Qinghai-Tibet Plateau wetting reduces permafrost thermal responses to climate warming

a Key Laboratory of Ministry of Education on Virtual Geographic Environment, Nanjing Normal University, Nanjing 210023, China b Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application, Nanjing 210023, China c School of Geographical Sciences, Nanjing University of Information Science & Technology, Nanjing 210044, China d Northwest Institute of Eco-Environment and Resources, Chinese Academy of Science, Lanzhou 730000, China


Introduction
Permafrost is one of the key components of the terrestrial system, which provides both positive and negative feedbacks to the climate systems (Cheng, 2004;Koven et al., 2011;Zimov et al., 2006). It was reported in Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report that the period of 1983-2012 was very likely the warmest 30-year period in the last 800 years and likely the warmest 30-year period in the last 1400 years in the Northern Hemisphere (IPCC, 2013). The Qinghai-Tibet Plateau (QTP) is the highest and most extensive permafrost region in mid-latitudes across the world, and it is considered to be one of the most sensitive areas to global climate change (Liu and Chen, 2000;Yao et al., 2012). Both observations and simulation studies indicate that the QTP has experienced prominent warming since the 1980s (Kuang and Jiao, 2016;Niu et al., 2004). As a result, substantial degradation of permafrost has been observed on the QTP Cheng and Wu, 2007;Kang et al., 2010). Meanwhile, in recent decades, both warming and concomitant wetting on the QTP were recorded (Gao et al., 2014;Li et al., 2010). Meteorological research shows that the QTP has become wetter as a whole since the 1980s, although there still exist large variabilities in space. For example, annual precipitation (AP) has decreased in the southern and eastern QTP and increased in the central part Zhang et al., 2017;Zhong et al., 2011;Zhou et al., 2019). The wetting climate has been most pronounced since the mid-1990s. Thousands of lakes on the QTP have been expanding due to increased precipitation (Lei et al., 2014;Yang et al., 2018). The wetting pattern has a strong positive correlation to increasing precipitation, with a correlation coefficient of up to 0.97 (Gao et al., 2015). The Community Climate System Model projected a general wetting trend over the entire QTP, with the greatest increase in the occurrence frequency of heavy precipitation. The wetting was projected to be stronger under the Representative Concentration Pathway (RCP) 8.5 than RCP 4.5 along with an increase in temperature (Gao et al., 2017). The coexistence of both warming and wetting processes and their interactions exert substantial influences on permafrost thermal responses to climate change on the QTP. Therefore, it raises important questions about how warming and wetting jointly affect permafrost thermal dynamics on the QTP, and what are the distinct roles of separated warming and wetting processes under this warming-wetting context.
Previous studies have mostly focused on the impacts of climate warming on permafrost changes and concluded that underway permafrost degradation was primarily driven by regional warming climate (Biskaborn et al., 2019;Cheng et al., 2019;Ran et al., 2018;Yao et al., 2019). However, the effects associated with the changing precipitation over the QTP cannot be neglected, as precipitation may greatly impact the freeze-thaw cycle within the active layer and modify thermal dynamics within permafrost. In recent periods, precipitation was reported to increase in high altitudes and arid areas, and to decline in wet regions (Gao et al., 2015). At high altitudes, changing precipitation may play an important role in altering permafrost thermal regimes alongside warming temperature. To date, only a limited number of studies have investigated the effects of wetting on permafrost dynamics. Recently, a relevant study found that an increase in year-around precipitation triggered a cooling effect in the active layer in frozen seasons and a heating effect in thawed seasons . Permafrost monitoring has indicated that increased summer precipitation spells critical impacts for deepening active layer thickness (ALT) (Wu et al., 2015). In another observational study, the increase in precipitation was closely connected to the reduction in ground surface temperature and slowing permafrost degradation (Cai et al., 2018). A study of precipitation extremes suggests precipitation impeded the rising rate of permafrost temperature (Zhu et al., 2017). The rainfall during summers may cool down the active layer such that increased rainfall may mitigate permafrost degradation (Wen et al., 2014). Those studies consistently suggest that wetting has the potential to reduce permafrost warming and potentially mitigate its degradation. However, due to the sparse distribution and limited representativeness of the monitoring sites, they may not be representative of the entire plateau that has a vast area and complex terrain. Consequently, it is imperative to quantitatively assess, through a modeling approach, the effects of wetting on the thermal dynamics of permafrost over the entire QTP.
