Air–Ice–Water Temperature and Radiation Transfer via Different Surface Coverings in Ice-Covered Qinghai Lake of the Tibetan Plateau

Abstract

There are numerous lakes in the Tibetan Plateau (TP) that significantly impact regional climate and aquatic ecosystems, which often freeze seasonally owing to the high altitude. However, the special warming mechanisms of lake water under ice during the frozen period are poorly understood, particularly in terms of solar radiation penetration through lake ice. The limited understanding of these processes has posed challenges to advancing lake models and improving the understanding of air–lake energy exchange during the ice-covered period. To address this, a field experiment was conducted at Qinghai Lake, the largest lake in China, in February 2022 to systematically examine thermal conditions and radiation transfer across air–ice–water interfaces. High-resolution remote sensing technologies (ultrasonic instrument and acoustic Doppler devices) were used to observe the lake surface changes, and MODIS imagery was also used to validate differences in lake surface conditions. Results showed that the water temperature under the ice warmed steadily before the ice melted. The observation period was divided into three stages based on surface condition: snow stage, sand stage, and bare ice stage. In the snow and sand stages, the lake water temperature was lower due to reduced solar radiation penetration caused by high surface reflectance (61% for 2 cm of snow) and strong absorption by 8 cm of sand (absorption-to-transmission ratio of 0.96). In contrast, during the bare ice stage, a low reflectance rate (17%) and medium absorption-to-transmission ratio (0.86) allowed 11% of solar radiation to penetrate the ice, reaching 11.70 W·m⁻², which increased the water temperature across the under-ice layer, with an extinction coefficient for lake water of 0.39 (±0.03) m⁻¹. Surface coverings also significantly influenced ice temperature. During the bare ice stage, the ice exhibited the lowest average temperature and the greatest diurnal variations. This was attributed to the highest daytime radiation absorption, as indicated by a light extinction coefficient of 5.36 (±0.17) m⁻¹, combined with the absence of insulation properties at night. This study enhances understanding of the characteristics of water/ice temperature and air–ice–water solar radiation transfer through effects of different ice coverings (snow, sand, and ice) in Qinghai Lake and provides key optical radiation parameters and in situ observations for the refinement of TP lake models, especially in the ice-covered period.

1. Introduction

The Tibetan Plateau (TP), often referred to as the “Asian Water Tower”, is home to a vast expanse of high-altitude lakes, collectively covering over 50,000 km², constituting more than half of China’s total lacustrine area [1,2,3]. These lakes, perched at elevated altitudes, are prone to seasonal freezing, with ice periods lasting from several months to as long as half a year and maximum ice thicknesses ranging from 0.58 to 0.83 m [4,5,6,7,8,9,10,11]. The formation of lake ice is not merely a natural phenomenon; it significantly influences the local climate by altering the radiation transfer from the air to the water, which in turn affects the thermal dynamics of the lakes [12,13].

The ice cover acts as a barrier to heat exchange, curtailing heat loss and evaporation, and acts as a pivotal component of the climate system on the Tibetan Plateau [14,15]. However, our current understanding of the physical processes and parameters governing lake ice formation is limited, leading to considerable uncertainty in the simulation of the plateau’s lake ice period by current climate models. The complexity of these processes, which include heat transfer, the evaporation of moisture, and changes in ice thickness, is compounded by the scarcity of direct observational data due to the challenging observational conditions in the region [15,16]. Uncertainties in parameters such as the optical properties and thermal conductivity of lake ice further increase the complexity of model simulations.

Existing models face challenges in estimating key parameters such as ice thickness, sub-ice water temperatures, and the ice surface albedo, which directly affect the simulation accuracy of the lake’s energy balance and the hydrological cycle process [4,15,16]. For instance, the increase in water temperature during the ice period and the rapid rise in water temperature near the ice layer towards the end of the ice melting are phenomena that current models struggle to simulate accurately [15]. Weather phenomena, including strong winds and precipitation, can significantly alter the internal radiation and thermal conditions within frozen lakes by influencing the ice cover’s composition and surface properties, such as albedo and light transmittance [17,18]. The reflection of solar radiation by snowfall, in particular, can lead to a decrease in lake temperature [19].

The thermal dynamics of fully ice-covered lakes are predominantly influenced by solar radiation, which can induce melting at various levels of the ice and create heat convection cells beneath the ice [4,20]. Observations indicate that, during the ice-covered period, lake water beneath the ice undergoes a gradual warming process attributed to the penetration of solar radiation through the ice layer [21,22]. Despite the presence of ice cover, a measurable fraction of solar radiation is transmitted, directly contributing to the incremental increase in water temperature throughout the freezing season [22,23,24]. In contrast, widely applied lake models for Tibetan Plateau (TP) lakes, such as the Freshwater Lake Model (FLake) and Community Land Model (CLM) coupled with lake schemes in the Weather Research and Forecasting Model (WRF), generally incorporate assumptions of negligible or non-existent solar radiation penetration through lake ice after freezing [15,21]. As a result, these models are unable to accurately represent the observed thermal dynamics, particularly the sustained warming of sub-ice water during the ice-covered period. This limitation underscores the need for incorporating more realistic representations of radiation transfer processes into lake models to enhance their capability in simulating energy balance and thermal dynamics under ice-covered conditions.

Field observations on the Tibetan Plateau are challenging due to the harsh climatic conditions, which has led to a focus on lake studies during the ice-free period. The scarcity of observational data during the freezing season, especially in the remote and sparsely populated Qinghai–Tibet Plateau, limits our understanding of the complex interactions between lake water, ice, and the atmosphere. The optical characteristics of lakes, including absorptivity and attenuation coefficients during the freezing period, remain elusive, hindering the advancement of lake modeling in the region [3,7,25,26]. Current models, such as the WRF-Flake [27], Flake [28], and LAKE2.0 [29], tend to overestimate the albedo of lake ice on the Tibetan Plateau, resulting in unsatisfactory simulations of energy balance and ice phenology during the ice-covered period [30]. To further refine global circulation models, regional climate models, and numerical weather forecasting models, it is essential to improve the representation of lake features within these models [31,32]. This highlights the urgent need for a better understanding of lake ice physics and the physical principles governing the seasonal evolution of ice.

