Abstract
To examine the rain-fed Aman rice yield fluctuation due to climatic anomalies overtimes in Bangladesh, we used climate-induced yield index (CIYI), ensemble empirical mode decomposition (EEMD), step-wise multiple regression, isotopic signature, wavelet transform coherence (WTC) and random forest (RF) model. In this work, daily multiple source climatic data which were collected between 1980 and 2017, from 18 weather stations and five atmospheric circulation indices were used for this purpose. The key findings were as follows; by employing principal component analysis (PCA), six temporal variability modes were identified as six corresponding sub-regions with various Aman rice CIYI fluctuations. The Aman rice CIYI in different sub-regions represented a noteworthy 3–4-year quasi-oscillation using the EEMD. The key climate variables (KCVs) including the potential evapotranspiration and cloud cover in September, the minimum temperature in August, and precipitation in July, August, and October were the best rice yield prediction signals in these sub-regions. The results suggest that Aman rice yield could likely decline by 33.59%, and 3.37% in the southwestern and southeastern regions, respectively, if KCV increased by 1 °C or 1%. The RF model suggests that the Indian Ocean Dipole (IOD) significantly influenced the rice yield. Isotopic signatures were employed to confirm the fluctuation and anti-amount effect during the Aman rice-growing period in Bangladesh. The results obtained in this study could be used as a guideline for sustainable mitigation and adaptation measures in managing agro-meteorological hazards in Bangladesh.