WaterSubject

The combined use of remote sensing and a distributed hydrological model have demonstrated the improved understanding of the entire water balance in an area where data are scarcely available. Water use and crop water productivity were assessed in the Upper Bhima catchment in southern India using an innovative integration of remotely sensed evapotranspiration and a process-based hydrological model. The remote sensing based Surface Energy Balance Algorithm for Land (SEBAL) was used to derive an 8 month time series of observed actual evapotranspiration from October 2004 to May 2005. This dataset was then used in the calibration of the Soil and Water Assessment Tool (SWAT). This hydrological model was calibrated by changing 34 parameters to minimize the difference between simulated and observed actual evapotranspiration. The calibration efficiency was assessed with four different performance indicators. The calibrated model was used to derive a monthly basin water balance and to assess crop water productivity and crop water use for the irrigation year 2004–2005. It was found that evapotranspiration is the largest water loss in the catchment and total evaporative depletion was 38,172 Mm³ (835 mm). Of the total evaporative depletion 42% can be considered as non-beneficial and could be diverted to other beneficial utilization. Simulated crop water productivities for sugarcane, sorghum and winter wheat are relatively high at 2.9 kg/m³, 1.3 kg/m³ and 1.3 kg/m³, respectively. The frequency distributions of crop water productivity are characterised by low coefficient of variation, yielding limited scope for improvement in the agricultural areas under the current cropping systems. Further improvements in water productivity may however be achieved by shifting the crop base from sugarcane to a dual crop and introducing a fallow period from March to May or by converting non-productive rangelands to bio fuel production or other agricultural land uses.

The Krishna River Basin (258,948 km²) in semi-arid southern India is the fourth largest in India in terms of annual discharge, and the fifth in terms of surface area. The basin covers parts of three south-Indian states: Maharashtra (27%), Karnataka (44%), and Andhra Pradesh (29%). After independence (1947), a major national objective was the rapid harnessing of the country’s water resources potential, which resulted in a surge of developments from 1960 onwards and a drastic reduction in river discharge. The massive proposed irrigation schemes promoted interstate water conflicts. The Krishna basin as a whole is now nearly a closed basin.

A major part of the water available in the Krishna basin originates from the humid regions of the Western Ghat Mountains where precipitation exceeds 5000 mm. The Upper Krishna and Upper Bhima catchments served by Western Ghats are, therefore, two very important catchments out of 12 major catchments in Krishna basin. These two catchments contribute significantly to Krishna river flows for downstream use. The Upper Bhima catchment is additionally important for the state of Maharashtra in the context of serving inter-sectoral water demands including hydropower, agriculture and drinking water supplies. Following the increase in utilization, the water released to the main stem of the Krishna from the Upper Bhima catchment has declined by 59% from an average of 8816 Mm³ in 1970–1980 to 3615 Mm³ during 1994–2004, and is mainly concentrated in the monsoon months June–September. During the last 20 years, a shift in agricultural practices towards more water consuming crops, such as sugarcane, took place. The sugarcane area, for example, has almost tripled during this period. An increased competition for water resources between agriculture and the industrial and domestic sectors may lead to a decrease in food production and to environmental degradation. The agricultural sector, being the largest consumer of water, should therefore focus on enhancing the productivity of water through (i) improving the production per unit of water consumed, or (ii) by maintaining the same production while reducing water use. Better knowledge on fresh water depletion and crop production patterns throughout a basin is essential for water managers and policy makers to improve water management in areas where water productivity is low. Traditional water management techniques often focus on water saving at field level by reducing irrigation water allocation to fields. However, a plot level saving may not necessarily lead to ‘real’ water savings at the basin scale, as excess water can be reused downstream. Water management should, therefore, focus on a reduction of water depletion by evapotranspiration and increasing water productivity, as this water is not available for reuse. Therefore, water productivity in this study is defined as the marketable crop yield over the seasonal water use by actual evapotranspiration (ET_act).

