Investigation of Thermal Effects of Lakes on Their Adjacent Lands Across Tibetan Plateau Using Satellite Observation During 2000 to 2022

Highlights

What are the main findings?

• In contrast to the “cold island” effect observed in summer (1.3 km), the “warm island” effect in autumn extends over a much larger area (5.5 km).
• A total of 79.2% of the lakes experienced declining LLTDs during 2000–2022.

What is the implication of the main finding?

• Atmospheric boundary layer stability contributes to the lake thermal effect.
• Land responds more rapidly to climate than lakes.

Abstract

Understanding the regulatory effects of lakes on land surface temperature is critical for assessing regional climatological and ecological dynamics on the Tibetan Plateau (TP). This study investigates the spatiotemporal variability in the thermal effect of lakes across the TP from 2000 to 2022 using the MODIS land surface temperature product and a model-based lake surface water temperature product. Our results show that the lake–land temperature difference (LLTD) within 10 km buffer zones surrounding lakes ranges from −2.8 °C to 3.4 °C. A declining trend in 79.2% of the lakes is detected during 2000–2022, with summer contributing most significantly to this decrease at a rate of −0.56 °C per decade. Assessments of the spatial extent of lake thermal effects show that the “warm island” effect in autumn (5.5 km) influences a larger area compared to the “cold island” effect in summer (1.3 km). Furthermore, southwestern lakes exhibit stronger warming intensities, while northwestern lakes show more pronounced cooling intensities. Correlation analyses indicate that lake thermal effects are significantly related to lake depth, freeze-up start date, and salinity. These findings highlight the importance of lake thermal regulation in heat balance changes and provide a foundation for further research into its climatic and ecological implications on the Tibetan Plateau.

1. Introduction

Compared to terrestrial surfaces, lakes exhibit distinct physical properties—lower albedo, higher heat capacity, and lower surface roughness [1]—with heat transfer primarily driven by turbulent mixing in the water column [2]. These dynamic and thermal properties of lakes influence local radiation budgets and energy balances [3,4,5], and modify surface–atmosphere interactions [6,7]. Owing to their large heat capacity, lakes buffer climatic fluctuations and play a significant role in modulating regional climate [8,9]. At the same time, climate warming accelerates lake evaporation [10], supplying more moisture for the atmosphere, and intensifying lake-effect precipitation events [11,12]. Furthermore, diurnal temperature contrasts between lakes and adjacent land areas drive thermally induced breeze circulations [13,14,15]. Therefore, understanding the thermal effects of lakes on adjacent land surfaces is crucial for regional climatological and ecological research [16,17,18]. Under global climate change, lakes worldwide are experiencing warming trends that lead to variations in lake ecosystems, microclimates along shorelines, and surrounding land temperatures, amplifying their impacts on coastal zones. The Tibetan Plateau (TP), which hosts more than 1100 alpine lakes larger than 1 km² situated above a 4000 m elevation [19], is particularly sensitive to climate change. Over the past few decades, TP lakes have undergone significant changes in water volume and thermodynamics. Although these lakes may play an important role in regulating local climate [20], the magnitude and underlying mechanisms of the thermal effects remain inadequately quantified [21].

Lake thermal effects—including the lake cooling effect (LCE) and lake warming effect (LWE)—remains poorly understood across the TP. A common method to measure the LCE/LWE is to identify the cooling/warming effect boundary of the lake, as well as the temperature difference between the cooling/warming effect boundary and the lakeshore [22,23]. The lake–land temperature difference (LLTD) serves as a critical index reflecting the thermal effects of lakes on adjacent terrestrial areas. Variations in LLTD are closely linked to energy exchanges between water bodies and their surroundings [24,25], rendering it a valuable parameter in climate modeling and environmental management [26]. Moreover, characterizing the spatial extent of lakes’ impacts on land surface temperatures is essential for understanding complex aquatic–terrestrial ecosystems interactions [27,28,29,30,31]. Current research on lake thermal effects presents two divergent perspectives: some studies suggest that lakes in temperate and high-latitude regions absorb more solar radiation in summer and release heat in autumn and winter, leading to lake surface air temperatures up to 17 K higher than the temperatures of surrounding land [4]. Alternatively, other studies suggest that lakes reduce surrounding temperatures [32,33,34,35] through evaporative cooling—for instance, surface temperatures of some temperate and high-altitude lakes can be 5 K lower than the temperatures of surrounding land in summer [4]. Thus, studies of lake thermal effects in climatically sensitive regions, such as the TP, is needed.