In this study, we aim to quantify the impacts of warming and wetting on permafrost thermal regimes using a water-heat coupled land surface model (LSM) through numerical experiments to account for various warming and/or wetting scenarios. First, we analyzed the spatiotemporal changes in air temperature and precipitation between two periods, i.e., 1983-1997 and 1998-2012. Then, we assessed permafrost responses to warming and wetting conditions by contrasting the scenarios representing different warming/wetting conditions with a control scenario that represents no warming and wetting trends between two study phases. Finally, we investigated the potential mechanism of the effects of wetting on permafrost dynamics.

Methods
The Noah LSM has long been widely used by the United States National Center for Environmental Prediction to provide land surface parameters for regional climate systems. It integrates the Penman-Monteith equation, multi-layer soil model, canopy evapotranspiration formula (Mahrt and Ek, 1984), and later a simple water balance model, and snow and frozen soil parameterization schemes (Chen et al., 1997;Ek et al., 2003). Many studies have confirmed the capability of Noah LSM in simulating thermal and hydrological processes of frozen ground (Chen et al., 2010van der Velde et al., 2009;Wu et al., 2018;Zheng et al., 2017).
The model used in this study is a modified version of Noah LSM 3.4.1 (Wu et al., 2018), with special treatments for permafrost simulation, including a modified thermal roughness scheme for sparse vegetation and improved parameterization of thermal and hydraulic conductivities to account for gravel and ground-ice content. The details of the modeling procedure, as used in the present study, were presented by Zhang et al. (2019). A scenario-based numerical simulation approach was developed to quantify the impacts of climate warming and wetting on permafrost. The experiment consists of four scenarios: a control scenario (CTL) representing no warming and wetting trends between two stages, three comparative scenarios representing warming and wetting combinations, i.e., warming alone, wetting alone, and historical (indicating both warming and wetting). The historical scenario was built upon the China Meteorological Forcing Dataset (CMFD) (Chen et al., 2011;He et al., 2020), which provides long term reanalyzed time series of seven atmospheric variables at a 3hr time step. It represents the real warming and wetting climate conditions, under which permafrost changes over the QTP have occurred. The resting scenarios were constituted by intentionally modifying air temperature and precipitation data while retaining the other atmospheric variables in the historical scenario. By contrasting the outcomes of the warming/wetting scenarios and the CTL, the impacts of warming/wetting on permafrost can be identified. Gao et al. (2014) reported an abrupt change in annual air temperature and precipitation over the QTP around 1998. A similar break point was also detected from the CMFD. Therefore, we divided the entire study period from 1983 to 2012 into two phases preceding and succeeding 1998, with 1998 as the pivotal year. The dataset covering the entire study period was separated into two parts, i.e., pre-(1983-1997) and post-(1998-2012) phases. The temperature time series was broken up into two segments: T pre and T post , denoting temperature data in pre-phase and postphase, respectively. Since historical air temperatures were increasing through the entire period, T pre and T post are able to represent low and high levels of air temperature that have already occurred, respectively. Likewise, precipitation data were separated into P pre and P post for the two phases. P post represents a condition of more precipitation than P pre . As shown in Table 1, three hypothetical scenarios were created by assembling different temperature and precipitation segments for the two phases while keeping other driving variables unchanged. The CTL that assumes no changes between the two phases has temperature and precipitation data in the post-phase identical to the counterparts in the pre-phase. The warming alone scenario duplicates the historical scenario except that precipitation in the post-phase is replaced by the counterpart in the pre-phase. Because T post is warmer than T pre , the warming alone scenario simulates an exclusively warming condition. Similarly, the wetting alone scenario constrains temperatures in the post-phase identical to the pre-phase and retains natural precipitation.
