Suprapermafrost groundwater (SPG) plays a critical role in hydrological and ecological functioning of permafrost regions, yet its spatiotemporal dynamics and controlling mechanisms remain poorly understood on the Qinghai-Tibet Plateau (QTP). Here, we integrated in situ observations, geophysical surveys, and machine learning (ML) models (including XGBoost, LightGBM, and RandomForest) to investigate the seasonal variation, drivers, and projections of SPG dynamics in alpine meadow (AM) and alpine wet meadow (AWM) ecosystems. Results showed that SPG tables ranged from -1.1 to -0.1 m in AM and from -1.3 to -0.2 m in AWM during the warm season. SPG fluctuations were primarily driven by thaw depth (TD) and rainfall infiltration and exhibited similar seasonal patterns across both ecosystems. A greater TD was associated with a deeper SPG table, as deeper thawing expanded the unsaturated zone and enhanced vertical drainage, indicating an exponential relationship between TD and SPG table position, and a linear relationship with aquifer thickness. In contrast, rainfall infiltration increased shallow soil moisture and elevated SPG tables, with responses influenced by rainfall intensity, duration, and infiltration pathways. Spatial heterogeneity in SPG distribution was further shaped by vegetation structure and microtopographic variation. Furthermore, ML models projected that mean summer SPG table depths in the 2090s would increase by 0.06 m under SSP126 and 0.64 m under SSP585 in AWM ecosystems, and by 0.37 m under SSP126 and 0.87 m under SSP585 in AM ecosystems. These findings provide new insights into how climate warming affects hydrological processes in permafrost regions of the QTP.

Suprapermafrost groundwater fulfils an important role in the hydrological cycle of the permafrost region. Under the influence of the soil freeze-thaw process in the active layer, the dynamic process of suprapermafrost groundwater is too complex to be fully quantified, which has limited our understanding of the features of groundwater dynamic processes in permafrost regions. To bridge this gap, the dynamic characteristics of the suprapermafrost groundwater level were systematically observed, and pumping tests were performed under different topographic conditions (e.g., altitude, slope orientation, and distance from the river). The results showed that the differences in the heat distribution and recharge source of groundwater at the different altitudes and slope orientations determined the phase and threshold of the variation in the suprapermafrost groundwater movement state. There was a significant Boltzmann function relationship between the groundwater level and soil temperature. The groundwater level in the downslope during melting increased earlier and that during freezing declined later than that in the upslope part during the initial thawing cycle and the initial freezing cycle, respectively. The groundwater level on the shady slope decreased twice as fast as that on the sunny slope at the initial freezing stage. There was a favourable exponential relationship between the hydraulic conductivity (K) and soil temperature in the study area. On the sunny slope, K was higher than that on the shady slope, and K was higher in the area near the river than in the area far from the river. When the melting depth of the active layer reached 2/3 of the maximum depth, K reached its maximum value. The study results also revealed that when the soil temperature was reduced to 1-0 degrees C, a strong linear relationship occurred between K and soil temperature.

To investigate the influences of land surface temperatures (LSTs) on suprapermafrost groundwater discharge, a river valley was selected in a typical permafrost region of Fenghuoshan (FHS) watershed on the central Qinghai-Tibet Plateau. We developed a two-dimensional model to simulate the suprapermafrost groundwater seasonal dynamics controlled by LSTs and the changing trends under a warming climate scenario (3 degrees C/100 year). We calibrated key parameters of our model by the field observations at FHS watershed and analysed the relationship between the different LSTs and the suprapermafrost groundwater discharge dynamics in the active layer. The results show that (a) by changing the permeability of the active layer, the LSTs have a significant effect on the suprapermafrost groundwater discharge. A higher LST causes more suprapermafrost groundwater discharge, resulting in a different discharge pattern and affecting the ability to replenish the nearby river in the permafrost area. (b) Under a warming climate, the most obvious change in the suprapermafrost groundwater occurs in the freeze initiation period (from October to December), and there is a significant increase in the suprapermafrost groundwater discharge rate. This study reveals that the LST has a controlling effect on the seasonal dynamics of shallow groundwater systems in permafrost regions, indicating that the impact of local topography on the suprapermafrost groundwater should not be ignored in suprapermafrost groundwater simulations. Moreover, the warming simulation results demonstrate that the freezing season is the significant transformation period of suprapermafrost groundwater dynamics under future climate change, which can be used to better understand hydrological and ecological process changes in permafrost regions under climate warming.