Remote sensing enables large-scale monitoring of surface temperature over both land and water, which is especially valuable in inaccessible or harsh environments such as the TP [36]. Consequently, it serves as a powerful tool for investigating the thermal effects across the region. Thermal infrared remote sensing is widely used to retrieve land surface temperatures by measuring thermal emissions via satellite sensors (e.g., MODIS, Landsat, Sentinel-3), followed by atmospheric correction and surface emissivity estimation. Common algorithms include the single-channel [37], split-window [38], and multi-angle methods [39], with the split-window algorithm being widely adopted for its high accuracy and broad applicability [40,41]. Additionally, machine learning techniques (e.g., random forest, neural networks) are increasingly applied to improve retrieval accuracy under heterogeneous surface conditions [42]. Passive microwave remote sensing shows advantages in terms of cloud-penetrating capability, although at coarser spatial resolution, it is suitable for large-scale analyses [43,44,45]. Some studies have merged thermal infrared and microwave data to improve the spatiotemporal resolution of surface temperature inversion [46]. Regarding lake temperatures, satellite-based observations have made significant progress in recent decades alongside traditional in situ measurements. Globally accessible satellite-based datasets include those from the ARC-Lake (ATSR Reprocessing for Climate: LSWT and Ice Cover) project [47] and the Global Lake Temperature Collaboration (GLTC) [48]. Over the TP, available satellite-derived lake temperature products are limited, though efforts using AVHRR [49] and MODIS [50] have been published. Alternatively, lake temperatures can be derived or reconstructed through well-calibrated models [51,52]. A lake surface water temperature dataset for the TP was developed by combining the strengths of satellite observations and modeling [53].

While previous studies have primarily focused on monitoring changes in lake surface temperatures via remote sensing, the influence of lakes on adjacent land under climate change remains inadequately understood. Therefore, this study aims to characterize the spatiotemporal patterns and seasonal variations in LLTD across the TP, quantify the effective influence distance of lakes on surrounding land temperatures, and examine the relationships between LLTD and lake characteristics. These insights are crucial for improving our understanding of the climate-regulatory function of lakes on the Tibetan Plateau.

2. Materials and Data

2.1. Study Area

The Tibetan Plateau is located between 26°00′–39°47′N and 73°19′–104°47′E, with an average elevation exceeding 4500 m and covering an area of approximately 2.5×10⁶ km². It is known as the “Asian water tower”, contributing to most major rivers in Asia [54]. The total area of lakes on the TP is estimated to be around 50,000 km², with most lakes located at altitudes between 4000 and 5000 m [55]. This study focuses on 120 lakes across the plateau, encompassing most major lakes with areas greater than 40 km² (Figure 1). These lakes were selected based on the Records of Lakes in China [56], and detailed lake information is provided in Supplementary Table S1.

2.2. Data

The land surface temperature (LST) around the lakes was obtained from the MOD11 version 6 product, accessible via the Earth Observing System Data and Information System (EOSDIS; https://earthdata.nasa.gov, accessed on 26 September 2025). MOD11 [57] provides LST and emissivity data retrieved at a 1 km resolution using the generalized split-window algorithm, and in 6 km grids using the day/night algorithm, mainly based on bands 31 (10.78–11.28 μm) and 32 (11.77–12.27 μm) from MODIS onboard the Terra and Aqua satellites, where Terra overpasses the equator at around 10:30 a.m. (10:30 p.m.) local time and Aqua at around 1:30 a.m. (1:30 p.m.). MOD11 includes products with temporal resolutions of daily, 8-day, and monthly, respectively. This study used MOD11A1 (Terra product of Land Surface Temperature/Emissivity Daily L3 Global 1 km) to extract the surface temperature around lakes. MOD11A1 is a tile of daily Level 3 product at a 1 km spatial resolution corresponding to the earth locations on the sinusoidal projection. Validation studies report an accuracy generally within 1 K [58,59], though errors can reach to 2 K in bare-soil and desert environments [60,61]. Validations against ground measurements in the TP have reported a mean bias ranging from −0.6 K to 2.36 K and a Root-Mean-Square Error (RMSE) typically within 2–5 K, with higher accuracy observed at night [61,62,63]. These errors are considered acceptable for regional-scale thermal analysis. The daily LST in this study is the mean of daytime and nighttime observations.

