Ultralight aircraft for conservation: AI-assisted aerial census reveals higher-than-expected mammal numbers in Zah-Soo National Park, Chad

Biodiversity in Central Africa is declining sharply, even within protected areas, making reliable and cost-effective wildlife monitoring essential. In open landscapes, traditional aerial surveys using light aircraft provide valuable data but are costly, risky, and prone to observer bias. Drones and very-high-resolution satellite imagery offer alternatives but are limited by spatial coverage or cost. Ultralight aircraft equipped with ultra-high-resolution (UHR) cameras represent a promising intermediate platform. This study assesses the suitability of ultralight aircraft for total aerial photographic censuses of protected areas. The study area was Zah-Soo National Park (815km², Chad). Flights were conducted using systematic transects, with two UHR cameras acquiring 72,892 georeferenced 61-megapixel images covering the entire area. The ultralight aircraft maintained the target flight altitude of 530ft (160m, SD=7m) and followed planned transects (mean deviation of 12m). Images were processed with a pretrained deep learning model (HerdNet), followed by human verification. Compared with the 2024 observer-based census, the 2025 census revealed higher minimum abundances for several wildlife species, particularly the roan antelope (129 vs.29), as well as high numbers of livestock along inner park borders. Elephant abundances were comparable (104 vs.101), emphasizing detection challenges in highly vegetated areas. Our findings demonstrate the feasibility of using ultralight aircraft as an aerial imagery platform for large-scale wildlife monitoring, allowing large mammal detection, enabling safer high-altitude flights, and providing a cost-effective alternative to light aircraft while surpassing drone coverage. Wider adoption will depend on local training, data management capacity, and further refinement of the already efficient automated processing pipeline.