GIS-Based Neural Network Modeling to Predict Suitable Area for Beetroot in Sri Lanka: Towards Sustainable Agriculture

Abstract

In this research, neural network technology with a Geographical Information System (GIS) was used to carry out land suitability analysis for beetroot. The suitability evaluation of beetroot (Beta Vulgaris) was based mainly on the methods described by the Food and Agricultural Organization (FAO). Study area of this research was in the upcountry of Sri Lanka. Soil properties, meteorological data, current land use and slope accessibility were considered as important factors to identify potential lands for beetroot. Average annual temperature and precipitation (1961-1990 data), topographic, soil and land use maps of the study area were used for the study. Crop requirement criteria were collected from a literature review and from the Department of Agriculture, Sri Lanka. Paper maps were scanned and screen digitized to prepare thematic layers (maps) and then converted to raster format and reclassified. Reclassified layers were converted to ASCII format. The Levenberg-Marquardt (LM) algorithm was used to perform the Artificial Neural Network (ANN) modeling. Finally, a suitability map was prepared according to the given criteria with four suitability categories; namely, highly, moderately, marginally and not suitable. According to the final suitability map of ANN modeling, 10.43%, 31.66% and 7.96% of lands fell respectively under highly, moderately and marginally suitable categories. The results revealed that there was no, “not suitable” land parcel in the present study area. According to these results, we conclude there is moderate potential for growing beetroot in the upcountry of Sri Lanka.