Lake surface water temperature (LSWT) was obtained from the Lake Daily Water Surface Temperature Dataset for the Tibetan Plateau (LSWT_TP). This long-term daily dataset was generated based on an improved air2water model, capable of simulating temperatures during frozen periods, overcoming a key limitation of the original model. The model was calibrated and validated against remote sensing data (MOD11A1), further enhancing the reliability of the simulations by enabling continuous, year-round simulations. Validation against in situ sequential observations from four lakes (R² = 0.97, 0.92, 0.90, and 0.97, respectively) and sporadic observations from 41 lakes (R² = 0.94, RMSE ≈ 2 °C) indicate acceptable accuracy. The dataset provides daily scale LSWT records for 122 lakes from 1978 to 2022, including ice-covered periods. For further details, refer to Guo et al. (2022) [53].

To analyze the relationship between lake thermal effects and lake characteristics, data on lake morphology, salinity, and lake ice phenology were collected. Lake area, water level, depth, and volume were primarily obtained from Records of Lakes in China [56] and the HydroLAKES dataset [64]. Lake ice phenology data came from the TPLIP dataset, which was developed by combining remote sensing and lake modeling approaches [65,66]. Validation against MOD10A1-based results showed high agreement, with R² > 0.96 and RMSE < 5 days for all four lake ice phenological indices (freeze-up start, freeze-up end, break-up start, break-up end) [67]. Lake salinity data were obtained from the Electrical Conductivity (Salinity) Data of Lakes over 10 km² on the Tibetan Plateau from 1982 to 2020 [20,68], which was developed using remote sensing inversion based on 87 in situ measurements (R² = 0.51).

3. Methods

3.1. Temperature Difference Between Lakes and Surrounding Lands

In this study, the average land surface temperature was computed within predefined buffer zones around each lake. The lake–land temperature difference (LLTD) was defined as the difference between the mean lake surface temperature and mean land surface temperature within buffers. Multiple buffer distances—3 km, 5 km, 10 km, 15 km, and 20 km—were analyzed. To minimize potential biases from urban areas, all pixels classified as urban within the buffer zones were excluded from the land temperature calculation, based on the 300 m resolution Land Cover product from the European Space Agency (ESA) Climate Change Initiative (CCI). We first investigated the spatial distribution of the multi-year average LLTD and its seasonal variations across the TP. Finally, long-term trends in the annual mean LLTD from 2000 to 2022 were examined.

3.2. Lake Influence Distance

Previous studies have primarily focused on the cooling effect of rivers, reservoirs, and lakes during summer, with results often well described by a sigmoid model [69,70]. In contrast, lake warming effects remain less studied. This study extends the sigmoid model to incorporate both cooling and warming effects.

The sigmoid function, also known as the growth curve, is suitable for capturing discontinuity at the temperature origin. To assess the lake thermal effects, continuous concentric buffers were delineated around each lake at 1 km intervals, extending up to 20 km from the shoreline. A directional averaging method was applied to reduce biases due to anisotropic thermal influences. Temperature data from the lake center to the shoreline were excluded from the curve fitting process. For each lake, the average temperature within each buffer zone was plotted against the distance from the shore, producing a fitted curve (Figure 2).

For the cooling effect, we adopted a sigmoid model from previous work [70], as shown in Figure 2a:

$$T\left(D\right)=T_o+{\displaystyle \frac{a}{1+e^{-b(D-D0)}}}$$

(1)

where $T\left(D\right)$ represents the land surface temperature at distance $D$, $T_o$ is the lake surface temperature, $D$ is the distance from the lakeshore (positive values only), $a$ is the lake cooling intensity (LCI), $b$ is a distance-adjusting parameter, and $D0$ is the distance at the shore, which set to 0 in this study. The value of $T\left(D\right)$ approaches an upper asymptote beyond the maximum influence distance.

