic data such as daily rainfall and pan evaporation of three meteorological stations were obtained from the Department of Agriculture, Government of West Bengal. The leaf area index was measured by canopy analyser. Besides these, soil samples were collected from different locations (for different soil types) and physical properties of soils such as textural classification, hydraulic conductivity and soil moisture characteristic curves were determined experimentally. The monthly groundwater table depth was measured at 19 different locations in the command and processed to prepare the pre and post-monsoon groundwater table maps.
Pre-processing
To develop a topographic map of the command, topographic data are digitized and interpolated using the topogrid interpolation technique available with GIS, Arc Info. A grid size of 200 x 200 m is used. The optical and RADARSAT, SAR data are used to develop the land use map of the study area (Fig.1). The command has six distinct land uses, namely, early paddy (10.9%), medium paddy (41.5%), late paddy (15.5%), fallow (12.5%), homestead (9.5%), and water bodies (10.1%). Paddy is the only crop in the Kharif season, which covers about 68.0% of command area. Fig. 2 shows the conventional soil distribution map of the command as obtained from NBSS, Calcutta. Four major types of soil covering the whole study area, namely, loamy sand (30.6%), sandy loam (35.3%), fine sandy loam (19.7%) and fine loamy sand (14.3%) are found. Sandy loam is the predominant soil of the command. Also, a soil distribution map is prepared using the IRS-1C LISS III data for December 1998 (Fig. 3). The percentage area coverage by different soils, loamy sand, sandy loam, fine sandy loam and fine loamy sand, are 13.1%, 29.6%, 32.7% and 24.5% respectively. Here also, the same four soils are found, however, the spatial distribution and the area coverage differs. Soil moisture retention curve was experimentally determined for each of these soils by using the pressure plate apparatus up to 15 atmospheres. The empirical van Genutchen model (Maidment, 1992) is fitted to the experimental data and the relationship is extrapolated over the required range for MIKE SHE (pF = 0 to 6, i.e., up to 1000 atmospheres). Fig. 4 shows the depth to the pyretic surface for the pre-monsoon periods of 1999. Water table depths for pre-monsoon period vary from 4 m at the upstream to 18 m at the downstream, whereas for post-monsoon period the depth varies from 4 to 16 m, with downstream end getting waterlogged. Fig. 5 shows the delineated polygons map having different total applied water depth (sum of irrigation and rainfall). This polygon map is prepared by overlaying the precipitation polygon map (three stations i.e., Burdwan, Kalna and Katwa) on the polygon map of distributary commands (13 reaches).


Fig. 1 Land use/cover map


Fig. 2 Soil distribution map (NBSS)


Fig. 3 Soil variability map(Remote Sensing)


Fig. 4 Irrigation and rainfall distribution polygon map


Fig 5. Ground water table variation map For the year-1999 (pre monsoon)


Fig. 6 Observed & simulated ground water table variations for the Kharif-1999
Methodology
The output from Arc Info digitising program is transformed into the type 2 format (grid) files of MIKE SHE using the developed link. Number of grids in the x- and y-directions are taken as 166 and 187 respectively with the cell size of 200 m x 200 m. The time series data files of potential evaporation, precipitation plus irrigation, and vegetation parameters, such as leaf area index and root depth as a function of time, are prepared for running the simulation for Kharif season.

MIKE SHE model is applied to 6 main canal system of DVC to see the feasibility of integrating RS data with DHM for the hydrological analysis of the command area. The model was calibrated for the study area during 1 May to 30th October 1999. The calibration was done by varying the saturated hydraulic conductivity, ‘Ks’, drainage constant and exponent ‘n’ in the hydraulic conductivity function and by comparing the post-monsoon observed and simulated groundwater levels at measured locations. The calibrated model was then used to simulate different scenarios.