This study sought to contribute to the understanding of atmosphere–ice–water interactions in Qinghai Lake through systematic multi-layer observations. Utilizing field observations and meteorological data, the impact of different weather processes on the lake surface and the effects of different ice coverings on radiation and temperature within the system were analyzed. This study represents one of the first attempts to conduct stratified observations of lake ice energy, analyzing how different ice cover materials influence temperatures across various ice layers. The findings contribute to a deeper understanding of how ice cover affects the lake’s energy balance and thermodynamic properties, offering valuable insights for refining TP lake model parameters and improving knowledge of the Qinghai Lake basin and the regional aquatic environment.

2. Materials and Methods

2.1. Study Area

The Qinghai Lake (36.53~37.25° N, 99.60~100.78° E), situated at the northeastern edge of the TP with an elevation of 3195 m, is the largest lake in China, which stretches approximately 106 km from east to west and 63 km from north to south (Figure 1a). The lake surface area is 4486.1 km², and the average depth is 21 m [33]. The average winter air temperature in the Qinghai Lake Basin ranges from −13.8 to −10.8 °C, with an extreme minimum temperature of approximately −33.4 °C. The lake water, with a salinity of 12.50~12.96 g·L⁻¹, is a weak alkaline solution with a pH ranging from 8.95 to 9.03. The lake’s drainage basin, with topography at higher elevations and an area of 2.97 × 10⁴ km², forms a closed inland basin [34].

Qinghai Lake undergoes a significant seasonal ice cover, with freezing typically commencing in December and culminating in substantial ice layers by January. These ice layers, averaging several tens of centimeters in thickness, gradually melt by March with the lake ice completely thawing by April [7]. Accompanying the ice cover, Qinghai Lake receives winter snowfall predominantly between November and February [35]. The annual precipitation in this period is variable yet generally ranges from low to moderate, influenced by the lake’s geographic setting and interannual climatic variability. Surrounding the lake, the dry and semi-arid zones are occasionally subject to dust and sand events. While infrequent, these dust storms, driven by strong winds carrying sand particles over the lake, are indicative of the region’s distinctive climatic characteristics, particularly during the winter months [36].

To comprehensively understand the various factors affecting the thermal balance of Qinghai Lake, we have conducted a study examining the optical properties of the lake water and ice, as well as the propagation of solar radiation through the ice layer.

2.2. In Situ Observation

To comprehensively understand the various factors affecting the thermal balance of Qinghai Lake, we conducted a field study examining the optical properties of the lake water and ice, as well as the propagation of solar radiation through the ice layer. An atmosphere–ice–water trinity observation program was conducted in Qinghai Lake in 6–24 February 2022. The observation site (Figure 1b) was located close to the shore in the Erlangjian Scenic Area of Qinghai Lake (36.59° N, 100.50° E), where the water depth is 18.5 meters. Observation data were collected with temporal resolutions of 1 minute for air temperature, wind, and radiation; 10 minutes for the ultrasonic distance meter system; and 30 minutes for underwater irradiance, as detailed in

Table 1. Introduction of observation instrument.

Observation Item	Sensor (Manufacturer)	Accuracy	Range	Height (Depth)
Temperature	PTWD (JST, Jinzhou, China)	0.2 °C	−40~80 °C	1.5 m
Wind speed	MaxiMet GMX 501 (Gill Instruments Ltd., Lymington, Hampshire, UK)	0.1 m·s−1	0.1~60 m·s−1	1.5 m
Global radiation	TBQ-2 (JST, Jinzhou, China)	<5%	300−3000 nm	1.5 m
Snow/sand depth (ice surface)	SR50A (Campbell Scientific, Logan, UT, USA)	0.01 cm	0.5~10 m	−0.6 m
Ice thickness (ice bottom)	Tritech PA500/6 (Tritech International Ltd., Westhill, Aberdeenshire, UK)	0.1 cm	0.1~10 m	−0.4 m
Ice temperature	PTWD (JST, Jinzhou, Liaoning, China)	<5%	−40~150 °C	−0.05, −0.10, −0.15, −0.20 m
Water temperature	PTWD (JST, Jinzhou, China)	<5%	−40~150 °C	−0.4, −0.5, −2.1, −6.7, −8.7, −12.7 m
Underwater irradiance	HOBO Pendant Temperature/Light 64K Data Logger-UA-002-64 (Onset Computer Corporation, Bourne, MA, USA)		175–1200 nm	−0.7, −2.1 m

. The underwater irradiance was measured in lux but converted to W·m⁻² here. The local noon at the site is within ±15 min of 13:30 h CST (China Standard Time; CST = UTC + 8 h). The solar radiation data with a solar elevation angle less than 15° were eliminated due to the weak solar radiation at sunrise and sunset. A practical problem in our setup was that snow accumulated around the instrument platform because of the winds. In the analysis, the lake water temperature data observed from February to April 2023 were also utilized to analyze the long-term characteristics of the water temperature during the ice cover period of Qinghai Lake.

The precipitation data were obtained from the National Meteorological Science Data Center (http://data.cma.cn/) (accessed on 10 May 2024). The dataset used was the China Surface Climate Data Daily (V3.0), which includes precipitation data from the Qinghai Lake 151 station with a temporal resolution of 1 hour. This station, the nearest ground meteorological station to the lake, is located in the southern side of the lake (36.58° N, 100.48° E) at an elevation of 3200.8 m. According to field observation, there was a thin layer of sand on ice, and the video surveillance shows a clear sand blowing.

2.3. Remote Sensing Instrumentation for Lake Surface and Ice Bottom Monitoring

In this study, two remote sensing instruments were utilized to monitor changes at the upper lake interface (covering or ice) and the ice bottom during the freezing period at Qinghai Lake. The SR50A, manufactured by Campbell Scientific (Logan, UT, USA), is an ultrasonic sensor that measures changes at the upper lake surface, which may include transitions between snow, sand, and ice. It operates by emitting high-frequency sound pulses and measuring the time it takes for the echo to return, achieving an accuracy of ±0.01 cm over a range of 0.5 to 10 m, with a deployment height of −0.6 m. This instrument’s quick response time and high resolution make it particularly effective for continuous monitoring in challenging environments where traditional measurement methods may be impractical or hazardous [37,38].

In parallel, the Tritech PA500/6 (Tritech, UK) measures changes at the underwater ice bottom using acoustic Doppler technology. This instrument provides distance measurements with an accuracy of ± 0.1 cm and operates over a range of 10 to 1000 cm, with a deployment depth of −0.4 m. By analyzing the frequency shifts in the reflected sound waves from particles within the water, the PA500/6 offers detailed insights into the growth and decay of the ice bottom [39,40].