Because all scenarios have the same temperature and precipitation conditions in the pre-phase on purpose, we focused on the outcomes in the post-phase. The CTL outcomes can serve as a contrast benchmark as there are no phase-wide variations in temperature and precipitation between the pre-and post-phase. By contrasting the simulation results from the warming/wetting scenarios to the CTL, the impacts of warming and wetting and their sole contributions can be determined. Permafrost dynamics from 1983 to 2012 were simulated for the four scenarios by the modified Noah LSM after a 32-year spin-up, for which the years 1979-1982 are used. The temporal and spatial resolutions are 3 hr and 0.1 • × 0.1 • , respectively. The modeling depth amounts to 15.2 m with 18 soil layers. The changes were measured in terms of some key thermal indicators for permafrost, which include ALT, temperature at the top of permafrost (TTOP), and permafrost area. Specifically, a grid cell is identified as underlying permafrost where the maximum monthly mean temperature remains below 0 • C for over 24 months in at least one soil layer as per the definition Table 1 Setup of scenarios by assembling air temperature and precipitation segments extracted from the historic records.

Scenarios
Scenario segments Pre-phase (1983Pre-phase ( -1997  ↑T post _P post ↑ Note: The control, warming alone, and wetting alone scenarios are hypothetical. Temperature segments, T pre and T post , and precipitation segments, P pre and P post , are extracted from the historical records. The subscripts, pre and post, represent the segments from pre-phase and post-phase, respectively. The symbol ↑ denotes a higher temperature/precipitation condition compared to the counterpart in the CTL. of permafrost. The ALT is determined by linear interpolation between two adjacent depths above and below the 0 • C isotherm to find the maximum depth in the thawed state. Similarly, the TTOP is estimated as an average temperature at the depth of ALT. For simplicity, the ALT and TTOP time series were obtained by averaging over the regions that are simulated as permafrost in the corresponding period under the CTL. The Albers equal-area conic projection was applied to ensure no distortion in calculating permafrost area. The differences between the results from warming/wetting scenarios and the CTL reflect the effects of temperature/precipitation changes on permafrost. By subtracting the CTL results from the three scenarios, we obtained the differences and drew the corresponding boxplots for the selected indicators. For example, if positive differences in a given indicator are shown in the boxplots for the warming alone scenario, it indicates that air tem-perature warming exerts positive impacts on that indicator. We also drew spatial difference maps relative to the CTL outcomes. The contributions of one-degree warming and 100 mm wetting to the changes in ALT and TTOP were computed. The contribution of warming/wetting is defined as the ratio of the difference in the indicator mean over the difference in the temperature/precipitation mean.

Spatiotemporal characteristics of climate warming and wetting over the QTP
The mean annual air temperature (MAAT) over the QTP is characterized by a rapid upward shift during 1983-2012 (red line in Fig. 1a). The warming rate was 0.49 • C/decade, broadly consistent with site observations. It was about three times greater than the global mean (Duan and Xiao, 2015). Such a rate has never been greater in any past periods on the QTP. The period 1983-2012 was also very likely the warmest 30-year period in the last 800 years in the Northern Hemisphere (IPCC, 2013). In contrast to the global warming hiatus in 1998-2012 (Huang et al., 2017), from 1998 onward steady multi-decadal warming continued to exist over the QTP. The strong El Niño event in the Northern Hemisphere beginning in 1998 was reportedly leading to an increase in global average temperature (Karl et al., 2015). The increasing trend of temperature was particularly evident on the QTP. In comparison to the phase of 1983-1997, the average MAAT in 1998-2012 increased remarkably by 0.89 • C. The difference map of MAAT between the two phases shows warming trends prevailing throughout the entire QTP (Fig. 1b). Widespread positive differences colored in red, i.e., warming trends, can be observed in most regions, except for some small portions in the northwest and southeast QTP where cooling trends were present. Both temporal and spatial patterns indicate that the QTP has experienced prominent warming after 1998.
