Measurements and modelling of evapotransiration to assess agricultural water productivity in basins with changing land use patterns : a case study in the São Francisco River basin, Brazil

Key words: Vineyards, mango, energy balance, evapotranspiration, water productivity, Bowen ratio, eddy correlation, water balance, natural vegetation, latent heat flux, sensible heat flux, biomass, water productivity, remote sensing, water management. . The São Francisco River basin in Brazil is marked by socio-eco¬nomic disparities and environmental vulnerabilities. Water managers in the semi-arid region of the basin are faced with several challenges, such as competition among different water user groups, local over-exploitation of aquifers, climate and land use changes, non-source pollution, erosion, and sedimentation. Water policy makers have to work out strategies for integrated water management, which rely on a proper knowledge base of the physical conditions encountered in the basin. The intensification of horticulture in the semi-arid north-eastern region of Brazil replaces natural vegetation (i.e. caatinga) by irrigated fruit crops. A proper knowledge of the water balance from these different agro-ecosystems is an essential pre-requisite for sound water resources planning in the basin context. Because of the importance of agricultural water management practices on basin hydrology, daily and seasonal actual evapotranspiration were measured in irrigated crops, along with experimental data collection over caatinga. Advanced radiation and energy balance measurements were conducted using the Bowen ratio and eddy correlation energy balance methods. Remote sensing algorithms are potentially suitable for the extrapolation of these local fluxes on a regional scale, and the opportunities of these tools were investigated. The key crop water parameters identified from this data set included actual evapotranspiration, actual transpiration, actual soil evaporation, evaporative fractions, aerodynamic resistances, surface resistances, crop coefficients, percolation fluxes and water productivity. The energy balance measurements on the irrigated fields revealed high evaporative fractions, which pointed out that soils are very wet and that a large majority of the net available energy is converted into latent heat flux. The average crop water consumption in wine grape were found to be 478 mm per growing season, while table grapes show 373 mm per growing season. The seasonal accumulated values for mango orchards were typically 1419 mm. On average the caatinga natural ecosystem evapotranspirated only 533 mm yr-1. The irrigation induced an incremental evapotranspiration of 2.2 mm d-1 or 8,030 m3 ha-1 yr-1. The water balances revealed that systematic over-irrigation is a common practice and that a continuous deep percolation flux occurs. The detailed results allowed expressing water consumption into specific bio-physical parameters, rather than only into more generic crop coefficients that lump together several individual crop water parameters. The stomata of irrigated crops seem to respond very tightly to atmospheric vapour pressure deficit while natural vegetation responds to the rainfall regime. The field results have been used further to calibrate and validate an existing remote sensing algorithm for the estimation of spatially distributed energy balance fluxes: the Surface Energy Balance Algorithm for Land (SEBAL). It was shown that it is required to apply the hot and cold pixel calibration for every individual image. A generic solution for the internal calibration of the sensible heat flux through the linear relationship between surface radiation temperature and vertical air temperature differences adjacent to the land surface could not be found. For daily scale, the values of the instantaneous evaporative fraction needed to be adjusted. The difference between field measurements and SEBAL was 4.4 % and 0.6% for natural vegetation and irrigated mango orchard, respectively, for annual scale. Further to the estimate of depleted water volumes in irrigated horticulture, it was investigated whether the incremental evapotranspiration values are productive. After calibration, the SEBAL algorithm was applied to determine regional scale evapotranspiration and biomass production. The remote sensing tools shows spatial variation of crop water productivity values and detects regions and farms where water can be saved. The net water withdrawal in the Low-Middle São Francisco River basin was also estimated. The biophysical water productivity based on actual evapotranspiration appeared to be around 0.90 L m-3, 2.80 kg m-3 and 3.4 kg m-3 for respectively wine grapes, table grapes, and mangos. The economic water productivities indicated that irrigated fruit crops have around 20 times more value per unit water consumed than irrigated arable crops. The area with fruit crops in the semi-arid region of the Low-Middle São Francisco River basin are expanding mainly with vineyards and mango orchards. The crop water consumption is high due to over-irrigation together with high thermal availability. The water is, however, productively used and creates a boost for the rural economy. The drawback is that agricultural drainage can adversely affect the water quality, and this requires a lower irrigation supply in the near-future. Knowledge of spatially variable actual evapotranspiration can help to optimize the necessary reduction in irrigation supplies