Together, these instruments measure the dynamic changes at the lake surface and the ice bottom to assess the total thickness of the lake cover and ice, contributing to a better understanding of the effects of different surface covers on ice melt.

2.4. Terra/MODIS Remote Sensing Imagery

The Moderate Resolution Imaging Spectroradiometer (MODIS), developed by NASA, was used alongside automatic weather station monitoring images to comprehensively assess the weather conditions and lake surface processes of Qinghai Lake in 6–24 February 2022. The MODIS instrument on the Terra satellite, using corrected reflectance and the Band 3-6-7 combination, provided false-color images that are particularly effective for snow and ice mapping due to the distinct reflective and absorptive properties of these features in different parts of the electromagnetic spectrum. The images, available from NASA’s Earth Data site (https://wvs.earthdata.nasa.gov/) (accessed on 10 May 2024), have a spatial resolution of 250 meters and a temporal resolution of 1 day.

2.5. Methodology

2.5.1. Albedo α

The surface albedo is the ratio of the upward solar irradiance E_u (unit: W·m⁻²) to the downward solar irradiance E_d just above the surface:

$$\alpha ={\displaystyle \frac{E_u}{E_d}}$$

(1)

It is an important parameter for the surface energy balance.

2.5.2. Lake Water Body Attenuation Coefficient K_dw and Lake Ice and Covering Attenuation Coefficient K_di

Infrared radiation is absorbed in a thin near-surface layer, and only photosynthetically active radiation (PAR) wavelengths (400–700 nm) are present in radiation that travels through ice [41]. Two radiation sensors were used to measure the downward radiation in the under-ice water body: sensor 1 at a depth of 0.7 m and sensor 2 at a depth of 2.1 m. By utilizing these PAR measurements, the attenuation coefficient K_dw (unit: m⁻¹) can be calculated using Equation (2). For a vertically optically homogeneous water body, the attenuation of radiation follows the exponential decay law [42]:

$$K_{dw}=-{\displaystyle \frac{1}{∆z}}\mathit{ln}{\displaystyle \frac{E_d\left(z_{0.7}\right)}{E_d\left(z_{2.1}\right)}}$$

(2)

Here, $∆z=z_{2.1}-z_{0.7}$, $E_d\left(z_{0.7}\right)$ and $E_d\left(z_{2.1}\right)$ (unit: W m⁻²) represent the downward radiation at depths $z_{0.7}$ and $z_{2.1}$, respectively.

The coefficient K_di is determined by utilizing the radiation at the ice bottom $E_d\left(z_{ice-water}\right)$ and the incident PAR on the lake surface. Based on previous research conducted at Xiaopo Lake at the eastern shore of Qinghai Lake, the average value of the PAR coefficient (ηPAR) in February in the Qinghai Lake basin is 0.42 [43]. Combining this with the thickness of the lake ice, K_di can be calculated.

2.5.3. Ice–Water Interface PAR z_i(PAR), Euphotic Zone Depth Z_eu, and Lake Ice Transmittance

By utilizing the PAR from Sensor 1, the distance between the ice bottom and sensor 1, and K_dw, one can estimate the PAR at the ice bottom during the observation period using Equation (3), where $z_{air-ice}$ represents the depth at the air–ice interface, $z_{air-ice}$ represents the depth at the ice–water interface, $E_d\left(z_{air-ice}\right)$ represents the irradiance at the air–ice interface, and $E_d\left(z_{ice-water}\right)$ represents the irradiance at the ice–water interface.

$$E_d\left(z_{ice-water}\right)=E_d\left(z_{air-ice}\right)e^{K_{dw}\left(z_{air-ice}-z_{ice-water}\right)}$$

(3)

The euphotic zone depth is defined as the depth at which the net primary production becomes zero, coincident with the depth of the layer of photosynthetic activity. The euphotic zone depth Z_eu is usually defined as the depth at which the irradiance is 1% of the PAR irradiance at the surface.

It can be calculated by incorporating $h_i$, the ice thickness, and $h_w$, the depth in water where irradiance reaches 1% of the surface PAR irradiance.

$$Z_{eu}=h_i+h_w$$

(4)

$$0.01=\eta PAR\ast \left(1-\alpha \right)e^{{-(h_{w}K}_{dw}+{h_{i}K}_{di})}$$

(5)

The transmittance refers to the ratio of downward radiation at the ice layer depth to the PAR on the lake surface. By using the radiation reaching the ice bottom, one can calculate the ice layer transmittance:

$$\tau ={\displaystyle \frac{E_d\left(z_{ice-water}\right)}{\eta PAR\ast E_d}}$$

(6)

3. Results

3.1. Background Field

Taking into account the prevailing meteorological conditions and the variability in surface coverage, the study period was delineated into three distinct stages represented by different shades in Figure 2: the snow stage, encompassing 6–11 February; the sand stage, occurring in 13–14 February; and the bare ice stage, observed in 19–24 February.

Table 2. Major weather phenomena and lake surface features.

Date	Weather Phenomena	Lake Surface Features
February 5–6, 10	Snow	Snow cover
February 12–14	Sand blowing	Sand cover
February 18	Strong wind	Bare ice

provides a summary of major weather phenomena and lake surface features during the study period. For reference, the MODIS remote sensing images and snapshots from the automatic weather station during the observation period are shown in Figure 3.

During the snow stage, the daily mean air temperatures dropped to about −15.23 °C on 7 February and then increased to −8.42 °C in the next stage. The meteorological observatory recorded the weather as snowfall on 5 February, with a total of 3.5 mm of snow falling during two periods: 02:00–03:00 and 23:00–06:00 in 5–6 February. A thin layer of snow forming on the ice surface was observed, with a thickness of about 2.03 cm on 5 February.

The average wind speed was 3.18 m·s⁻¹ during the observation period. There were four days experiencing speeds exceeding 6.00 m·s⁻¹ accompanying two significant wind events. The first wind event spanned in 12-14 February, with daily average wind speeds ranging from 6.07 to 6.69 m·s⁻¹. The intensity of these winds resulted in the deposition of fine sand particles onto the ice surface and started the sand stage. Subsequently, the second significant wind event was recorded on 18 February, characterized by a daily average wind speed of 6.09 m·s⁻¹ and instantaneous winds of up to 17.7 m·s⁻¹. This event led to the dispersion of sand particles, thereby exposing the underlying ice surface. The bare ice period began, which is a common characteristic of TP lakes due to less snow than the low-altitude lakes with high latitudes.