Unlike prevalent warming that was widely recognized, precipitation change patterns over the QTP appear more complicated. The AP averaged across the QTP was significantly increasing from 1983 to 2012 at a rate of 45.8 mm/decade (green line in Fig. 1a). Especially after 1998, the AP has maintained at a high level with 2005 being the wettest year. The average AP in 1998-2012 was higher by 86 mm than the average of 1983-1997. Abrupt changes around 1998 can be detected by both MAAT and AP. This corroborates with the findings of the same pivotal year in the previous studies (Gao et al., 2014;Kuang and Jiao, 2016;Sun et al., 2020). Precipitation changes between the two phases varied in space (Fig. 1c). It is evident that the spatial differences occurred were mostly positive, implying that most QTP became wetter in the second phase. This wetting correlates with an increase of water vapor convergence due to circulation changes . More specifically, the AP intensively increased on the northern QTP, with the largest increase appearing on the central QTP, known as Qiangtang High Plain. In some areas of the southeastern and southern QTP, the AP declined to some extent. By superimposing a dry-wet climate division on the AP distribution (Fig. 1c), it is obvious that the change patterns of AP differed by climate zone. It was becoming wetter in arid and semi-arid zones, whereas relatively drier in some areas of semi-humid and humid zones. In general, the QTP tended to be wetter in the second phase. These results confirm that the QTP has experienced pronounced warming and concomitant wetting since the mid-1990s although there were still regional exceptions. Fig. 2 illustrates the changes in permafrost indicators (ALT, TTOP, and permafrost area) simulated under three warming/wetting scenarios and the CTL scenario. The results show pronounced distinctions in terms of permafrost responses to warming and/or wetting conditions. Positive differences in ALT, TTOP, and loss of permafrost area relative to the CTL consistently appeared in the warming alone scenario, suggesting that more permafrost degradation had occurred than in the CTL. Meanwhile, persistent negative differences that occurred in the wetting alone scenario indicate that more permafrost has been developed than in the CTL, attributable to more precipitation in the wetting alone scenario than the CTL. In terms of all three indicators, the warming alone curves and the wetting alone curves depart from the CTL curves in an opposite direction. It signifies that permafrost over the QTP has opposite responses to climate warming and wetting. In other words, climate warming thermally degraded permafrost whereas climate wetting favored permafrost formation. Therefore, as expected, the historical scenario with both warming and wetting in place produced in-between results as a consequence of offsetting each other's impacts. Under this scenario, a short-term reversal of all indicators has occurred after 2000, i.e., ALT andTTOP decreased in 2000-2004 Fig. 1a. Alongside the positive role of warming in permafrost degradation, these comparisons reveal that increased precipitation exerts an important role in modulating permafrost thermal regimes.

Permafrost responses to climate warming and wetting
Spatially, in most areas over the QTP, the trends in terms of the differences in permafrost indicators under the three warming/wetting scenarios (Fig. 3) are consistent with those discovered from the temporal analyses (Fig. 2). To better represent the spatial variations in permafrost area, we used transition of frozen ground type as a proxy indicator. Conversion of previous permafrost into seasonally frozen ground is termed as positive degradation. Positive differences in all indicators against the CTL dominated in the warming alone scenario, while prevalent negative differences were observed in the wetting alone scenario. However, the distributions of positive and negative differences were comparable in the historical scenario. By design, the warming alone scenario represents higher temperature conditions than the CTL. Considerable permafrost degradation has occurred over the QTP in response to a warmer climate than in the CTL, as evidenced by thickened ALTs, higher TTOPs, and extensive permafrost losses across the plateau (top row in Fig. 3), with exceptions in some regions on the northwest QTP that rather affected by regional cooling. Although the visible permafrost degradation was not much in space (Fig. 3g), there was intensive thermal degradation within the permafrost body on the central QTP that is an alpine continuous permafrost region, where ALT and TTOP were simulated to dramatically increase in the warming alone scenario. The spatial patterns obtained for this scenario are highly correlated with the MAAT differences in space (Fig. 1b). In contrast, thinner ALTs, lower TTOPs, and increased permafrost areas were simulated over the most portions of QTP in the wetting alone scenario (middle row in Fig. 3), as more precipitation took place in the wetting alone scenario than the CTL. Moreover, distinct from warming induced patterns, precipitation induced patterns show notable regional differences. The overall spatial patterns broadly match the summer precipitation variabilities in space (Fig. 1c). Because precipitation over the QTP is mostly concentrated in summer with less than 10% in winter (Wang et al., 2018), the spatial