Changes in Lake Ice Phenology of Sayram Lake from 2000 to 2019

QIN Qiyong¹,²,³, LI Xuemei¹,²,³, ZHANG Bo¹,²,³, LI Chao¹,²,³, SUN Tianyao¹,²,³

¹Faculty of Geomatics, Lanzhou Jiaotong University, Lanzhou 730070, Gansu, China
²Gansu Provincial Engineering Laboratory for National Geographic State Monitoring, Lanzhou 730070, Gansu, China
³National Local Joint Engineering Research Center of Technologies and Applications for National Geographic State Monitoring, Lanzhou 730070, Gansu, China

Abstract

Lake ice phenology represents a sensitive indicator of climate change. To clarify changes in ice characteristics and their influencing factors in the Tianshan Mountains—a region with distinct alpine features—this study examines Sayram Lake in Xinjiang, China. Located in a closed alpine basin as a brackish water lake, Sayram Lake's unique geographical setting ensures that its evolution and ice conditions are minimally influenced by human activities, making it an ideal study area for lake ice phenology in this region. Using MODIS imagery, Chinese lake datasets, and meteorological data, we analyzed the phenological characteristics of lake ice on Sayram Lake from 2000 to 2019 through threshold-based extraction and trend analysis, and identified temporal changes in the climatic factors influencing lake ice phenology. The results indicate: (1) The average dates for freeze-up start, freeze-up end, break-up start, and break-up end on Sayram Lake occurred on November 2, January 18, April 26, and May 17, respectively. The average complete ice duration and ice duration were 99 days and 196 days, respectively. (2) Both break-up start and break-up end dates showed advancing trends, while the freeze-up end date also advanced, correlating positively with corresponding monthly average temperatures. The complete freezing period extended, whereas the overall ice duration shortened. (3) Sayram Lake exhibited identical spatial patterns in freezing and thawing, with littoral zones being the first to freeze and the first to thaw. (4) Changes in Sayram Lake ice phenology result from the combined effects of intrinsic lake conditions (morphometric factors, shoreline complexity, etc.) and climate change (temperature, accumulated negative temperature, etc.).

Keywords: lake ice; phenology; MODIS; Sayram Lake; climate change

1. Study Area Overview

Sayram Lake, historically known as "Jinghai" (Clear Sea), is the highest-altitude and largest alpine brackish water lake in Xinjiang \cite{}. Situated between 80°39′~81°30′E and 44°27′~44°45′N, the lake is nestled within a closed alpine basin \cite{}. The lake basin is surrounded by mountains with abundant natural resources, including wetlands, grasslands, and forests distributed across elevation gradients, along with existing glaciers. The lake plays a crucial role in conserving mountain water sources and regulating climate and environmental conditions in the region and throughout northern Xinjiang \cite{}. The lake surface is elliptical, covering an area of 2071 km² with a maximum water depth of 23.4 km \cite{}. The region experiences a temperate continental semi-arid climate. Due to its high elevation, ample water vapor from the Atlantic Ocean is lifted by topographic effects, creating abundant local precipitation that forms a "wet island" in the arid northwestern region. Simultaneously, under global warming, substantial snow and ice melt has increased inflow to the lake. Sayram Lake has no surface outlet, relying entirely on natural evaporation and seepage through the lake bottom to maintain dynamic water balance \cite{}. Furthermore, owing to its unique geographical location, the lake's evolution and ice conditions are rarely affected by human activities, with humans and nature coexisting harmoniously and the natural ecosystem maintaining basic equilibrium.

2. Data and Methods

2.1 Data Sources and Processing

Lake ice data: Previous research indicates that MODIS data achieves overall accuracy of 87.5%~94.0% for land cover identification in the Xinjiang region, effectively reflecting the actual conditions of Tianshan landforms \cite{}. We selected MODIS snow product data with 500 m spatial resolution provided by NSIDC to retrieve attribute parameters including freeze-up start, complete freeze-up, break-up start, and complete break-up of Sayram Lake ice. This data is based on the Normalized Difference Snow Index (NDSI) and represents a comprehensive classification image suitable for this study's requirements.

Chinese lake dataset: \cite{} utilized Landsat imagery and topographic maps, employing semi-automatic water extraction with manual visual inspection and editing to complete research on the number and area changes of detailed Chinese lakes (greater than 1 km²) over the past 50 years.

Meteorological data: Temperature data were downloaded from the China Meteorological Data Network (http://data.cma.cn/). Due to data availability limitations, 2018-2019 data were obtained through linear fitting. Data preprocessing involved format conversion and reprojection of MODIS images, followed by batch clipping using ArcGIS to extract the Sayram Lake area.

2.2 Research Methods

Lake ice phenology research investigates the periodic formation and melting of ice covers on water bodies and their temporal variations due to seasonal and interannual climate changes. Ice begins to form when lake surface temperature drops below 0°C. Ice cover formation affects water bodies at high altitudes where cold season temperatures remain below 0°C for extended periods. To avoid extreme weather impacts on ice phenology extraction, researchers employ different methods to obtain four key temporal parameters: freeze-up start, complete freeze-up, break-up start, and complete break-up.