Remote Sensing and distributed hydrological models are indispensable tools in objectively quantifying water depletion, water balance components, agricultural yields and water productivity in areas where data is scantily available. This paper shows how the hydrological model Soil and Water Assessment Tool (SWAT) can be applied to simulate the catchment’s water balance, quantify water depletion per land use, and to analyze crop water productivity per agricultural system. Using an innovative methodology, the SWAT model is calibrated using Remotely Sensed ET_act based on the Surface Energy Balance Algorithm for Land (SEBAL) algorithm. This approach is unique in the sense that Remote Sensing is completely integrated in the calibration of a hydrological model. Traditionally hydrological models are calibrated using measured hydrographs. Lack of data and absence of natural flows generally compromise the calibration of such models. The approach demonstrated in this paper provides an innovative methodology to assess water resources in drought prone catchments with limited data availability.

The total simulated evaporative depletion and precipitation indicate that the upper Bhima catchment was nearly closed in 2004–2005. This was confirmed by independent discharge measurements, which show that only 5.6% of the total precipitation was released to the main stem of the Krishna River. If even a water supplying catchment, such as the Upper Bhima, is nearly closed the future for the entire Krishna basin looks grim and a structural rethinking of the planned expansion of irrigated agriculture is warranted. The water productivities are already relatively high and there seems limited scope for further improvement. Based on the analysis two ways to use or allocate water more effectively in the catchment are proposed. Firstly, a diversion from sugarcane to a dual cropping season (similar to SIA) and the introduction of a fallow period from March to May (when precipitation is absent and ET_ref is extremely high) are proposed. Sugarcane is grown throughout the year and its water consumption is highest of all land use classes that were distinguished. A shift towards a dual cropping systems will most likely increase the catchment’s discharge. Of course some caution is warranted and social–economic considerations need to be taken into account. Secondly, by converting non-productive rangelands to rain fed agriculture, the beneficial ET will further increase on the expense of non-beneficial ET. A viable alternative could be the introduction of Jatropha curcas, that is considered to be an excellent source of bio-diesel and that can be grown in wastelands across India (Francis et al., 2005). The Government of India is keen on reducing its dependence on coal and petroleum to meet its increasing energy demand. Promoting the cultivation of Jatropha is a crucial component of its energy policy, which should lead to energy independence by the year 2012. A conversion to other agricultural land uses could also be viable, but requires a careful land use suitability assessment.

The study faced a number of data limitations for which further improvements are recommendable. Firstly, the land use map could diversified to a larger number of crops by more elaborate ground-truthing and inclusion of more remote sensing data sources in the classification. Secondly, a more detailed soil map (including soil physical parameters) would yield more detailed results and a more efficient calibration. Thirdly, the calibration period could be extended in the monsoon season by including other datasets that do not depend on cloud cover (e.g. radar). The SEBAL algorithm depends on radiances in the visible, near-infrared and thermal infrared part of the spectrum, and measurements are hampered by clouds. The current calibration period covers 8 months and a longer time series would be preferable. However, realistic simulations during the dry period from October to May are more important considering that water management issues related to evapotranspiration management, water shortage and irrigation are the dominant hydrological issues relevant to agriculture and water managers. Finally the simulation period should be extended to cover a multi-year periods covering a range of possible climate conditions.

It can be concluded that the integration of Remote Sensing in the calibration of a distributed hydrological model is highly innovative and enhances our insight in the hydrological pathways in drought prone areas with limited data availability. Catchments, such as the Upper Bhima, are difficult to model given the large number of anthropogenic disturbances, such as reservoirs, dams and irrigation canals, that render stream flow unusable for calibration. By using remotely sensed ET_act this problem is overcome and a detailed calibration of a hydrological model was performed and assessed by a number of efficiency indicators. The use of a hydrological model has clear advantages over using remote sensing alone. A model provides insight in the entire hydrological cycle, fluxes between the different water balance components and the crop growth cycle, while remote sensing provides only insight in one component of the water balance at high spatial detail at a specific point in time. A calibrated model also offers opportunities to analyse future scenarios, e.g. land use change and climate change. It is the combination of the strength of both approaches that provides a wealth of possible future applications.