Level Estimation Model for Cultivation Suitability based on Meteorological Parameters

Abstract

From the early ages, rice is the staple food of Sri Lankans. Paddy is usually cultivated as a wetland crop, whichever rain fed or irrigated. There are number of factors that affect the productivity of paddy cultivation. These factors can be categorized in to methodology, climatic, economic, physical location. Despite technological advancements imposed to the paddy cultivation related to any of the above factors, climate still remains as the major cause of agricultural productivity. Hence this research focuses on determining the effect of climatic factors to the paddy yield through computer science based approach. In order to estimate unknown paddy yield levels, the relationship between meteorological factors (e.g. temperature (max), temperature (min), rainfall and evaporation) and paddy yield has been modeled using machine learning techniques. As the initial step, a parameter value forecasting model was developed. Forecasted outputs used as input to cultivation level estimation module and based on cultivation levels, seasonal shifts identification process was carried out. RapidMiner data mining tool has been used for model generation and artificial neural networks for learning purpose. Both Forecasting module and cultivation level estimation module showed satisfactory level of accuracy. As a tropical country, Sri Lanka may not have expected highest level of accuracy for parameter value forecasting. However 2 this forecasting model was able to achieve more than 50% of average prediction trend accuracy for all the considered meteorological parameter fore castings. The accuracy can be boosted with the availability of precise and significant data and increasing the number of considering factors. Through this research, the research objectives have been achieved up to satisfactory level by applying computer science concepts and evaluation by the domain experts endorsed the applicability of the research concept.