variabilities of summer precipitation bear great resemblance to the AP pattern (Fig. 1c). Differing from the fabricated sole factor scenarios, the historical scenario characterizes the conditions of simultaneous warming and wetting relative to the CTL. The coexistence of extensive positive and negative differences in space can be observed (bottom row in Fig. 3). While wetting can offset the impacts of warming, in a certain region one factor might dominate. The most negative differences appeared in the northwest QTP, where the regional cooling and the wetting, both of which are conducive to permafrost formation, multiplied the differences as compared to the CTL. While the most conspicuous thermal offset occurred on the central QTP in Fig. 3d, the impacts of wetting to permafrost in the same region were also huge (Fig. 3e). The consequent differences (Fig. 3f) under the historical scenario remained positive on most central QTP. We can imagine permafrost on the central QTP would have undergone unprecedented thermal degradation in response to climate warming assuming no wetting in the second phase. Therefore, by contrasting the outcomes of three warming/wetting scenarios, it was discovered that climate wetting on the QTP inhibits permafrost thermal degradation when exposed to a warming climate. The control scenario (CTL) serves as a contrast benchmark, representing no warming and wetting between the two phases. The warming/wetting alone scenarios reflect the separated impacts of climate warming/wetting relative to the CTL. The historical scenario represents historical permafrost changes under real climate conditions, which are with both warming and wetting. Differences in ALT, TTOP, and permafrost loss, obtained by subtracting the CTL results from the results of the warming/wetting alone and historical scenarios during 1998-2012, are expressed in the boxplots. The boxes represent 25-75% quartiles and the whiskers are 1.5 interquartile ranges from the medians that are shown in black lines in the boxes. The gray dots indicate mean values, and the star symbol (*) indicates outlier values.

Permafrost responses to wetting in dry-wet climate zones
We noticed that the spatial patterns for the differences in permafrost indicators under the wetting alone scenario (middle row in Fig. 3) show distinctive regional differences. It is suggestive of differing permafrost responses to wetting in dry-wet climate zones, which to our knowledge has never been reported in previous studies. The most sensitive responses, indicated by the largest negative differences in ALT and TTOP ( Fig. 3b and 3e), were found in the arid zone (AZ), followed by the semi-arid zone (SAZ). The AZ and SAZ account for over 60% of the total permafrost area on the QTP, with most of them located in continuous permafrost regions. Despite most of the permafrost in those zones remained stable in type, a lot of seasonally frozen ground in the transition areas has converted into permafrost driven by increased precipitation in this scenario (Fig. 3h). Taken together, these results suggest that the increases in summer precipitation exert very strong cooling effects on frozen ground in the AZ and SAZ. By contrasting the outcomes from the three warming/wetting scenarios, it is clear that the substantially increased precipitation in 1998-2012 prevented massive permafrost degradation resulted from simultaneous climate warm-ing in the AZ and SAZ. Through this simulation, we confirmed the critical role of wetting in regulating permafrost dynamics over the QTP, particularly in the vast AZ and SAZ.
The effects of wetting on frozen ground diminished gradually from dry areas to wet areas. Fig. 4 presents the time series of differences in ATL, TTOP, and summer precipitation ( P s ) for the four dry-wet zones. The differences were computed by subtracting the CTL results from the wetting alone scenario. The more divergence from the zero horizontal line, the more impacts received from wetting. Both ALT and TTOP vary considerably in response to the changed summer precipitation in the four zones. The most obvious cooling effect was observed in the AZ, where ALT and TTOP declined substantially. With P s increasing progressively, the negative differences for ALT and TTOP became larger and the maxima were reached in the wettest year 2005. When P s exceeded 100 mm, the cooling effect basically stopped growing in the AZ. The effect was less intense in the SAZ even though P s in this zone became greater than in the AZ after 2005. Therefore, the strength of the cooling effect depends on not only the amount of increased precipitation, but more importantly the dry-wet zone where permafrost underlies. In the SHZ, P s continued to increase over time with a magnitude less than in AZ and SAZ, the cooling effect on permafrost was rather mild. Unlike the three zones, P s in the HZ oscillated around the zero line with a generally declining trend over the study period. As a result, small positive differences in ALT and TTOP, contrary to those in the other three zones, were found in the HZ. The trends between permafrost indicators and P s are negatively correlated. From these results, it can be concluded the increases in summer precipitation exert strong cooling effects on permafrost thermal regime and slow down thermal degradation in arid and semi-arid areas, whereas in humid or semi-humid areas the effects are rather limited or even opposite.

