Abstract: Objective　As the most critical north-south geographical boundary and a pivotal ecological barrier in China, the Qinling Mountains play an indispensable role in regulating regional climate, conserving water resources for the South-to-North Water Diversion Project, and maintaining ecological balance. Endowed with unique geographical location, topographical features, and climatic conditions, this region exhibits high sensitivity to climate change. Therefore, grasping the spatiotemporal evolution characteristics of land surface temperature (LST) in the Qinling Mountains not only facilitates an in-depth understanding of the thermal environment evolution process but also offers robust scientific support for long-term ecological change monitoring, ecological protection, and rational management in the region. Methods　To achieve the research objective, MODIS LST data (MYD11A1 product) of the Qinling ecological area from 2003 to 2021, which is based on the split-window algorithm, were acquired based on the Google Earth Engine (GEE) platform. Firstly, the quality control (QC) band of MYD11A1 data was utilized for preprocessing to eliminate interference from cloud cover and low-quality pixels. Subsequently, long-term series of cloud-free daytime and nighttime average LST data were generated by combining image superposition and mean synthesis methods. Concurrently, the normalized difference vegetation index (NDVI), net primary productivity (NPP), dryness index (NDBSI), and transformed humidity components (WET) for 2021 were calculated using MOD13Q1, MOD17A3HGF, and MOD09A1 images on the GEE platform, and Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) data with a spatial resolution of 30 m was also obtained, then all datasets of the five parameters were resampled to a unified spatial resolution of 1000 m to ensure consistency for subsequent analyses. The applicability of MODIS LST data was validated using LST data from 16 ground meteorological stations in the Qinling ecological area provided by the Shaanxi Meteorological Bureau. The results showed that the average absolute error (P) between MODIS LST sampling values and meteorological station observations ranged from 0.51 ℃ to 1.56 ℃ with an average of 0.89 ℃, meeting the accuracy requirements of this study. In terms of research methodologies, firstly, a normalized classification method was employed to categorize LST into five grades, namely, lowtemperature zone (0 ≤ NLST < 0.2), sub-low-temperature zone (0.2 ≤ NLST < 0.4), medium-temperature zone (0.4 ≤ NLST < 0.6), sub-high-temperature zone (0.6 ≤ NLST < 0.8), and high-temperature zone (0.8 ≤ NLST ≤ 1). Then, the Sen's slope estimation method was adopted to analyze the variation trend of LST from 2003 to 2021, and the Mann-Kendall test was applied to conduct significance testing of the trends. Finally, the multiscale geographically weighted regression (MGWR) model was selected to analyze the action scale and effectiveness of LST-influencing factors, with R² (which measures the proportion of the variation in the results explained by the model; the higher the value, the stronger the explanatory power of the model) and AICc (used for model evaluation and selection; the lower the value, the better the model) selected as the key evaluation indicators, because compared with traditional ordinary least squares (OLS) and geographically weighted regression (GWR) models, MGWR model has advantage of allowing the relationship between the dependent variable and different independent variables to vary across spatial scales. Results and Discussion　The research findings are summarized as follows: (1) From 2003 to 2021, the annual average LST in the Qinling ecological area exhibited a weak growth trend, with a rate of 0.0166 ℃/a. Specifically, the annual average daytime LST was 19.12 ℃ , showing an insignificant downward trend at a rate of 0.0148 ℃/a, while the annual average nighttime LST was 7.62 ℃, presenting a significant upward trend at a rate of 0.0479 ℃/a. The interannual variation of LST displayed obvious stage characteristics: a slight increase period (2003–2006), a continuous decrease period (2006–2012), a rapid recovery period (2012–2013), a sharp decline period (2013–2015), and a subsequent slight continuous upward period (2015–2021). Seasonally, the anomalous fluctuations of LST were most pronounced in winter, which contributed the most to the annual average LST variation. (2) During the 19 years, the LST in the Qinling ecological area consistently exhibited a "north-low, south-high" spatial pattern, primarily attributed to the climate differences between the north (warm temperate semi-humid and semi-arid monsoon climate) and the south (subtropical humid monsoon climate) of the Qinling Mountains. While the warming trend showed a "north-high, south-low" spatial distribution for this area during the 19 years. Among the thermal zones, the medium-temperature zone and sub-high-temperature zone dominated, accounting for 43.21% and 37.92% of the study area, respectively, and their proportions increased from 2003 to 2021, while the proportions of the lowtemperature zone, sub-low-temperature zone, and high-temperature zone showed a decreasing trend. (3) As for the relationship with altitude, the annual average, annual average daytime, and annual average nighttime LST on both the north and south slopes showed a significant negative correlation with altitude. As the altitude increased, the LSTs gradually decreased, with a distinct turning point at the altitudes of 2000–2400 m and the minimum LST observed at altitudes of 3200- 3600 m. The vertical decreasing rate of LST on the north slope was slightly higher than that on the south slope, and the nighttime LST decreasing rates on both slopes were lower than the corresponding daytime decreasing rates. (4) Compared to the OLS and GWR models, the MGWR model demonstrated superior performance with R² = 0.902 and AICc = 2583.110. Among the selected influencing factors (DEM, WET, NPP, NDVI, NDBSI), NPP and NDBSI had a positive impact on LST, while WET, NDVI, DEM, and constant terms exerted a negative influence. And DEM exhibited a strong action scale (bandwidth = 130, accounting for 3.6% of the total samples) and the strongest action effectiveness, serving as the most significant negative driving factor for LST. Conclusions　This study systematically analyzed the spatiotemporal variation characteristics of LST in the Qinling ecological area from 2003 to 2021, and quantified the action scale and effectiveness of various influencing factors using the MGWR model. The research results provide a scientific basis for the long-term monitoring of ecological changes, the formulation of ecological protection strategies, and the rational management of the Qinling Mountains, while also offering a reference for related research on thermal environment evolution in similar ecological areas. However, this study still has certain limitations, such as the division of the study area not being sufficiently refined, and the impact of water surfaces on LST not being considered, which may have exerted a certain adverse effect on the research results. In future research, more refined regional division and comprehensive consideration of various underlying surface factors will be implemented to improve the accuracy and reliability of the findings.

Key words: land surface temperature, moderate-resolution imaging spectroradiometer, Qinling Mountains, spatiotemporal variation, gradient descent rate, multiscale geographically weighted regression