An Improved COLD Approach for Monitoring Construction Dynamics Using HLS and LULC Data: A Case Study in the New Capital of Egypt
Keywords: Construction Change, HLS, LULC, COLD, Time Series
Abstract. Accurate change detection is essential for understanding land disturbances. The Continuous Monitoring of Land Disturbances (COLD) algorithm is a widely used method for detecting rapid ground changes using Harmonized Landsat and Sentinel (HLS) data. Despite its advancements, change detection accuracy is often limited by frequent ground alterations, such as large-scale construction during urbanization. This study proposes an improved COLD algorithm that integrates land disturbances identified by COLD with publicly available land use/land cover (LULC) maps from Esri, covering the period from 2017 to 2024, to map and quantify construction activities. To evaluate the performance of the proposed method, this study takes the new capital of Egypt as the study area, which is experiencing a surge in national infrastructure projects. We focused on tracking construction dynamics from 2016 to the present. The spatiotemporal detection of land disturbances uses the COLD algorithm with a dataset of 559 images from Harmonized Landsat and Sentinel missions. The identified COLD breaks correspond to transition periods, capturing changes from July of one year to July of the next. Then, two distinct overlapping analyses were performed: first, we aligned the COLD-detected disturbances with the LULC maps of the same year; second, we overlaid the LULC maps of the following year with the COLD results. While both methods yielded similar insights, the latter approach identified a more extensive area classified as undergoing construction, providing a more accurate depiction of progressive development. We validated the results by visually comparing detected construction activities over time and cross-referencing with historical satellite imagery from Google Earth. This approach has proven effective for monitoring and mapping construction changes and holds potential for application in other regions with available LULC maps.