| Automatic LULC Mapping Using Google Earth Engine: A Case Study of Egypt's New Delta project |
| Paper ID : 1020-ICRSSSA-FULL |
| Authors |
|
Mohsen Nabil *1, Eslam Farg1, Marwa S. Mostafa2, Nagwan mahmoud mahmoud afify1, Sayed M. Arafat1, Mohamed M. Elsharkawy3 1Agriculture applications department, National Authority For Remote Sensing & Space Sciences (NARSS) 2Data Reception, Analysis and Receiving Station Affairs, National Authority for Remote Sensing &Space Sciences (NARSS) 3Soils Department, Faculty of Agriculture, Beni-Suef University, Beni-Suef, Egypt |
| Abstract |
| Accurate and up-to-date land use/cover (LULC) spatial maps are essential inputs to numerous agriculture, environmental and economic applications. Those maps are also essential for decision-makers to track illegal land-use activities or in planning for new projects. Egypt's government is currently running construction and land reclamation mega projects to provide citizens with housing, food, and jobs. Those projects cause significant LULC changes, which must be quantified accurately and frequently to assess their environmental and economic consequences. Remote sensing satellite images can be used for accurate and updated land cover monitoring. Hence, the current study proposes an automated machine-learning algorithm for assessing LULC changes using freely available remote sensing data. The developed approach relays on Sentinel-2 time series imagers and high-quality LULC datasets recently produced for 2020 to train the ML algorithm on Google Earth Engine (GEE). The proposed approach automatically produced five yearly LULC maps for 2018-2022 with an overall accuracy of 89.8% for the 2018 map and 87% for 2022. The accuracy of the change between the 2018 and 2022 maps was also validated and achieved 83% overall accuracy and 0.81 kappa coefficient. The developed algorithm can be run on GEE for any customized area to produce LULC annual maps automatically without the need for expensive and time-consuming field visits. |
| Keywords |
| Image Classification, Machine Learning, Random Forest, Land use change, Agriculture Expansion |
| Status: Accepted (Oral Presentation) |