Quantification of permafrost responses to warming and wetting
The impacts of warming and wetting on permafrost thermal indicators (ALT and TTOP) were quantified as listed in Table 2. As the impacts of wetting vary in climate zones, the quantification was also made for each zone separately. The results show, the ALT increases by 0.46 m and the TTOP by 0.53 • C in response to an average of one degree increase in air temperature in the permafrost regions over the QTP. Taking the QTP as a whole, wetting with 100 mm more summer precipitation results in a reduction of the ALT by 0.35 m and the TTOP by 0.36 • C. In the AZ, the ALT declines by as much as 0.75 m and the TTOP by 0.71 • C, in response to additional 100 mm summer precipitation. In the SAZ and SHZ, the reductions are less, i.e., 0.24 m in ALT and 0.22 • C in TTOP for the SAZ, and 0.10 m and 0.19 • C for the SHZ. Note that the difference of summer precipitation in the HZ was minimal between the post-phase and the pre-phase, and the HZ covers a small fraction (∼14%) of the total permafrost area, most of which is of the type of island permafrost. Thus, we excluded the HZ from quantifying contributions. Obviously, the strongest responses of wetting occur in the AZ, and the next is in the SAZ. The quantification confirms that while climate warming induces permafrost thermal degradation over most parts of the QTP, the concomitant wetting can effectively reverse the degradation in dry areas (AZ and SAZ).

Mechanisms of the effects of climate wetting on permafrost
In cold and dry permafrost areas, precipitation as a major source of soil moisture can alter soil thermal regimes (Lupascu et al., 2014). Percolating rainfall delivers heat to the depth, influencing thermal and hydrological processes in the active layer and permafrost. Our modeling studies show that increases in summer precipitation have strong cooling effects on permafrost thermal dynamics in arid and semi-arid areas, as evidenced by remarkable declines in ALT and TTOP ( Fig. 3 and 4). This may be explained by accompanying heat exchanges when rainfall reaches the soil surface, infiltrates and percolates to the deep soils. More precipitation increases soil surface wetness and soil moisture, leading to increasing soil surface evaporation, which consumes heat energy and reduces the amount of heat transporting to the deep. Rainfall heat exchanges with the surrounding soils in a form of convection. At the time when phase changes are occurring within the active layer, heat transfer will be tremendously increased by heat convection. Meanwhile, increased soil moisture can absorb abun-  represents average increments between two phases, and the subscripts pre and post represent pre-phase (1983-1997) and post-phase (1998-2012), respectively. P s and T represent the averaged increments of P s and MAAT in post-phase relative to pre-phase, respectively. dant heat as liquid water has a high specific heat capacity and diminish heat transfer to the deep. When heat is consumed more by evaporating and percolating liquid water than that transferring to soils, it will put a cooling effect on permafrost. As simulated in this study, the cooling effect prevails in arid zones. Such effect of precipitation on permafrost has been already reported in some previous observation-based studies such as Zhang et al. (2001). Our modeling study, from a regional perspective rather than a point perspective as in previous studies, confirms the cooling effects that abundant precipitation may bring to mountainous permafrost, and more importantly, highlights the spatial variability of cooling effects in dry-wet climate zones, with being most prominent in arid and semi-arid zones. However, when a large amount of rainfall percolates into the deeper soils, it may cause an opposite, warming effect on permafrost. Summer rainfall usually comes with higher temperature than in the deep soils and the convection will effectively increase the soil temperature when the liquid percolates through the soil. Even if we do not consider the convective heat transfer as in the Noah LSM, more liquid water accumulating at the bottom of the active layer will alter the process of phase change occurring there. In the freezing period, bidirectional freezing will take place. The upward freezing occurs at the interface between the thawing soils and permafrost and this process will release massive latent heat, thus affecting the thermal regime of the permafrost body. The Noah LSM assumes the liquid water holds the temperature equal to the ambient soils. Therefore, after heavy rainfall rapidly reaches the base of the active layer in summer, those extra water