3.2. Lake Surface-Covering Transformation and Ice Thickness

As seen in Figure 4, the distance between the underwater ultrasonic device and the ice base did not change much during the entire observation period, decreasing by only 1.95 cm in 6–24 February. This suggests that the sinking of the lake ice bottom was slow, at a rate of about 0.11 cm·d⁻¹.

The lake boundary comprises the coverings of snow or sand on the top and ice on the bottom. During the snow stage, the ultrasonic device was positioned 60.37~61.15 cm from the surface. The average thickness of ice and snow layers remained constant at 34.30~35.85 cm. The snow cover was relatively thin, around 2 cm.

In the sand stage, with sand covering the surface, the distance of the ultrasonic device from the ice to the sand surface rapidly decreased from 60.59 cm to the minimum value of 50.32 cm. There was a significant increase in the thickness of the ice and sand layer from 35.85 cm to the maximum of 47.53 cm. Wind activity resulted in a measured sand and snow layer thickness of approximately 8 cm in the observation area, where accumulation was notable due to observation structure effects causing substantial deposition. The spatial distribution of sand in the Qinghai Lake is heterogeneous due to its considerable size, leading to thin sand layers in other regions.

In the bare ice stage, only with the bare ice, the daily average lake ice thickness decreased and stabilized at 36.36~36.89 cm. The daily fluctuation in ice thickness ranged from 2.20 cm to 2.70 cm, exhibiting the most significant diurnal variation observed during the study period. This could possibly be attributed to increased ice sublimation and deposition during the daytime and nighttime.

Based on the analysis above, the ice bottom had a minimal variation, and the ice thickness maintained about 36.6 cm during the whole observation period that happened to be the stable ice period. The thickness between the lake surface and ice bottom varied mostly because of the changed coverings.

3.3. Lake Water and Ice Temperature

3.3.1. Lake Water Temperature Under Ice

The lake water hovered around 0 °C with fluctuations not exceeding 0.54 °C (Figure 5a) during the whole observation period. The temperature of the lake water near the bottom of the ice (0.4 m or 0.5 m below the surface of lake ice) was lower than 0 °C, which was because Qinghai Lake is a saline lake (12.50 g·L⁻¹), which has about a −0.69 °C negative freezing point [44]. The deep lake water (12.7 m) temperature remained below 0 °C until 18 February, after which it exceeded 0 °C.

The water in all layers with snow and sand coverages was colder than that with only bare ice. The mean temperature at the water depth of 0.4 m (12.7 m) below the ice surface was −0.24, −0.29, and −0.10 °C (−0.18, −0.17, and 0.15 °C) during snow, sand, and ice stages, respectively (

Table 3. Maximum, minimum, and mean of lake water temperature at different depths below the ice surface in three stages.

Depth	Max (°C)			Min (°C)			Mean (°C)
	Snow	Sand	Ice	Snow	Sand	Ice	Snow	Sand	Ice
0.4 m	−0.11	−0.22	0.05	−0.33	−0.36	−0.22	−0.24	−0.29	−0.10
0.5 m	−0.11	−0.22	0.06	−0.32	−0.35	−0.20	−0.24	−0.29	−0.10
6.7 m	−0.12	−0.17	0.11	−0.31	−0.34	−0.23	−0.23	−0.27	−0.07
8.7 m	−0.12	−0.17	0.11	−0.31	−0.33	−0.23	−0.24	−0.27	−0.08
12.7 m	0.01	−0.14	0.36	−0.31	−0.28	−0.06	−0.18	−0.17	0.15

).

The water temperature was almost uniformly mixed throughout the water column in the snow and sand stages, with the vertical gradient within 0.1 °C between the shallow and deep layer. During the bare ice stage, the vertical temperature difference gradually increased to 0.37 °C.

The short duration limited the visibility of this subtle warming trend in Qinghai Lake in the February 2022 observation period (Figure 5a), but the extended observations in 2023 confirmed a gentle warming tendency throughout the ice period and a notable temperature rise of 3.87 °C before ice melting from late March to early April (Figure 5b). Qinghai Lake, characterized as a brackish body and the biggest lake in the TP and China, also exhibited similar typical characteristics with increasing seasonal under-ice water temperatures of the Tibetan Plateau’s lakes. The phenomenon had been observed in TP lakes including Ngoring Lake Bangong Co, Zhari Namco, and Dagze Co, etc. [15], but the related conditions of lake ice and air–ice–water radiation transfer have not been studied or supported by the in situ observations.

3.3.2. Ice Temperature

Significant disparities in ice temperature in the three distinct stages were observed (Figure 6). The ice located 5 cm beneath the ice surface had the smallest daily minimum temperature and biggest daily maximum temperature (−10.50 and −0.40 °C) during the bare ice stage compared to that covered with snow (−5.46 and −3.00 °C) and sand (−4.41 and −2.99 °C). Thus, its mean temperature (−5.29 °C) was lower in the bare ice stage than in the other stages (−4.02, and −3.68 °C), while the diurnal ice temperature fluctuations in the bare ice stage (8.52 °C) is 5~8 times higher than the daily temperature fluctuation observed in the snow (1.58 °C) and sand (1.04 °C) stages. The deeper ice layers (10 cm, 15 cm, and 20 cm below the ice surface) shared similar stage temperature characteristics, and they were generally warmer and exhibited smaller temperature ranges (

Table 4. Maximum, minimum, and mean of lake ice temperature at different depths below the ice surface in three stages.

Depth	Max (°C)			Min (°C)			Mean (°C)
	Snow	Sand	Ice	Snow	Sand	Ice	Snow	Sand	Ice
5 cm	−3.00	−2.99	−0.40	−5.46	−4.41	−10.50	−4.02	−3.68	−5.29
10 cm	−1.70	−2.60	−1.51	−4.03	−3.41	−7.90	−3.00	−2.98	−4.39
15 cm	−0.42	−2.30	−1.90	−3.00	−2.80	−6.10	−2.14	−2.52	−3.69
20 cm	−0.32	−1.77	−1.90	−1.90	−2.00	−4.50	−1.25	−1.88	−2.93

).

