Quantitative and Qualitative Evaluation of Rainfall Forecasts and Its Propagation into Irrigation Schedules: A Farmers’ Perspective

Despite the availability of India meteorological department (IMD) weather forecasts for more than a decade, their effective utilization in irrigation schedules and crop simulations is least practiced. To bridge this gap, we propose a farmer-centric approach wherein IMD published short-term rainfall forecasts (with 1–5 days lead) are represented in three formats (two qualitative: binary, categorical; and one quantitative: numerical) and further used in irrigation schedules. For each combination of forecast representation and lead time, water balance and crop simulations were performed on a daily basis (120 days) and the results are compared with conventional irrigation (which does not account for rainfall forecast) and perfect forecast irrigation (ideal) scenarios, with the objective to advocate the optimal forecast scenario for use in irrigation. Applicability of the proposed framework is tested on a water-intense rice crop for one crop season (Jun–Oct, 2022). Conventional irrigation has resulted in the highest use of irrigation (1,320 mm) and percolation (1,647 mm) yet achieved a low yield (6.67 t ha−1), highlighting the need for considering forecast-driven irrigation. All forecast scenarios have overestimated the rainfall, recommending low irrigation amounts. Reliability of binary forecast is found to be high at all lead times (accuracy > 0.48, bias < 1.77), however, its further use in water balance has heavily underestimated irrigation amounts (about 40%). Though less reliable (RMSE > 15 mm), numerical rainfall forecast has resulted in optimal irrigation amounts and relatively higher yield (7.1 t ha−1). Our results suggest that data representation of IMD rainfall forecasts has precedence over lead time in scheduling irrigation and simulating crop yield.