Seasonal variations and challenges in estimating populations and identifying species of Korean ungulates using drone-derived thermal orthomosaic maps

Drones equipped with thermal infrared (TIR) cameras offer significant time and labor savings in estimating wild ungulate populations. However, accurately monitoring forest-dwelling ungulates remains challenging due to their elusive behavior and complex habitat. This study evaluated the feasibility of using TIR orthomosaic maps derived from drone surveys to estimate the population size of three Korean ungulate species: Water deer (Hydropotes inermis), Roe deer (Capreolus pygargus), and Long-tailed goral (Naemorhedus caudatus) in a semi-controlled environment. We generated 15 paired TIR and RGB orthomosaic maps from drone flights conducted in March and June 2024. Ungulate counts from TIR imagery were cross-verified using Red, Green, and Blue (RGB) maps. Two error metrics—counting error and detection error—were calculated using the maximum verified count per month as a proxy for the true population size. Regression analysis indicated that Ground Sample Distance (GSD) was positively associated with counting error, while no clear relationship was found between GSD and detection error. These results suggest that environmental and behavioral factors may influence detection reliability more strongly than image resolution alone. In addition, we analyzed thermal body measurements to explore the potential for species identification. While TIR orthomosaic maps were generally effective for estimating body size, variation in posture—particularly lying positions—substantially affected measurement accuracy and limited their usefulness for distinguishing species. This study highlights both the capabilities and limitations of using TIR orthomosaic maps for ungulate monitoring and offers practical considerations for applying drone-based surveys in more complex natural settings.

Methods
Experimental Design

The data was collected from 15 paired thermal infrared and RGB orthomosaic maps generated from drone flights over a 20,227㎡ fenced deer park housing approximately 19 water deer (Hydropotes inermis), 5 roe deer (Capreolus pygargus), and 4 long-tailed goral (Naemorhedus caudatus)1. Two drone platforms were utilized: the DJI Matrice 350 RTK equipped with a Zenmuse H20T camera and the DJI Mavic Pro 3 Thermal, both capturing thermal images with 640×512 pixel resolution.

Data Collection Methodology

Thermal signatures of ungulates were manually identified and measured using QGIS software from the orthomosaic maps1. All measurements were cross-verified with corresponding RGB orthomosaic maps to ensure accurate species identification1. The dataset includes body measurements from 149 thermal signatures, with animals classified by species and posture (standing vs. lying). Measurements were taken for four distinct body parts: total body length (T1), shoulder width (T2), waist width (T3), and hip width (T4).

This dataset supports research on thermal-based wildlife monitoring techniques and provides baseline morphometric data for Korean ungulate species as measured through drone-based thermal imaging technology.