Evaluation of kharif rainfed rice yield predictability using Extended Range Forecast System over Madhya Pradesh, India

Abstract

Abstract The estimation of rice yields using weather observations of three consecutive seasons: 2019, 2020 and 2021, and comparing them with spliced up extended range forecasts of every week from Southwest monsoon seasons JJAS-2019, 2020 and 2021, respectively, gives a glimpse into usability of ERF in future, for the purpose of agriculture. The performance assessment in effectively simulating rainfed rice yields has been done for over 10 districts of Madhya Pradesh, using crop simulation model DSSAT v.4.8, for the seasons 2019, 2020 and 2021, covering JJAS (June, July, August and September) monsoon season. The correlations carried between Yield, Harvest Index (HI) forecasts obtained using category 1 dataset (only observed weather) and category 2 (observed + ERF +Normal) and category 3 (observed + Normal) shows that replacing climatological normal with ERF at vegetative phase of the rainfed rice crop doesn’t show much improvement in its performance, except for season 2019 HI. In reproductive phase, only in 2021 season, yield and HI forecasts obtained with category 2 dataset with ERF splicing, were found to have better correlation coefficients with those obtained with category 1 dataset. Yield forecast obtained by splicing ERF data during October were better correlated to that obtained using observation weather, in ripening phase of the crop. It was also observed that only in 2021 season, stage 2 (End of Juvenile -Panicle Initiation phase) rainfall accumulation with ERF splicing was found to be better correlated with observed rainfall than that without ERF. Rainfall accumulated in Panicle Initiation -End of Leaf Growth phase (stage 3) achieved by splicing ERF data, was significantly correlated to observed rainfall than category 3 weather dataset rainfall without ERF, in season 2019. Rainfall accumulated in grain filling phase (stage 5) by splicing ERF data, was significantly correlated to observed rainfall only in season 2020 at date 14th October as initial (r= 0.922**), and was better than category 3 dataset whose correlation coefficient was 0.663* to observed rainfall.