During the snow, sand, and bare ice stages, the minimum values of the daily average temperatures recorded beneath the ice surface at 5 cm (−4.76, −4.16 and −9.71 °C) and 20 cm (−1.34, −1.95, and −4.05 °C) were all documented between 07:30 and 08:30. The maximum values of the daily average temperatures at these depths (5 cm: −3.26, −3.11, and −1.19 °C; 20 cm: −1.09, −1.79, and −2.16 °C) were observed between 17:00 and 18:00 during the three stages. The vertical temperature difference between the surface and deeper layers of ice was greater in the morning (−3.41, −2.20, and −5.66 °C) than in the afternoon (−2.17, −1.33, and 0.97 °C). There was a negative vertical temperature difference throughout the observation period, except during the bare ice stage in the afternoon, when a positive vertical temperature difference was observed. The vertical temperature gradient was steeper during the morning of the bare ice stage than during the other two stages.

Observational findings have unveiled the impact of lake ice and its coverings on the temperature variations across different layers of lake water and ice. However, the intrinsic physical mechanism of increasing in water temperature during the freezing period in plateau lakes remains inadequately explained. Subsequently, we will elucidate the causes of these temperature discrepancies by analyzing variations in radiation and the optical properties of both the lake water and ice.

3.3.3. Correlation Between Air Temperature and Ice/Water Temperature at Different Depths

The correlation between air temperature and ice/water temperature at various depths shows significant inconsistency (

Table 5. Correlation between air temperature and ice and water temperature at different depths.

	Ice				Water
Depth (m)	0.05 m	0.10 m	0.15 m	0.20 m	0.4 m	0.5 m	6.7 m	8.7 m	12.7 m
Correlation	0.76 *	0.74 *	0.66 *	0.5 *	0.13	0.14 *	0.21 *	0.16 *	0.11 *

Note: * denotes correlation is significant at the 0.001 level.

). The surface ice temperature almost immediately reflects changes in air temperature. Correlation analysis reveals a high correlation between air temperature and surface ice temperature at depths of 0.05 m and 0.10 m, with values of 0.76 and 0.74, respectively. This suggests that variations in air temperature have a significant impact on surface ice temperature, indicating the sensitivity of surface ice to environmental climate changes.

As depth increases, the correlation between ice temperature and air temperature decreases significantly. At depths of 0.15 m and 0.20 m, the correlations are 0.66 and 0.50, respectively, indicating that the effect of air temperature on deeper ice temperatures gradually diminishes. At depths of 6.7 m, 8.7 m, and 12.7 m, the correlation with water temperature further decreases to 0.21, 0.16, and 0.11, respectively. This suggests that deeper water temperatures are less influenced by air temperature due to the insulating effect, and the variations in air temperature, unlike solar radiation, are difficult to penetrate and affect a deeper water body.

3.4. Radiation and Optical Parameters

3.4.1. Downward Shortwave Radiation

Throughout the entire observation period (6–24 February 2022), as shown in Figure 7, the downward shortwave radiation reaching the lake surface exhibits a weak upward trend with no significant changes, averaging 235.82 W·m⁻². The exceptions occurred with a decrease during the day in 10, 18 and 20 February, attributed to cloudy weather (with daily means of 191.97, 209.75, and 210.49 W·m⁻², respectively), and sand blowing on the 12 February (daily mean of 193.72 W·m⁻²). However, the radiation level remained high during the remaining, mostly sunny observation period. It is worth noting that the snowfall on the 5-6 February occurred at night and did not impact the radiation during the daytime.

3.4.2. Upward Shortwave Radiation and Albedo

In the snow stage, the albedo of the freshly snow-covered surface was 0.68 at maximum, and the upward shortwave radiation was 159.56 W·m⁻² on 6 February. In the case of a thin snow cover here, the underlying medium influences the albedo, resulting in the present value lower than reported for optically thick, new snow (around 0.9) [45,46]. Subsequently, due to snow metamorphosis, the albedo decreased to 0.57. Snowfall occurred on the morning of the 10 February, blanketing the lake surface with fresh snow, which possesses a higher albedo than old snow. This maintained the albedo approximately at 0.55, with the upward shortwave radiation of 124.21 W·m⁻² on 11 February. Throughout the snow stage, the lake surface maintained high reflectivity (albedo ranging from 0.55 to 0.68) and exhibited strong upward shortwave radiation (from 110.58 to 159.56 W·m⁻²).

In the sand stage, the albedo of the sand-covered ice was low, leading to a sudden decrease from 0.49 to 0.37 in 12–13 February, while the corresponding upward shortwave radiation decreased from 93.94 to 53.54 W·m⁻². By the 16 February, the sand cover remained on the ice surface, and the albedo continued to decrease, stabilizing at 0.15 due to the melting and deterioration of the snow present in the sand. The daily mean upwelling shortwave radiation further declined, reaching a minimum of 32.90 W·m⁻² on 18 February in the sand stage. The average daily upward shortwave radiation in the sand stage varied greatly (32.90~93.94 W·m⁻²), and the albedo remained low (0.15~0.49).

During the bare ice stage, much of the particles (snow and sand) on the lake surface were dispersed by strong winds, leaving the ice surface bare. The albedo during 19–24 February at noon remains stable, with a consistent range of 0.15 to 0.18 without any notable fluctuations. The upward shortwave radiation increased slightly during the bare ice stage but still maintained a low level (32.72 to 50.78 W·m⁻²).

3.4.3. Net Solar Shortwave Radiation

During the snow stage, the net shortwave radiation incident on the lake surface was relatively small (89.46 W·m⁻²), due to the high reflectivity of the snow. In the sand stage, the sand absorbed a large quantity of radiation (174.00 W·m⁻²). As the wind blew the sand and snow away, the net shortwave radiation rose quickly and eventually stabilized. Higher net radiation (209.39 W·m⁻²) was absorbed by the ice and the under-ice water column during the bare ice stage due to the increased solar radiation intensity and decreased albedo. The net shortwave radiation incident on the lake surface during the bare ice stage was larger than that during the snow and sand stages due to the higher reflectance of snow and the snow–sand mixture compared to bare ice. The snow stage and the sand stage accounted for only 43% and 62%, respectively, of the net radiation in the bare ice stage.

