An Early-Warning System for Endangered Wildlife in Kazakhstan's Renewable-Energy Transition

Presented at the 7th International Conference on Electrical, Communication and Computer Engineering (ICECCE) 2026, Kuala Lumpur, Malaysia, 23 October 2026. This is the author's accepted manuscript.

Kazakhstan is investing significantly in solar and wind energy across the country to hit its targets for renewable energy development; however, the open, dry steppe that this infrastructure is built upon is also crucial habitat for endangered wildlife. Hence, this ecological cost is an unpriced economic externality, only visible after construction. Furthermore, most maps give one confident answer regardless of actual data and survey coverage, while actual nationwide habitat data is sparse, with few records for most species. Hence, this study proposes an early-warning system that flags this risk, giving policymakers insight prior to site selection. The system fits a presence-background species distribution model for the steppe eagle (Aquila nipalensis), evaluates it with spatial-block cross-validation, and finally wraps it in split-conformal prediction at 90% coverage. Each 1 km cell receives one of three verdicts, confident habitat, confident non-habitat, or abstention, which an overlay on solar suitability turns into three planning alerts. The model reaches a spatial-holdout area under the curve of 0.65 and a Continuous Boyce Index of 0.85. It identifies a confident-habitat core of about 2,700 km² and an abstention zone of about 54,000 km², of which 16,000 km² lies on solar-suited land. Across ten background draws the core ranges from 1,300 to 3,200 km², yet 0.33 ± 0.04 of it lies on solar-suited land and 0.86 ± 0.04 within 25 km of the grid. A leave-one-region-out audit gives coverage of 0.91 against the 0.90 target. The system can screen sites before construction, flagging where a survey is required, and transfers to other species and drylands.