For the warming effect, a symmetrically modified sigmoid model was used, as shown in Figure 2b:

$$T\left(D\right)=T_o-{\displaystyle \frac{a}{1+e^{-b(D-D0)}}}$$

(2)

where $T\left(D\right)$, $T_o$, $D$, $b$, and $D0$ have the same definitions as in Equation (1), while $a$ here represents the lake warming intensity (LWI). $T\left(D\right)$ stabilizes at a lower asymptote beyond the maximum influence distance.

The temperature function increases/decreases by 95% from the water surface to the maximum/minimum temperature when $D-D0=3b$. Accordingly, the lake cooling/warming distance (LCD/LWD) is defined as $D_{effect}={\displaystyle \frac{3}{b}}$, while the lake cooling/warming intensity (LCI/LWI) is defined as $I=a$. LCD and LCI are used as the metrics to quantify the LCE [22,71], and similarly, LWD and LWI are used as the metrics to evaluate the LWE. In this study, the months exhibiting the most pronounced cooling and warming effects are first identified for each lake, and then the effect distance of the lake’s influence on land surface temperature during these months is computed.

4. Results

4.1. Spatial Distribution of LLTD

Under idealized conditions, seasonal temperature contrasts between lakes and adjacent land surfaces could offset each other under a complete energy balance, resulting in an annual LLTD close to zero. In practice, however, the observed LLTD values typically deviate from 0 °C due to complexities in energy balance and related climatic processes such as evaporative cooling and precipitation. Figure 3 illustrates the spatial distribution of the LLTD within 10 km buffer zones across the TP, both on an annual basis and by seasons. The results show that the LLTD values range from −2.8 °C (Wanquan Lake at Gerze County, Ngari Prefecture) to 3.4 °C (Gyesar Co at Coqên County, Ngari Prefecture). Spatially, lakes in the arid region of the northern TP generally exhibit a negative annual LLTD, indicating that mean lake surface temperatures are lower than those of the surrounding land, while semi-arid regions demonstrate greater spatial heterogeneity. Overall, 54% of lakes within the 10 km buffer zone show an annual mean LLTD below 0 °C, suggesting that most lake surfaces remain cooler than their surroundings throughout the year. This pattern likely results from the near-zero LLTD in winter (due to lake ice cover) and strongly negative summer LLTD, which together outweigh the positive LLTD in autumn. These findings highlight the role of lakes in modulating climate warming on the TP. LLTD was also computed for multiple buffer zones (3 km, 5 km, 10 km, 15 km, and 20 km), with the results (Appendix A Figure A1 and Figure A2) showing consistent spatiotemporal distribution patterns. Therefore, subsequent analyses of seasonal variation and interannual trends (Section 4.2 and Section 4.3) are based on the 10 km buffer, as all studied lakes exceed 40 km² in area.

In spring (March–May), the LLTD values are predominantly negative, ranging from –6.0 °C at Taro Co to 0.7 °C at Gyesar Co. The summer (June–August) values vary between –6.5 °C at Toson Lake and 2.4 °C at Gyesar Co. In autumn (September–November), the LLTD becomes mainly positive, ranging from −2.3 °C at North Hulsan Lake to 7.1 °C at Taiyang Lake. Winter (December–February) values span from −2.6 °C at North Hulsan Lake to 7.9 °C at Xuru Co. These results indicate clear seasonal patterns in thermal influences. Specifically, TP lakes generally cool adjacent land in spring and summer but warm it in autumn. During winter, lakes in the south maintain a warming effect, while those in the north show no statistically significant thermal influence.