3.4.4. Underwater Radiation of 0.7 m

Underwater radiation refers to solar shortwave radiation that penetrates to liquid water through the ice cover. During the snow stage, there was an increasing trend of underwater radiation at 0.7 m depth, rising from 4.94 W·m⁻² to 7.89 W·m⁻² during 6–9 February (Figure 8). This corresponds to the decreased surface albedo due to the aging of snow and to the increased incident radiation. The high albedo of new snow in the early morning of the 10 February resulted in a reduction in underwater radiation to 6.41 W·m⁻² on the 10 February. Subsequently, with the aging of snow, the radiation increased to 8.10 W·m⁻² on the 11 February.

The radiation level at the depth of 0.7 m substantially reduced during the sand blowing on the 12 February, reaching a minimum of 2.22 W·m⁻² on the 13 February. The reduction in radiation was attributed to the surface coverage around the site by an approximately 8 cm thick layer of sand, decreasing the penetration of radiation through the sand layer. Thus, the sand largely changed the solar forcing of ice into a surface boundary condition rather than a distributed source term. As the wind dispersed the sand particles over the lake, the thickness of the sand layer gradually decreased. This gave rise to a continuous increase in the underwater radiation, which reached 8.62 W·m⁻² at 0.7 m depth on the 16 February, exceeding the maximum of the snow stage. During the bare ice stage, the radiation at the depth of 0.7 m showed a gradual increase from 9.39 W·m⁻² to 12.97 W·m⁻² during the 20–24 February. Without the absorption of a deposited covering, the shortwave solar radiation penetrates the ice surface in large quantities.

Comparatively, the mean radiation at 0.7 m depth was significantly higher in the bare ice stage (11.70 W·m⁻²) than during the snow cover stage (7.44 W·m⁻²), with the sand stage recording the lowest values (3.42 W·m⁻²). The diurnal variation in underwater radiation peaked just after local noon (12:30–13:30), with maxima recorded at 27.71, 13.91, and 44.33 W·m⁻² for each stage, respectively. The radiation reaching the water is influenced by the thickness and optical properties of snow, sand, and ice. The underwater radiation is significantly affected by the thickness and optical attributes of the overlying snow, sand, and ice. Snow, even when thin, provides exceptional reflection, while sand absorbs considerable incoming solar radiation. The variation in underwater radiation is primarily driven by snow’s high albedo and sand’s absorptivity of solar shortwave radiation.

3.4.5. Underwater Radiation of 2.1 m

The mean underwater radiation at 2.1 m depth during the bare ice stage (6.63 W·m⁻²; range: 5.58~7.71 W·m⁻²) was higher compared to that covered with snow (average: 1.39 W·m⁻²; range: 1.10~1.66 W·m⁻²) and sand (average: 2.25 W·m⁻²; range: 1.75~2.74 W·m⁻²). The diurnal variation in underwater radiation at 2.1 m depth peaked just after local noon (12:30–14:00), with maxima recorded at 5.97, 9.33, and 27.14 W·m⁻² for each stage, respectively (Figure 8c). During the snow stage, the radiation exhibited stable fluctuations (0.56 W·m⁻²), while, in the sand and bare ice stages, it showed a clear increasing trend, with rates of 2.03 W·m⁻²·d⁻¹ and 0.34 W·m⁻²·d⁻¹, respectively. The increase during the bare ice stage was due to the gradual rise in solar radiation and thinning ice, leading to more incoming radiation, whereas the rapid rise during the sand stage was primarily caused by wind thinning the surface sand, allowing quicker radiation penetration. Additionally, radiation at 2.1 m was consistently lower than that at 0.7 m in all three stages, which can be attributable to the substantial absorption of radiation by the water layer, with the respective differences of 5.27, 0.99, and 4.56 W·m⁻². The disparity in the magnitude underscores impacts of surface conditions on the penetration of solar radiation in the deeper water.

3.4.6. K_dw and Ice Bottom Radiation

The diurnal variation of K_d for both the lake water body and ice was computed using Equation (2) (Figure 9). Throughout the snow stage, the average K_dw was 1.17 (±0.06) m⁻¹, decreasing sharply to a minimum of 0.27 (±0.05) m⁻¹ during the sand stage. In the bare ice stage, K_dw was a consistent average value of 0.39 (±0.03) m⁻¹ for an extended period, which was lower than during the snow stage but higher than during the sand stage. There are no data of the water quality in the three stages; however, the differences in K_dw may be due to variations in the spectral distribution of underwater radiation. The attenuation coefficient of snow, ice, and water, along with the albedo, shows diurnal cycles from dawn to sunset, particularly under clear skies. Generally, both K_dw and albedo are higher in the morning and afternoon compared to the solar noon largely due to the solar elevation angle. This fluctuation is most prominent during the bare ice stage, while the snow and sand stages tend to exhibit milder variations, as the covering materials (snow and sand) block solar radiation, allowing it to penetrate into the ice layer only when the radiation is sufficiently strong. Another potential factor behind the daily variation is the melting and the consequent presence of liquid water, even in small amounts.

Equation (2) was utilized to calculate the radiation at the ice bottom considering the underwater radiation at 0.7 m depth, K_dw, and the distance from the ice bottom. The results show changes consistent with the underwater radiation level. The daily peak values of the radiation at the ice bottom are 30.01, 14.77, and 46.24 W·m⁻² in the three stages (Figure 8d). The mean radiation is higher in the bare ice stage (11.70 W·m⁻²) than in the snow cover (7.44 W·m⁻²) and sand (3.42 W·m⁻²) stages. The determined attenuation coefficient of the ice is greater than that often predicted for bare ice due to the small quantity of sand present on the ice; the analysis is unable to differentiate between the attenuation in ice and sand due to the lack of data.

3.4.7. K_di and Lake Ice Transmittance

Equation (3) was employed to calculate the K_d of lake ice with its coverings, considering the radiation at the ice bottom, the radiation entering the lake surface, and the thickness of the ice. In the snow stage, the K_di of the snow-covered ice layer is 4.63 (±0.25) m⁻¹, while the K_di of bare ice is 5.36 (±0.17) m⁻¹, which is higher than that of the ice–snow layer but lower than the sand-covered ice layer, which is 6.78 (±0.47) m⁻¹. Even a small amount of sand on ice will increase the K_di, reflecting the strong absorption characteristics of sand. This is also one reason why the K_di of the bare ice stage is larger than that of the snow stage.