Simulation scenarios
Model simulations are done with four different combinations of Remote Sensing (RS) and conventional data. The different combinations are:

  1. Soil map (conventional) and land use (RS)
  2. Soil map (conventional) and land use map (conventional)
  3. Soil map (RS) and land use map (conventional)
  4. Soil map (RS) and land use map (RS)

Simulations for the above-stated four scenarios are carried out over 1st May to 30th October 1999. Kharif irrigation period is chosen because the canal system is primarily designed to cater the needs of Kharif crops.
Results and Discussion

Model Calibration
The calibration of the model was done by matching the post-monsoon water table fluctuations at fifteen locations in the command area. The model parameters varied during calibration are saturated hydraulic conductivity, Ks (1.07 cm/h to 2.33 cm/h), drainage constant (2e-6) and exponent (n) in the hydraulic conductivity function (7.1 to 11.5). Fig. 6 shows relationship between observed and predicted ground water elevations. Regression analysis between observed and predicted post monsoon groundwater level yielded a high (0.84) coefficient of determination. To statistically evaluate the calibration performance of the model two other criteria, suggested by ASCE committee (1993), Nash-Sutcliff coefficient and Student’s t-test are also carried out. Nash-Sutcliff coefficient is found as 0.775, which is reasonably good and Student’s t-test reveals that the model results are acceptable at 1% level of significance.

Simulation Results
The calibrated model is used to perform simulations with different scenarios to study the effect of using remote sensing data over conventional data with distributed hydrological model. The results are presented below:

Effect of spectral soil distribution
The first and fourth scenarios are used to study the effect of spectral soil distribution on the water balance simulations with the distributed hydrological model. Figs. 7 and 8 present temporal water content variation at a selected node in head reach for conventional and RS soil, respectively. It is evident that soil moisture content variations are different due to dissimilar spatial variability of soil obtained from different sources. This is obvious because the same node may represent different soil in RS soil map and NBSS soil map. Fig. 9 shows the moisture content variation at tail reach node for the first simulation scenario. Comparison between Figs. 7 and 9 reveal that head reach nodes are having high moisture content throughout the season as compared to the tail reach nodes. This indicates that head reach receives more irrigation water. Similar soil moisture pattern between head reach and tail reach nodes was also observed for the fourth simulation. Furthermore, there are dry patches at head reach as well as tail reach of the command for both simulation scenarios, in spite of heavy rainfall during the season. This implies that additional irrigation is essential for attaining the potential crop yield. The additional water requirements are calculated at head reach and tail reach of the command for both first and fourth scenarios. Due to variation in the soil moisture regime, the additional water requirements differ at head reach and tail reach in both scenarios. The additional irrigation water requirements to bring the soil moisture content of the paddy field to 75 percent of saturation value at head reach and tail reach are found as 55.7 cm and 63.0 cm for the season respectively for first scenario. The same for fourth scenario are found as 41.5 cm and 50.8 cm respectively. It is evident that total additional water requirement is comparatively less in fourth scenario (considering RS soil map) than that in first scenario (considering NBSS soil map) at head reach as well as at tail reach. It is also observed that total irrigation water requirements are more at tail reach than that at head reach with both soil maps. This implies that less water is available to the tail reach, a usual phenomenon in most of the Indian irrigation commands.


Fig. 7 Soil moisture variation in a node using RS soil (loamy sand)

Fig. 8 Soil moisture variation in a node using NBSS soil (sandy loam)

Fig. 9 Soil moisture content variation for tail reach node using NBSS-soil map
Effect of spectral land use distribution
The third and fourth scenarios are used to study the effect of spectral land use distribution on the water balance simulations with distributed hydrological model. The soil moisture variations are slightly different at the same node in RS land use pattern (categorised in six classes, early paddy, medium paddy, late paddy, fallow, homestead and water bodies) and conventional land use pattern (considering early paddy in whole area except for paved area), at head reach as well as tail reach. This is because a particular node represents different crop transplanting dates in RS land use and conventional land use, i.e., the node representing late paddy in RS land use pattern, represents early paddy in conventional land use pattern.
Additional irrigation Water Requirement in Distributaries Command
The amount of additional irrigation water is calculated for two separate simulation scenarios, namely simulation with canal irrigation under existing canal release schedule and simulation without canal irrigation. Fig. 10 presents the additional water required to bring the soil moisture content up to 75 percent of saturation for these scenarios. It is observed that even in the first scenario (simulation with canal irrigation) there is a need of additional irrigation water in spite of existing canal irrigation in all the distributary commands, i.e., the released canal irrigation water is not sufficient for potential yield in the command. Water requirement at head reach distributary commands is less, indicating that there is more canal supply through head reach distributaries. Whereas, at the tail reach distributary commands the requirement is high, indicating less canal supply through tail reach distributaries, except distribuatry K. The water requirements at head reach and tail reach distributary commands are not met uniformly by the existing canal release schedule, as evidence from the non-uniform difference between water requirements for two simulations. This shows that there is a need to reschedule the existing canal release pattern.

