Estimating and uncertainty analysis of potential evapotranspiration under climate change in a semi-arid region

Global greenhouse gases increase could be a threat for the sustainable agriculture under climate change due to affecting important meteorological and hydrological variables. Potential evapotranspiration is an effective key factor influences on the production of agricultural crops and lacking an appropriate understanding of its values could endanger food and water securities. Therefore, in this research, the amount of this important variable was estimated under various emission scenarios in general circulation models of the atmosphere (GCMs) up to 2100.
The projected effects of global warming on the values of potential evapotranspiration and the related estimation uncertainties were analyzed in Shiraz city based on the outputs of 15 GCMs under three scenarios of A1B, A2 and B1. The large scale data of GCMs were downscaled using the statistical method of LARS-WG in Shiraz station in three periods of 2011-2040 (initial period), 2041-2070 (middle period) and 2071-2100 (late period). To do so, the model was first calibrated and validated based on daily weather data during base period (1981-2010) and then was applied for downscaling process. For estimating potential evapotranspiration, the capability of empirical models, linear regressions and artificial intelligence methods including adaptive neuro fuzzy inference systems and support vector machines was compared with FAO-Penman-Maonteith method. Then, the amount of potential evapotranspiration in future was estimated using the selected model. Finally, the range of uncertainty for the estimated values of potential evapotranspiration under different GCMs were determined for annual, seasonal and monthly time scales.
Results of t-test and the amount of criteria indices showed that the selected downscaling model is capable enough for estimating precipitation and cardinal temperatures up to 2100. Support vector machines model had the lowest error for estimating potential evapotranspiration based on the values of root mean square error (0.42 mm) and model efficiency coefficient (0.97) indicating its suitability for estimating the parameter in the future climate of Shiraz. Comparing the average results of 35 ensembles of the selected models (15 GCMs under three emission scenarios) as well as the median values for PDFs under the three scenarios of A1B, A2 and B1 for 2011- 2100 period with those of the base period indicted an increase in potential evapotranspiration for annual, seasonal and monthly time scales. The highest increase in potential evapotranspiration under global warming will happen in middle and late periods of 21th century (10.3-15.6 %), high rainy seasons (5.4-31.9 %) and also December, January and February will have compared to the base period. Analyzing the cumulative probability distribution functions showed that the range of uncertainty for estimating annual, seasonal and monthly potential evapotranspiration were, respectively, 180-250, 47.1-132.7 and 19.6-56.4 mm.
Conclusions The finding of this research demonstrated that the increase in atmospheric demand in rainy months could threaten both rainfed and irrigated agriculture through decreasing soil moisture content for spring cultivation and increasing the green water deficit in autumn cultivations. The issue requires planning for coping with this global challenge. Nevertheless, it should be considered that long-term planning will be more risky than short ones due to having higher uncertainties for estimating potential evapotranspiration.