Figure 10 illustrates the evolution of light transmittance in February together with the daily variation described by Equation (5). The total transmittance in the snow, sand, and bare ice stages are 8%, 3%, and 11%, respectively. Combined with the reflectance values of each stage, the absorption rates of the ice with the different coverings are 31%, 65%, and 72%, respectively. In terms of the daily variation, light transmission first appears at 07:00 (08:00) CST during the sand and bare ice stages (but was delayed by one hour in the snow stage). Thus, not only do the coverings influence the intensity of light transmission but also the period with underwater light is shorter, thereby reducing the heating impact of radiation. All physical property parameters of lake ice and coverings mentioned in this chapter are shown in

Table 6. Daily average of main physical parameters in each stage.

	Snow	Sand	Bare Ice
Albedo of coverings or ice	61%	32%	17%
Absorptivity of coverings and ice	31%	65%	72%
Transmittance of coverings and ice	8%	3%	11%
Absorption-to-transmission ratio (β)	0.79	0.96	0.86
Kdw of lake water (m−1)	1.17	0.27	0.39
Ice bottom radiation (W·m−2)	7.44	3.42	11.70
Kdi of coverings and ice (m−1)	4.63	6.78	5.36
Zeu(PAR) (m)	-	-	11.50

.

4. Discussion

4.1. Mechanism of Lake Ice Temperature Variations Across Three Stages

Lake water temperature remained near the freezing point during the period of this experiment and hardly changed, indicating that heat transfer from the water body to ice was small and the lake ice heat content is not much controlled by the water layer below the ice. Meanwhile, the growth of ice at the bottom was small (1.95 cm) over the 19-day study period; thus, the latent heat released by the stage transitions was small. The transfer of strong solar radiation on the TP can be taken as the foremost factor to explain the lake ice temperature differences in the three stages (−4.02, −3.68, and −5.29 °C at a depth of 0.05 m in snow, sand, and bare ice stages, respectively).

The net solar radiation over the lake surface during the snow stage was much lower than in sand and bare ice stages (89.46, 174.00, and 209.39 W·m⁻², respectively) due to the higher reflectance of snow (61%) to that of sand and bare ice (32%, 17%). Furthermore, the net solar radiation was lessened more by the higher absorptivity of sand (the absorption-to-transmission ratio: 0.96) during the sand stage than in snow and bare ice stages (0.79, 0.86). Thus, as depicted in Figure 11, the net radiation penetrating from surface into water during the snow and sand stages was much lower than in the bare ice stage (7.44, 3.42, and 11.70 W·m⁻², respectively) owing to the strong absorptivity of sand and the strong reflectivity of snow, which was consistent with the lake ice temperature.

Another noteworthy point is the diurnal variation in the three stages. The snow layer that contained air served as a blanket due to the very low thermal conductivity. Therefore, the diurnal range of ice temperature was only −5.46~−3.00 °C in the snow stage, much less compared with the bare ice stage (−10.50~−0.40 °C). The thick and high-density sand layer almost totally blocked the ice from radiation and made the ice temperature diurnal range (−4.41~−2.99 °C) even less than in the snow stage.

In the bare ice stage, the incoming solar radiation could directly reach the ice without any loss, and, at night, due to the lack of covering, the long-wave radiation loss was enhanced. This caused dramatic fluctuations of lake ice temperature and the lowest average temperature, unlike in the surface-covering stages.

4.2. Mechanism of Lake Water Temperature Variations Across Three Stages

The ice cover acts as a barrier to wind-driven momentum and prevents the wind-induced vertical mixing within the lake [17], which would be contributed by the density-driven mixing owing to the gradient of salinity and temperature of lake water. Observations showed that the salinity gradient was only 0.33 g·L⁻¹·m⁻¹ and had a small influence on vertical density difference. All the observed water temperatures were below the maximum density temperature (274.43 K) [47,48]. Thus, the density would always increase with the increasing temperature warming due to solar heating and daytime convection events in interaction with the evolution of the vertical temperature–salinity distribution. Throughout the observation period, the lake’s water body persistently conveyed heat to the ice at the ice/water interface, driven by the ice’s lower temperature. The water maintained a relatively stable temperature, which can be partially credited to the solar radiation that compensated for the heat absorbed by the ice, sustaining the water’s temperature around 0 °C, with only subtle variations observed across three distinct stages. Specifically, at a depth of 0.4 meters, the water temperature registered at −0.24, −0.29, and −0.10 °C during the snow-covered, sand-covered, and bare ice stages, respectively (Table 3). In these stages, underwater radiation predominantly dictated the thermal exchange within the lake. To elucidate these temperature discrepancies, an analysis of the ice bottom radiation was conducted to ascertain the phenomenon of energy penetration into the aqueous medium.

During the sand stage, the sand and superficial snow layer significantly reflected the incident solar radiation (32%), with the residual radiation absorbed by the intervening snow, sand, and ice layers (65%), resulting in minimal energy (3%, equivalent to 3.42 W·m⁻²) permeating the water body, aligning with the lowest observed lake water temperatures. In the snow-covered stage, the high reflectance of the snow (61%) led to a substantial reflection of downward radiation, coupled with absorption by the snow and ice layers (31%), thereby reducing the radiation reaching the water. Consequently, the underwater radiation during this stage was relatively diminutive, amounting to 7.44 W·m⁻² (8%), which sustained the lake’s lower water temperature. In contrast, during the bare ice stage, the absence of coverings allowed solar radiation to be minimally reflected by the ice surface (17%) before being substantially absorbed by the ice layer (72%), thereby heating the shallow water adjacent to the underside of the ice. Ultimately, the incident radiation penetrated the water layer with the greatest intensity (11%, equivalent to 11.70 W·m⁻²), with this direct heating effect and enhanced radiation absorption contributing to a comparatively elevated lake water temperature, marking the highest temperatures during the bare ice stage.

The transmittance of solar radiation through the ice layer and coverings plays a decisive role in determining the lake’s water temperature. Given that the ice during the bare ice stage is not entirely transparent, it exhibits a strong absorptive capacity for solar radiation, particularly in the long-wave spectrum, with this capacity intensifying as the ice thickness increases [42]. In Qinghai Lake, with an ice thickness of approximately 36.6 cm, the transmittance rates for solar radiation are as follows: 11% for the bare ice stage, 8% for the snow-covered ice stage, and 3% for the sand-covered ice stage. These data underscore the significant impact of the ice layer and coverings’ characteristics on the lake’s water temperature, with the ice layer’s absorptive and transmittance capabilities of solar radiation as pivotal factors influencing the lake’s thermal regime.

