Temperatures and Normalized Difference Vegetation Index Forecasting in the Tropical Rainforest: Hala-Bala Wildlife Sanctuary, Thailand using Artificial Neural Networks

摘要

Hala-Bala Wildlife Sanctuary is one of the preserved areas in Thailand. It is situated at the boundary of Thai-Malaysian which covers the area of San Kala Khiri Mountain. The prediction of temperatures and normalized difference vegetation index (NDVI) will help predict fertility changes in this area. This research aimed to study the prediction of temperatures and NDVI in Hala-Bala wildlife sanctuary, Thailand. The data used in this research is time series data for seventeen years from 2001 to 2017. Artificial Neural Networks was used in this research for forecasting these data after that tracking signal was used to find the appropriate period. The predicted result from Artificial Neural Networks with sigmoid activation function (Mean Square Error = 2.17, Mean Absolute Error =1.10) for temperatures in Hala-Bala wildlife sanctuary found that the suitable prediction period is one period only. Besides, the suitable prediction period for NDVI (Mean Square Error = 0.00072, Mean Absolute Error = 0.017) is five periods, respectively.