WildlifePulse: A Temporal Population Forecasting Framework for Endangered Species Conservation Management

Abstract

The conservation of endangered species is becoming reliant on the need for predicting population loss in advance. Although the current census method of assessing population trends is very useful, it is not only retrospective but also lacks the ability to take into consideration the temporal trends that affect extinction risks. In this study, we develop WildlifePulse as an approach for making population forecasts of endangered species based on their historical demographic, environmental, and habitat factors. The approach leverages time series learning along with ecological feature engineering to make short- and medium-term population forecasts of endangered species. Instead of the traditional static estimation approach, the proposed system is more focused on identifying analysis of trends over a number of years in order to detect trends such as increasing rate of decline or habitat-induced stress factors. The performance of our system was tested using population and environment data of various endangered species. The results demonstrate the effectiveness of temporal modeling in increasing the reliability of forecasts over traditional static approaches, providing conservation managers with a useful tool for decision-making. In addition, the modularity of the proposed framework also facilitates its adaptability across different species according to varying data availability and monitoring efforts. This article introduces WildlifePulse as an effort towards a proactive approach in conservation management driven by forecasting, instead of reaction.