Prediction of potential habitat areas of Mergus merganser in Xizang by MaxEnt Model based on parameter optimization

As a key indicator species for plateau wetland ecosystems, the Common Merganser (Mergus merganser) faces increasing threats from rapid infrastructure development and climate change in the Xizang Autonomous Region. This study aimed to delineate its potential suitable habitat and quantify the relative influence of hydrological, climatic, topographic, and anthropogenic gradients using a rigorously optimized Maximum Entropy (MaxEnt) model. To achieve this, we sought to produce an evidence-based habitat suitability map and identify key environmental drivers, thereby supporting the design of river and lake-centered conservation strategies amidst ongoing environmental changes. We compiled 186 georeferenced occurrence records from 2024—2025 field surveys, literature, and the Global Biodiversity Information Facility (GBIF). A set of 24 environmental covariates representing climate (from WorldClim), topography (from the Geospatial Data Cloud), and human disturbance (distance to roads and water) was assembled at a 1 km resolution. After screening with Pearson correlation (|r|≥0.8 excluded) and ENMTools-based collinearity checks, 13 informative predictors were retained. Using the ENMeval package, we evaluated 48 candidate MaxEnt models with various feature class combinations: linear (L), quadratic (Q), hinge (H), product (P), and threshold (T), and compared regularization multipliers ranging from 0.5 to 4.0 based on minimum ΔAICc. The optimal model (LQH+RM 1.0) was run with ten bootstrap replicates, allocating 75% of the points for training and the remaining 25% for testing; its performance was assessed using mean test AUC and threshold-based omission rates. The tuned model performed excellently, attaining a mean test AUC of 0.971±0.004 and showing minimal divergence between training and test data, indicating strong predictive ability and robustness. Jackknife and permutation importance tests identified four dominant variables: distance to water (26.2%), distance to roads (23.5%), temperature seasonality (17.2%), and maximum temperature of the warmest month (10.0%), jointly explaining 76.9% of the model gain. Response curves revealed that suitability exceeded 0.85 within 0-100 m of permanent water, peaked at approximately 1 km from major roads, declined sharply when annual temperature variability surpassed 4 °C, and approached zero above about 22 °C in summer. Using the mean MTSS threshold of 0.11, the logistic output was reclassified into four suitability tiers. High-suitability areas (scores between 0.50 and 1.00) occupied 1.22×10⁴ km², representing 1.01% of Xizang, and were concentrated along the Nyang River, middle-lower Yarlung Zangbo River, Nu River, Lancang River, and around lakes such as Yamdrok Yumco, Bangong Co, and Puma Yumco. Medium and low-suitability areas spanned 8.99×10⁴ km² (7.47%) and 2.58×10⁵ km² (21.46%), respectively, while 8.43×10⁵ km² (70.06%) were unsuitable. Overlay analysis with protected areas revealed that only 2.65×10³ km² of high-suitability habitat (1.06% of the total protected area) falls within existing nature reserves, indicating substantial conservation gaps. Largely confined to narrow riparian-lacustrine corridors vulnerable to road expansion, we recommend establishing 100 m core protection zones along waterbodies, integrating road-planning buffers, and safeguarding ecological flow regimes to close these gaps. Future research should prioritize long-term monitoring of population dynamics and the impacts of climate change on habitat suitability to refine adaptive management strategies.