Comparing different weather generator algorithms for daily temperature

Weather Generators (WGs) are widely used in water engineering, agriculture, ecosystem and climate change studies because observed climatic series have deficiencies related to length, completeness and spatial coverage. WG models can simulate daily temperature data in a few parametric ways. One way is to simulate daily minimum (Tmin) and maximum (Tmax) temperatures using an autoregressive model. Another way consists of the simulation of daily average temperature (Tav) and daily temperature range (R) and then the calculation of Tmin and Tmax indirectly. In this study, four different algorithms were assessed for daily temperature in combination with a well-tested weather generator. M1 and M2 algorithms simulated the daily Tav and R values directly and Tmin and Tmax indirectly. M3 algorithm simulated Tmin and Tmax and M4 algorithm simulated Tmin and Tav directly and the other variables indirectly. The results showed that each algorithm could perform better in simulating primary variables (which are simulated directly). This issue was more considerable in relation to daily R values. M2 overestimated the cross-correlation coefficients of this variable because of the assumption of a strong autocorrelation structure between primary variables in the WG model. M3 and M4 outperformed the other algorithms in relation to most studied indices. This study showed the importance of choosing the best temperature generation algorithm according to the requirements