Ecology, Environment and Conservation Paper

COMPARATIVE ANALYSIS OF INTEGRATING AND COMBINING THERMAL TIRS AND OLI DATA TO SUPERIOR CHANGE DETECTION USING GEOSPATIAL TECHNIQUES

Hayder Dibs, Hashim Ali Hasab, Mustafa Ridha Mezaal and Nadhir Al-Ansar

Abstract

Obtaining up-to-date geospatial information of Land cover (LC) change detection (CD) with high quality and accuracy results is very confusing, especially with a large number of classifiers and different dataset types were adopted in the last decades. The thermal dataset has valuable information to investigate the CD patterns. This study aims to investigate the effectiveness of integrating thermal data, to find the accurate CD methodology by (1) employing noise removing correction models. (2) Images resampling and pan sharpening thermal and visible datasets using Grim Schmidt spectral (GS) method. (3) Combining them to perform. (4) Different image classifications with Mahalanobis distances (MD), Maximum Likelihood (ML) and Artificial Neural Network (ANN) methods applied on two Landsat 8 satellite images captured by Operational Land Imager and the Thermal Infrared Sensors of 2015 and 2020 to produce twelve thematic Maps of LC. (5) Statistical comparison was made between each classifier’s results. The Results proved that applying the ANN approach on integrated and combined OLI and TIRS data can enhance and produce an accurate result of CD compared to other conventional methods, with overall accuracy about 96.31% and 98.40% and kappa coefficients about 0.94 and 0.97 of 2015 and 2020 respectively. However, the ML performance slightly better compared to the MD method. (6) A confusion matrix method was adopted to test the rustles. Finally, the pan-sharpening and combination of the thermal data enhance the accuracy of (5% - 6%) for the employed classifier methods.