4.2. Seasonal Variations in LLTD

Figure 4 shows the seasonal variation in the mean LLTD across all lakes on the TP, represented by the black solid line. The LLTD begins to gradually decrease in spring as intensified solar radiation raises land surface temperatures more rapidly than lake temperatures, which lag due to the higher heat capacity of water. This decreasing trend continues into summer, reaching an annual minimum in June. During autumn, lakes retain heat more effectively than the rapidly cooling land, resulting in a continuous rise in LLTD that peaks in late October. Throughout winter, the LLTD shows a gradual decrease. Notably, a near-zero phase occurs for about one month starting in mid-August. This transitional equilibrium results from several factors: monsoon season induces cloud cover, reduces solar elevation angles, and extends nocturnal cooling, collectively moderating land surface warming, while the thermal inertia of lakes delays their temperature response relative to land. These counteracting thermal response mechanisms drive the LLTD toward temporary balance in August.

Seasonal LLTD variations exhibit distinct patterns across different basins. The Inner Basin, which contains the largest proportion of lakes (

Table 1. Number of lakes in each basin or climate zone.

Basin	Number	Climate Region	Number
Inner	96	HIIC	7
Brahmaputra	3	HIID	9
Indus	5	HIB	2
Qaidam	6	HIC	77
Yangtze	3	HID	25
Yellow	7
Total	120	Total	120

), largely governs the overall LLTD behavior of the TP. For the ten lakes within the Yangtze and Yellow River Basins (Figure 4a), the LLTD remains negative from spring through June, increases thereafter, and declines again in September. This pattern likely arises from the interplay among radiative forcing, thermal storage release in lakes, and evaporation effects—where terrestrial evaporation suppresses land temperature while lake evaporation modulates water surface cooling. In the Qaidam Basin, lakes with prolonged ice cover and delayed melt exhibit an LLTD minimum in early October, followed by a rapid increase. The Indus Basin shows the highest LLTD peak (7.65 °C) in late November, surpassing other basins; if sustained, this increase could influence local humidity and lake thermal stratification. Although seasonal LLTD patterns are generally consistent across climatic zones (Figure 4b), arid regions [72]—including the high-altitude sub-frigid arid zone (HID) and the high-altitude temperate arid zone (HIID)—reach their LLTD minimum in late July, lagging behind other zones.

4.3. Interannual Variations in LLTD

Figure 5 shows the interannual variations in LLTD within 10 km buffer zones across different seasons (Figure 5a–d) and on an annual basis (Figure 5e) from 2000 to 2022, as determined by the Mann–Kendall trend test [73,74]. Annually, 79.2% of the lakes exhibited a decreasing trend in LLTD, which can be attributed to multiple factors, such as glacial meltwater inflow, evaporative cooling, lake expansion, snow cover loss, urbanization, and more rapid climatic response of land compared to lakes. Seasonally, the average summer LLTD decreased at a rate of –0.56 °C per decade, representing the largest contributor to the annual decline (Figure 5b). This pattern is mainly driven by two factors: (i) the predominance of glacier-fed lakes, which constitute 63% of the studied lakes and exhibit suppressed summer warming compared to adjacent land, and (ii) enhanced evaporative cooling that counteracts lake temperature increases [75]. In contrast, during winter, 87.5% of the lakes showed increasing LLTD trends (Figure 5d), with an average rate of 0.3 °C per decade. This trend occurs because TP lakes warm more significantly in winter than in other seasons [53]. Furthermore, reduced lake ice cover enhances radiative heat absorption, resulting in more pronounced lake surface warming relative to land during winter. Spatially, lakes located in the southern and northwestern TP show a declining trend in annual LLTD, with statistically significant decreases (p < 0.05) during summer. In autumn, southern lakes show a decreasing trend in LLTD, whereas northern regions experience a more pronounced warming-driven increase in LLTD in winter. The decrease in autumn LLTD over southern TP lakes is consistent with regional “warm-wet” trends, where lake expansion and increased evaporation may further slow down the lake warming effect.

4.4. Effect Distance and Effect Intensity of Lakes

As shown in Figure 3 and Figure A1, LLTD varies with selected buffer distance, indicating that lakes influence adjacent land within a limited spatial range, although the exact effective influence distance remained unclear. To address this, we first identify for each lake the months with the most pronounced cooling (typically occurring in spring/summer) and warming (typically occurring in autumn/winter) effects, as summarized in Supplementary Table S1. Using the method described in Section 3.2, we then quantified the influence distance (Figure 6a,b) and influence intensity (Figure 6c,d) during these key months. The regression models for both cooling and warming effects demonstrated robust performance. Specifically, 105 lakes were retained for the cooling effect and another 105 lakes for the warming effect, based on goodness-of-fit (R²) values exceeding 0.4 in their respective models.

