Trend Analysis and Forecast of Precipitation, Reference Evapotranspiration, and Rainfall Deficit in the Blackland Prairie of Eastern Mississippi

Abstract

AbstractTrend analysis and estimation of monthly and annual precipitation, reference evapotranspiration ETo, and rainfall deficit are essential for water-resources management and cropping-system design. Rainfall, ETo, and water-deficit patterns and trends at Macon in eastern Mississippi for a 120-yr period (1894–2014) were analyzed for annual, seasonal, and monthly periods. The analysis showed historical average annual rainfall, ETo, and dryness index (DI) in the location to be 1307 mm, 1210 mm, and 0.97, respectively. Monthly rainfall and ETo ranged from 72 to 118 mm and from 94 to 146 mm, respectively, between May and October, resulting in a monthly rain deficit from 22 to 62 mm. Annual rainfall showed an increasing trend of 1.17 mm yr−1 while annual ETo exhibited a decreasing trend of −0.51 mm yr−1, resulting in an annual DI reduction of 0.001 per year. Seasonal trends were found for rainfall in autumn (1.06 mm yr−1), ETo in summer (−0.29 mm yr−1) and autumn (−0.18 mm yr−1), and DI in autumn (−0.006). An autoregressive, integrated, and moving-average (ARIMA) approach was used to model monthly and annual rainfall, ETo, and DI and to predict those values in the future. Low values of the root-mean-square error (RMSE) and mean absolute error (with both statistics being normalized to the mean of the observed values), low values of average percent bias, and low values of the ratio of the RMSE to the standard deviation of observed data, along with values of 1.0 for Nash–Sutcliffe modeling efficiency and the index of agreement, all suggest that the performance of the models is acceptable. The ARIMA models forecast 1319 mm of mean annual rainfall, 1203 mm of mean annual ETo, and 0.82 of mean annual DI from 2015 to 2024. The results obtained from this research can guide development of water-management practices and cropping systems in the area that rely on this weather station. The approaches used and the models fitted in this study can serve as a demonstration of how a time series trend can be analyzed and a model fitted at other locations.