contents will serve to warm the permafrost body during the subsequent freezing period. It is liable to occur in the event of heavy rainfalls in summer. In the semi-arid areas over the QTP (Fig. 5d), we observed in our simulations the presence of higher soil temperatures at deeper soils (6 m and 10 m) in the wettest year of 2005 under the wetting alone scenario (Fig. 5a and 5b), while conversely both ALT and TTOP measured at a relatively shallow depth (2.5 m on average) were simulated to be reduced. By examining the driving forces, the warming effect is related to the occurrence of heavy summer precipitation in the semi-arid zones (Fig. 5c). It implies the existence of dual effects on permafrost in semi-arid areas, i.e., a cooling effect in shallow depths and a warming effect in deep depths. The warming effect of excessive precipitation can be supported by some permafrost site observations (Luo et al., 2016), where soil temperatures have increased while no apparent local warming trends were found. Luo et al. (2016) attributed the warming permafrost to increasing summer precipitation. A similar finding was also provided by a simulation work performed by Subin et al. (2013). These findings together provide evidence that increasing summer precipitation in semi-arid high altitudes is likely to have dual effects on permafrost thermal regimes. Through our experiments, it can be found that the cooling effect by increasing precipitation is not obvious in relatively wet areas. However, permafrost in dry areas is more susceptible to wetting. As such, future wetting could mitigate permafrost responses to climate warming. The complication behind the mechanisms calls for more in situ observation as well as theoretical studies towards understanding and assessing the vulnerability of permafrost to future climate warming and wetting. The Noah LSM used in this study is limited in the lack of considering convective heat transfer and the role of convection in exchanging heat in a freeze-thaw cycle is worthy to further study. In addition, the current study is based on a single driving dataset. The uncertainties associated with the driving data and model parameters, in particular, the soil parameters, may bias the results.

Conclusions
This study provides a new perspective on permafrost responses to climatic warming and wetting through numerical experiments. From the results, we have drawn the following conclusions. The QTP has experienced pronounced warming and wetting especially after the mid-1990s. Permafrost over the QTP has opposite responses to climate warming and wetting. While climatic warming leads to permafrost thermal degradation over the QTP, wetting prevents from the degradation, especially in the vast arid and semi-arid areas. The increases in summer precipitation exert strong cooling effects on permafrost thermal regimes in arid and semi-arid areas, whereas in wet areas the effect is rather little. It was estimated that one-degree warming of air temperature in permafrost regions of QTP leads to an increase of 0.46 m in ALT and 0.53 • C in TTOP. An increase of 100 mm in summer precipitation causes a mean reduction of 0.35 m in ALT and 0.36 • C in TTOP. Regional differences are remarkable in terms of permafrost responses to wetting. The largest decrease of 0.75 m in ALT and 0.71 • C in TTOP in response to 100 mm additional summer precipitation occurred in the arid zone, followed by 0.24 m and 0.22 • C in the semi-arid zone, 0.10 m and 0.19 • C in the semi-humid zone. We also found that increased summer precipitation in semi-arid high altitudes has dual effects on permafrost dynamics. When heavy summer rainfall occurs in those areas, it imposes a cooling effect on the active layer and a warming effect on the underlying permafrost body. As shown by the experimental simulations, apart from climatic warming, wetting is also a critical factor regulating the current thermal regimes of permafrost over the QTP and it will become even more important when precipitation has been projected to continuously increase in the future. Insights into these aspects are expected to contribute a better understanding of the roles of warming and wetting in altering permafrost thermal responses.

CRediT authorship contribution statement
G.Z. conceptualized this research, performed simulations, analyzed data, and wrote the original draft. Z.N. conceived the idea, supervised the study, reviewed, and edited the manuscript. Y.L. and G.Z. conducted climatic zonation. L.Z. and G.C. provided very valuable suggestions for this paper. All authors reviewed and accepted the manuscript.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.