Analyzing relationships between rainfall and paddy harvest using artificial neural network (ANN) approach: case studies from North-western and North-central provinces, Sri Lanka

Abstract

Purpose: Food and agriculture are frequently affected from on-going climate change. A significant percentage of annual harvest is lost due to extreme climatic conditions in different parts of the world. Sri Lanka is considered as a country which is vulnerable to climate change. Therefore, this research presents a detailed analysis to find out the non-linear relationships between the rainfall and paddy harvest in two major provinces of Sri Lanka. Research Method: North-central and North-western provinces as two major agricultural areas were selected for the study. Rainfall trends were identified using non-parametric Mann-Kendall and Sen’s slope estimator tests. The artificial neural network (ANN) approach was used to establish non-linear relationships between rainfall and paddy yield. Findings: There was no significant (p > 0.05) linear correlation between rainfall amount and the rainfed paddy yield in tested locations. However, no clear relationship between the rainfall and rain fed yield were found in the 14 predefined functions (polynomial, logarithmic, exponential and trigonometric) derived using ANN where the calculated coefficients of determination were less than 0.3. Research limitations: Due to lack of other climate variables such as temperatures, a significant relationship was not observed in this study. Originality/value: We have shown that non-linear artificial neural network approach can be used to study the impact of climate on agricultural production in Sri Lanka.