Publications

Maithani, S; Nautiyal, G; Sharma, A; Sharma, SK (2022). Simulation of Land Surface Temperature Patterns Over Future Urban Areas-A Machine Learning Approach. JOURNAL OF THE INDIAN SOCIETY OF REMOTE SENSING, 50(11), 2145-2162.

Abstract
The present study attempts to simulate land surface temperature (LST) as a function of built-up density using an artificial neural network (ANN) algorithm. The LST was retrieved from Landsat thermal datasets for years 2000, 2010, and 2019 using a single channel algorithm. The retrieved LST was validated using downscaled MODIS LST products of corresponding dates and an average root-mean-square error (RMSE) of 1.3 degrees C and 1.9 degrees C was obtained on winter and summer datasets, respectively. The ANN was trained using corresponding built-up density and LST values for the year 2000, 2010, and 2019. The mean absolute error (MAE) of 1.1 degrees C and 0.7 degrees C was achieved on summer and winter training datasets, respectively, while on summer and winter testing datasets MAE of 1.5 degrees C and 0.9 was obtained. Various urban growth scenarios in the year 2029 were simulated using sim weight which is a nonparametric instance-based machine learning algorithm, based on K nearest neighbour procedure. The simulated built-up densities for the year 2029 were input into the trained ANN and LST patterns for the year 2029 were predicted. Analysis of the results showed that for Dehradun city which is located in a humid subtropical climate zone in India, compact urban growth led to lesser areas having LST greater than 42 degrees C in summers, as compared to other urban growth patterns.

DOI:
10.1007/s12524-022-01590-z

ISSN:
0974-3006