Global near-surface temperatures have risen persistently since the late 19th century, with China experiencing an accelerated warming trend in recent decades. To quantitatively assess the spatio-temporal distribution and factors, we constructed a multi-source dataset. ERA5-Land demonstrated superior accuracy during validation and was selected for attribution analysis. Specifically, we analyzed multi-year average temperature (MAT) and annual trends (TAT) across China's sub-regions from 2000 to 2023. Using a Whale Optimization Algorithm-based Random Forest (WOA-RF) model integrated with multi-source data, quantifying the key influencing factors of near-surface temperature (NST) across climatic sub-regions. Our results show that 96.6% of China experienced significant decadal-scale variation at a rate of 0.29 degrees C/decade, with the most pronounced warming (0.3 degrees C/ decade) observed in the sensitively Plateau Mountain climate region. Spatially, the central-southern Tibetan Plateau exhibited significant warming (p < 0.001), while localized cooling trends were identified in the Loess Plateau and Tarim Basin. WOA-RF model revealed that MAT spatial patterns were predominantly governed by Specific Humidity (SH), Elevation (DEM), AOD and Soil Moisture(SM) serving as critical regional regulators. Conversely, while CO2 drove national-scale warming trends (TAT), Cloud Cover (CC) emerged as the primary in monsoonal regions.