Role of Short-Term Weather Forecast Horizon in Irrigation Scheduling and Crop Water Productivity of Rice

Crop simulation models in conjunction with weather forecasts help in understanding soil-water-plant interactions in real time, and thus in effective management of irrigation water. However, selection of the optimal forecast horizon for use with irrigation scheduling is challenging in the context of uncertainties associated with weather forecasts. This study is aimed at evaluating the effectiveness of the Indian Meteorological Department (IMD) short-term weather forecasts with different forecast horizons (1, 3, and 5 days) in simulating crop water and yield dynamics of rice. Rule-based irrigation is triggered by specifying thresholds on the current day’s ponding depth ($d_o$), precipitation $(P_o)$, and evapotranspiration $({\mathrm{ET}}_o)$ as well as their forecast values ($\sum _{i=1}^n{P}_i$; $\sum _{i=1}^n{\mathrm{ET}}_i$; $n=1,3,5$). The IMD forecasts of temperature, humidity, wind speed, and cloud cover were translated into ${\mathrm{ET}}_o$ forecasts using the Penman-Monteith equation and region-specific crop coefficients. The Soil–Water–Atmosphere–Plant (SWAP) model was modified to simulate soil–water and plant growth conditions by considering the three forecast scenarios along with conventional irrigation (ignoring weather forecast), and a hypothetical perfect 5-day forecast (reference). Experiments were conducted in four paddy fields in a command area of south India for two monsoon seasons (2018 and 2019) to calibrate and parameterize the SWAP model. The accuracy of forecast variables is increased with increase in length of forecast horizon (root-mean-square error = $12.5\pm 0.98$, skill score = $0.6\%\pm 0.03\%$ at 1-day lead to $5.9\pm 2.5$, $22.4\%\pm 7.8\%$ at 5-day lead) due to averaging. Conventional irrigation has resulted in the highest use of irrigation water ($912\pm 12.5\mathrm{mm}$), percolation losses ($\mathrm{1,245}\pm 19.5\mathrm{mm}$), and electricity ($806.5\pm 100.5\mathrm{kW}$), achieving a low yield ($\mathrm{2,331}\pm 138\mathrm{kg}{\mathrm{ha}}^{-1}$). Irrigation scheduling with 5-day forecast horizon outperformed other scenarios (69% water saving and 23% higher yield), and is slightly inferior to the hypothetical perfect forecast. Our results conclude that IMD forecasts, though moderately reliable at multiple lead times, can serve as a valuable tool in scheduling irrigation activities for sustainable management.