Fig. 10 Comparison between water requirements for simulations
with canal irrigation and without canal irrigation

Conclusions
The hydrological water balance of the Damodar 6 main canal system, West Bengal is carried out using MIKE SHE. The MIKE 11 simulation derived irrigation depths of different canal reaches along with respective rainfall amount are used as precipitation input to the model. MIKE SHE model is calibrated for the study area by comparing the observed and simulated post monsoon water table levels at fifteen observation wells. Observed and simulated post monsoon water table levels are found to be in close agreement at most of the points. Remote sensing soil map and NBSS soil map shows different soil types even at the same node. The additional irrigation water requirement estimated using RS soil map is less as compared to that of NBSS soil map at both head and tail reaches. There is not much variation in soil moisture distribution due to different land use patterns (obtained from RS and conventional) in the unsaturated zone during the simulation period at head and tail reaches. Simulation results show the occurrence of low moisture content in tail reach regions due to non-uniform distribution of irrigation water. There is a need for additional irrigation water both at head and tail reaches under the existing canal release and this requires rescheduling of the existing delivery pattern. Results of MIKE SHE along with RS data show the potential of the integrating remote sensing data with Distribute Hydrological Models towards improved information generation on hydrological regime of the command.

References

  1. ASCE Task Committee on Definition of Criteria for Evaluation of Watershed Models. (1993). Criteria for evaluation of watershed models. Journal of Irrigation and Drainage Engineering, ASCE, 119 (3), 429-442.
  2. Bastiaanssen, W. G. M., (2000). Irrigation performance indicators based on remotely sensed data: a review of literature. Irrigation and Drainage Systems, 13(4), 291-311.
  3. Droogers P. and Bastiaanssen W. (2002). Irrigation performance using hydrological and remote sensing modelling.” Journal of Irrigation and Drainage Engineering, ASCE, 128 (1), 11-18.
  4. Fortin, J. P., Turcotte, R., Massicotte, S., Moussa, R., Fitzback, J. and Villeneuve, J. P. (2001). Distributed watershed Model Compatible with Remote Sensing and GIS Data. I: Description of Model, Journal of Hydrologic Engineering, 6, 91~99.
  5. Jensen, K.H. (1983). Simulation of water flow in the unsaturated zone including the root zone technical series paper no. 33. Institute of Hydrodynamics & Hydraulic Engineering, Technical University of Denmark, Copenhagen, Denmark.
  6. Kristensen, K.J. and Jensen, S.E. (1975). A model for estimating actual evapotranspiration from potential transpiration. Nordic Hydrology. 6, 70 – 88.
  7. Maidment, D. R. (1992). Handbook of Hydrology. McGraw Hill Publications. 5.1-5.39.
  8. Ministry of Water Resources, (1997). Report of the Planning and Programme Implementation Government of India, New Delhi India.
  9. Refsgaard, J.C. and Storm B. (1995). MIKE SHE Computer models in watershed hydrology. V.P.Singh ed. Water Resource Publications, Colorodo, USA. 806-846.
  10. Rogers, C.C.M., Beven, K.J., Morris, E.M., and Anderson, M.G. (1985). Sensitivity analysis calibration and predictive uncertainty of Institute of Hydrology Distributed Model. Journal of Hydrology, 81,179-191.
  11. Sanmugnathan, K. and Bolton, P. (1988). Water management in third world irrigation schemes-Lesson from the field. ODU Bull, 11, Hydraulic Research, London, UK.