The results show that the average lake cooling distance across the TP is approximately 1.3 km, ranging from a minimum of 0.3 km (Garing Co at Coqên County, Ngari Prefecture) to a maximum of 4.7 km (Bairab Co at Gerze County, Ngari Prefecture). In contrast, the mean lake warming distance is 5.5 km, varying from 0.3 km at Co Ngoin to 17.6 km at Aqqik Kol. Thus, the “warm island” effect in autumn influences a broader spatial extent than the “cold island” effect in summer, despite similar influence intensities. This pattern is attributed primarily to lake thermal inertia and seasonal differences in atmospheric boundary layer stability. In autumn, lakes release heat stored from summer due to high heat capacity, acting as persistent heat sources that transfer energy over long distances. The more stable atmospheric boundary layer in autumn suppresses vertical turbulence and promotes horizontal advection, which enables heat to be transported farther. In contrast, in summer, rapid diurnal temperature fluctuations over land and strong vertical mixing limit its horizontal extent of cooling. Quantitatively, summer cooling intensities across TP lakes range from 1.3 °C to 9.8 °C, while autumn warming intensities vary between 0.4 °C and 10.7 °C. Spatially, lakes in the southwest and northwest TP show stronger cooling and warming intensities than other regions, exceeding 5 °C. Notably, several lakes within the Yellow River basin—including Donggeicuona Lake, Ngoring Lake, Kuhai Lake, and Longyang Reservoir—demonstrate warming intensities greater than 7.5 °C. The influence distances and intensities of lake thermal modulation are crucial for assessing lacustrine impacts on regional microclimates, which exhibit morphometric dependencies, as further discussed in Section 5.1.

5. Discussion

5.1. Relationships Between Lake Thermal Effects and Lake Characteristics

To mitigate multicollinearity among lake characteristics, this study calculated partial correlation coefficients (Rp) between lake characteristics and LLTD, as shown in Figure 7. Lakes with greater water storage capacity exhibited a significant (p < 0.05) negative correlation with spring LLTD (Rp = –0.23), which can be attributed to their higher thermal inertia delaying temperature increases relative to adjacent land. In contrast, in winter the LLTD correlated positively with lake water volume, with lake depth identified as the dominant factor (R = 0.33, p < 0.05). Moreover, higher lake elevation was associated with greater annual LLTD (Rp = 0.38, p < 0.05), a trend particularly evident in winter (Rp = 0.43). This likely results from thinner atmosphere at high elevations, enabling lakes to absorb more solar radiation than surrounding land, and thus enhancing the “warm island” effect.

Lake ice phenology and LLTD are related. Lakes with later freeze-up dates prolonged open-water phases, allowing sustained sensible heat flux release from the lake surface. This thermodynamic process helps maintain lake temperatures above those of surrounding land during winter, resulting in a significant negative correlation (Rp = −0.41) between the freeze-up start date (FUS) and winter LLTD. The persistence of thermally active surfaces further enhances the role of lakes as local heat sources.

Many lakes on the TP are saline, and their salinity is closely linked to thermal behavior. High salinity reduces the specific heat capacity, accelerating the lake temperature response to radiative forcing, while simultaneously promoting latent heat dissipation via surface tension, which cools the lake surface. A significant negative correlation was identified between salinity and autumn LLTD (Rp = −0.38), as well as LLTD within the warming effect distance (Rp = −0.40). This relationship can be attributed to more stable stratification in saline lakes, which suppresses vertical heat transfer from deeper layers, along with the effect of evaporative cooling.