4.3. K_di, K_dw, and Euphotic Zone Depth in the Qinghai Lake

The mean (±standard deviation) K_di in Qinghai Lake is 5.36 (±0.17) m⁻¹ calculated with the observation data, which is higher than that in nine Estonian and Finnish boreal lakes and the brackish Santala Bay of the Baltic Sea (0.51~3.54 m⁻¹) and in the central Asian arid climate zone of the Wuliangsuhai (0.21 m⁻¹) [49,50,51]. Higher K_di in Qinghai Lake could be explained from three aspects. Lake ice contains impurities that consist of gas bubbles, liquid inclusions, and particles, which originate from the water body, bottom sediments, or atmospheric deposition [52]. The gas bubbles in the ice have a great impact on the scattering of light and in the liquid inclusions of brackish ice eventual CDOM (chromogenic dissolved organic matter), and algae can absorb light [53]. Additionally, sand particles brought by strong winds would thinly cover the wrinkled lake ice surface in the TP [27], and sand particles would be stuck in the ice by the freeze–thaw process, forming sand layers that absorb much light. Although the K_di in Qinghai Lake is quite big, the incident radiation is large due to its high altitude and low latitude, and sunlight does have an important role in heating the water and in providing photons for primary production beneath the ice. During the ice stable stage, with only bare ice that is the normal status for the TP lakes, there are 11.70 W·m⁻² of photosynthetically active radiation penetrating into the water in Qinghai Lake, much higher than that in northern lakes with high latitude. With the consideration of penetrated solar radiation from the lake ice bottom in lake models, the originally simulated flat under-ice water temperature in TP lakes will be improved.

Compared to the rather clustered K_di, the K_dw ranges widely from 3% to 15 m⁻¹. Also, water bodies could lead to disparate levels of attenuation of solar radiation, and the euphotic zone depth varies (0.3 to 60 m), with significant discrepancies in deep water temperatures [38,49,54]. It is very necessary to have the accurate K_dw in lake models. According to the observed PAR during the freezing period in the Qinghai Lake, the average K_dw is 0.39 (±0.03) m⁻¹, and the euphotic zone depth is estimated to be 11.50 m, which is smaller than the average water depth (21 m). In fact, many lakes on the TP have a low K_dw of 0.10~1.17 m⁻¹ (the mean is 0.26 m⁻¹) [55]. For example, the K_dw of Namco Lake in the northwestern region of the TP is only 0.14 m⁻¹ [51]. Meanwhile, many lakes in Eastern China exhibit a relatively high K_dw, such as Chaohu Lake (1.56~18.01 m⁻¹ in winter), West Lake (0.49~2.25 m⁻¹), Taihu Lake (2.45~10.42 m⁻¹), and Longgan Lake (0.71~3.72 m⁻¹). That is because these lakes are more eutrophic and shallower than the TP lakes. The dynamic effect of wind-induced turbulent eddies gives rise to the resuspension of inorganic particles from the sediment in shallow lakes of Eastern China, while the ice cover and large depth damp the resuspension in the Qinghai Lake, even under strong wind conditions.

5. Conclusions and Future Works

Based on the systematic field experiment on air–ice–water temperature and radiation transfer conducted during the ice-covered period of Qinghai Lake in February 2022, combined with high-resolution remote sensing technology (ultrasonic instruments, acoustic Doppler devices) and MODIS imagery to analyze changes in ice thickness and surface conditions, this study examined characteristics of water and ice temperatures, air–ice–water radiation transfer, and corresponding optical properties, as well as the effects of different coverings (snow, sand, and ice) on temperature and radiation transfer.

Common weather processes (e.g., snowfall, sand blowing, and strong winds) on the Tibetan Plateau can significantly alter the surface conditions of the ice cover. These coverings play an important role in the variations in ice and water temperatures. The mean ice temperature at 0.05 m beneath the ice surface for the three stages—snow cover, sand cover, and bare ice—was found to be −4.02, −3.68, and −5.29 °C, respectively. The daily ice temperature variation ranges were smaller in the snow and sand stages (1.58 and 1.04 °C) compared to the bare ice stage (8.52 °C). In contrast, the water temperature slightly increased around 0 °C (the water depth of 12.7 m: −0.18, −0.17, and 0.15 °C) with fluctuations not exceeding 0.54 °C during the entire observed ice-stable period. The lake water temperature at the depth of 2.1 m in ice-covered Qinghai Lake continued to increase to 3.87 °C before ice melted and was similar to most Tibetan Plateau lakes.

The different coverings (snow layer and sand layer) on ice exhibit distinct properties, dividing the incident radiation into reflected, absorbed, and transmitted components. For bare ice, the contributions of these three components were 17%, 72%, and 11%, respectively. In contrast, a thin 2 cm snow cover resulted in corresponding values of 61%, 31%, and 8%, while an 8 cm sand cover yielded values of 32%, 65%, and 3%. These differences explain why the ice (water) temperatures in the snow and sand stages were similar but influenced by different mechanisms: the high reflectivity of snow (61%) and strong absorption-to-transmission ratio of sand (0.96). During these stages, less solar radiation penetrated into the water (8% for snow and 3% for sand), resulting in lower water temperatures (−0.18, −0.17 °C). Additionally, the reduced absorbed solar radiation by ice (31% for snow and 65% for sand) limited diurnal temperature variations (1.58, 1.04 °C) due to the insulation of snow and sand. In contrast, during the bare ice stage, lake ice had the lowest temperature (−5.29 °C) and the greatest diurnal variations (8.52 °C). This was attributed to the absorption of 72% of solar radiation (171 W·m⁻²) by 37 cm of ice, which had a light attenuation coefficient of 5.36 (±0.17) m⁻¹.

Percentages of 8%, 3% and 11% of solar radiation penetrated into the lake water in snow, sand, and bare ice stages, respectively, which resulted in colder lake water in the first two stages. The averaged radiation of approximately 11.70 W·m⁻² penetrating through the ice layer during the bare ice stage primarily contributed to the warming of lake water during this specific period. The averaged light attenuation coefficient of water K_dw was 0.39 (±0.03) m⁻¹, which corresponded to a euphotic zone depth of 11.50 m and influenced the special thermal conditions of water temperature.