Genetic Programming for Amphibian Presence Prediction to Support Sustainable Land-Use Planning

Abstract

The application of artificial intelligence-based techniques in spatial planning provides a low-cost and time-saving solution to extensive ecological field surveys. One spatial task is to predict the presence of various amphibian species while planning roads and bridges. At the place of fieldwork, we may use satellite imagery data to extract various features and then use these features to predict the presence of amphibians. Amphibians data set, available on the UCI repository, contains the features, created from satellite imagery and information collected from the inventories of two road projects in Poland. The objective is to forecast the presence of amphibians by utilizing these features. This work proposes a genetic programming-based approach for predicting amphibian species based on features generated from satellite imagery data. Experimental evaluation shows that the proposed method achieves better performance than SVM, standard GP, and existing literature across key metrics such as accuracy, sensitivity, specificity, and AUC. Furthermore, accurate amphibian-presence prediction can assist agricultural planners and environmental authorities in minimizing ecological disturbances during rural infrastructure development, contributing to sustainable land-use planning and biodiversity conservation in agricultural landscapes.