Mapping the thermal footprint of a municipal solid waste landfill using remote sensing and artificial intelligence

This work demonstrates the value of combining remote sensing, regression models, random forest (RF) algorithms, and artificial neural networks (ANN) to provide crucial information for landfill management in Jordan. The process of predicting land surface temperature (LST) using linear and nonlinear regression models, ANN, and RF depended on past LST time series retrieved from Landsat images for the years 2000 to 2018. Additionally, the study utilized the normalized difference vegetation index (NDVI), normalized difference moisture index (NDMI), as well as data on humidity, wind velocity, and ambient air temperature. The deployed ANN model exhibited a coefficient of determination of 0.87 and a mean squared error of 6.40*10^-8. Similarly, the RF model accurately identified 93.88% of the LST values. The findings revealed that the LST at landfills was consistently higher than the summer air temperature, and that the LSTs of open landfill cells exceeded those of closed cells. Moreover, the predicted LST values from ANN and RF models surpassed those from linear and nonlinear regression models. Notably, the R^2 value of 0.81 indicates a strong correlation between ANN and RF findings.