5.2. Limitations and Outlook

This study investigates the thermal effects of lakes on surrounding land temperatures, with a focus on the spatiotemporal and seasonal variations in LLTD, and the distance and intensity of thermal influence. While the average influence distance around lakes was quantified, it should be noted that this distance is modulated by localized topography and prevailing wind patterns, resulting in directional heterogeneity in thermal effects. Future research could incorporate additional topographic and environmental variables to better capture the anisotropic variations in thermal propagation. Furthermore, although this study focuses on long-term and large-scale trends in LLTD, pronounced diurnal temperature fluctuations—such as those driving lake–land breezes—introduce complexities that are not yet fully understood over the TP. These diurnal dynamics may contribute uncertainties in quantifying the LLTD and thermal influence distances, warranting further investigation into multiscale atmosphere–lake–land interactions.

LLTD arises from the lag in temperature between lakes and adjacent land, driven by the thermal inertia of water. This delayed thermal response significantly influences climatic and ecological process. In spring, delayed lake warming maintains cooler adjacent zones, thereby suppressing convection and precipitation. During summer, the thermal lag generates cool, humid microclimates along shorelines that serve as critical thermal refugia for species adapted to plateau conditions. In autumn, the slower cooling of lakes extends frost-free periods, prolongs growing seasons, and influences animal behaviors such as migration and reproduction—for example, affecting the survival of Anser indicus and Artemisia wellbyi [76]. Moreover, the thermal buffering capacity of lakes mitigates extreme climatic events through energy redistribution, an effect particularly pronounced in larger lakes [21]. Interannual variability (Section 4.3) reveals a strengthening of the summer “cold island” effect, enhancing the role of lakes in reducing heat stress mortality for cold-adapted species. Concurrently, intensified winter “warm island” effects increase sensible heat flux from lake surfaces, attracting overwintering waterbirds, accelerating permafrost thermal degradation, and altering lake-mediated carbon cycling. Therefore, quantitative assessments of these climate–ecosystem feedback processes are essential for understanding future hydrological and ecological dynamics on the plateau.

6. Conclusions

Lakes play a critical role in modulating local land surface temperature and climate variability due to their distinct thermal properties compared to terrestrial surfaces. Advances in satellite-based observations and lake energy balance modeling have greatly improved the availability of surface temperature products spanning both lakes and surrounding land at large spatial scales. As lakes on the Tibetan Plateau (TP) are particularly sensitive to climate change, this study investigated the thermal effects of lakes on adjacent land through the integration of remote sensing data and modeling data across the TP. The main findings are summarized as follows:

(1)

Lakes on the Tibetan Plateau significantly modulate local surface temperatures, with the lake–land temperature difference (LLTD) within a 10 km buffer ranging from –2.8 °C to 3.4 °C. Spatially, lakes in the northern arid regions generally exhibit a negative LLTD (cooling effect), while seasonal variations reveal a negative LLTD in spring and summer due to higher lake heat capacity, transitioning to positive values (warming effect) in autumn as lakes release stored heat. During winter, southern TP lakes maintain a pronounced warming influence.

(2)

Considerable spatial and temporal heterogeneity in LLTD is observed across basins and climate zones. Lakes in the Inner Basin dominate the overall LLTD pattern of the TP, while arid-region lakes show the lowest LLTD values. Temporally, 79.2% of lakes exhibited a declining trend in LLTD from 2000 to 2022, with summer contributing most significantly to this decrease (–0.56 °C/decade), whereas winter LLTD increased (0.3 °C/decade).

(3)

The spatial extent of lake thermal effects, quantified via a sigmoid model, indicates a broader “warm island” effect in autumn (5.5 km) compared to the summer “cold island” effect (1.3 km). Geographically, southwestern lakes exhibit stronger warming intensities, while northwestern lakes show more pronounced cooling intensities.

(4)

Key lake characteristics—including depth, freeze-up phenology, and salinity—significantly influence lake thermal effects. Lake depth shows a strong positive correlation with winter LLTD (R = 0.33), and an earlier lake freeze-up start date is negatively correlated with winter LLTD (R = –0.41), highlighting the role of persistent open water in enhancing local heating. Higher salinity reduces autumn LLTD, likely due to reduced heat capacity and increased latent heat dissipation. Future research should focus on the climatic and ecological implications of these thermal effects under ongoing climate change.