การเปลี่ยนแปลงอุณหภูมิพื้นผิวดินทางตอนเหนือของโบโกตา ประเทศโคลอมเบีย ตั้งแต่ปี ค.ศ. 2001-2020

Abstract

This purpose of this research was to examine the seasonal patterns and trends of Land Surface Temperature (LST) and to investigate the predictive models and (term lag and Normalized Difference Vegetation Index (NDVI) that related to LST variability in the upper north of Bogota, Columbia from 2001-2020. The observation data used in this study were obtained from the National Aeronautics and Space Administration (NASA) website as Moderate Resolution Imaging Spectroradiometer (MODIS) LST Data, which was collected every 8 days from January 1, 2001 to December 27, 2020 (a total of 920 data observations) from 9 regions. In this study, cubic spline was used for seasonal patterns analysis and simple linear regression was used for analyzing the trend of the average temperature change for 20 years. The results showed that the average temperature in the upper north of Bogota has been slightly decreasing, at around 0.021 degrees Celsius every year. The data has been divided into 70%-30% proportions for training and testing data sets, respectively. Multiple Linear Regression (MLR) methods and Random Forest (RF) were utilized as the prediction models and factors correlated to LST variability. Root mean square error (RMSE) and R-square were used to compare the predicting performance among constructed models. The results showed that the most important variable in all regions is NDVI. The RF model gained the smallest RMSE from testing both training and testing data sets. The R- square values of MLR model were between 23.68 % to 45.65 % while those of RF model were between 29.90% to 53.29%. However, it cannot be guaranteed that the same performance for each model will be the same for other study areas