Some researchers calculate the percentage of open water area relative to maximum lake area throughout the year to determine lake ice phenology, while others calculate the percentage of ice-covered area relative to maximum lake area \cite{}. Taking Sayram Lake in 2001/2002 as an example (Fig. 1), the correlation between open water area and lake ice area is extremely strong, with both methods yielding identical results. Based on literature review, the freeze-up start date is defined as the date when pure pixels of ice first appear on the lake surface in the second half of the year; the complete freeze-up date is when the ice proportion reaches 90% and can be continuously maintained; the break-up start date is when the ice proportion falls below 90% and can persist in a melting state; and the complete break-up date is when the ice proportion first reaches 10% \cite{}.

During a study period, different researchers define the lake freezing period differently. Four variables describe ice cover duration: (1) Ice duration (ID) is the period from freeze-up start to complete break-up, representing the interval between freeze-up start (FUS) and break-up end (BUE); (2) Complete ice duration (CID) is the interval from complete freeze-up to break-up start, representing the period between freeze-up end (FUE) and break-up start (BUS) \cite{}. For simplicity, ice cover spanning calendar years is counted as belonging to the same ice cycle.

The calculation methods are as follows \cite{}:

$$
\begin{align}
FUS, & \text{if } IA \geq 0.1 \times LA \
FUE, & \text{if } IA \geq 0.9 \times LA \
BUS, & \text{if } IA \geq 0.9 \times LA \
BUE, & \text{if } IA \geq 0.1 \times LA
\end{align}
$$

where $IA$ is lake ice area, $LA$ is maximum lake area, $FUS$ is freeze-up start, $FUE$ is freeze-up end, $BUS$ is break-up start, and $BUE$ is break-up end.

The ice phenology algorithm extracts the histogram from all MODIS images in the time series and plots annual curves of the ratio of lake ice area to lake area. Ice phenology events are identified as threshold crossing points on these curves at 10% and 90% values. If multiple threshold crossings occur during freeze-up and complete break-up periods—potentially caused by wind events and refreezing during melting—the exact dates are determined through linear interpolation between the last point below threshold and first point above threshold.

2.3 Validation of Lake Area Extraction from MODIS 8-Day Composite Data

This study validates the accuracy of lake ice area derived from MODIS 8-day composite data using the Chinese lake dataset and supplementary literature on Sayram Lake area derived from Landsat data. The validation compares the sum of lake ice and open water areas extracted from MODIS 8-day composites against reference values. Results show that the maximum lake area derived from MODIS is more accurate than the average (Table 1), with a maximum lake area error rate of only 0.998% (p < 0.01). Additionally, the strong correlation between open water area and lake ice area (R² = 0.998, p < 0.01) demonstrates that lake ice area extracted from MODIS 8-day data has good accuracy, making phenology calculations reliable.

3. Results and Analysis

3.1 Meteorological Conditions

Based on temperature data from Wenquan Meteorological Station—the nearest station to Sayram Lake—Sayram Lake experiences an average annual temperature of 4.1°C and average annual precipitation of 255.07 mm. Monthly temperature and precipitation distributions (Fig. 3) show that precipitation concentrates in May-September, while temperatures peak in June-August. The lake ice gradually forms from November onward when temperatures drop below 0°C, beginning in shallow areas and expanding toward the center until complete freeze-up. In spring, rising air temperatures above the ice and water temperatures below cause the ice layer to thin and eventually break. Overall, the climate of Sayram Lake creates favorable conditions for ice development while temperature increases promote ice melt.

3.2 Lake Ice Phenology Characteristics of Sayram Lake

Using MODIS imagery from 2000-2019, we obtained four key phenological parameters: freeze-up start, complete freeze-up, break-up start, and complete break-up (Table 2). The average freeze-up start date was November 2, with the earliest on October 18 and latest on November 24, requiring an average of 77 days to reach complete freeze-up by January 18 of the following year. The earliest complete freeze-up occurred on January 5 and latest on February 5. Break-up typically began in late April, with an average start date of April 26 (earliest April 15, latest May 5). After approximately 21 days of melting, complete break-up occurred on average by May 17 (earliest May 1, latest May 30).

Over the 20-year period, Sayram Lake showed significant interannual variability in both ice duration and complete ice duration. Average ice duration was 196 days (range: 146-237 days), while average complete ice duration was 99 days (range: 64-114 days). Both break-up start and complete break-up dates showed advancing trends at rates of -0.03 d·a⁻¹ and -0.44 d·a⁻¹, respectively. The complete freeze-up date also advanced at -0.25 d·a⁻¹, correlating with decreasing January temperatures. Complete ice duration showed a lengthening trend (0.01 d·a⁻¹), while overall ice duration shortened (-0.43 d·a⁻¹). Temperature emerged as the primary factor influencing lake ice phenology.

3.3 Freezing and Thawing Patterns of Sayram Lake

The ice formation process typically begins along shallow littoral zones. Sayram Lake exhibits freezing/thawing patterns consistent with Qinghai Lake, starting from edge shallow areas and gradually expanding to the center. During freeze-up, the lake may experience complex cycles where ice begins forming, then partially melts, then refreezes until complete coverage. Similarly, during break-up, melting may be interrupted by refreezing events, demonstrating the complexity of Sayram Lake's ice regime.

Spatially, Sayram Lake shows identical patterns in freezing and thawing: littoral zones freeze earliest and melt first. Ice begins breaking in late April, starting from the northeastern and eastern shores and gradually expanding toward the center. Compared to the freezing process, melting proceeds more rapidly, lasting only about 40 days. This pattern differs distinctly from lakes on the Tibetan Plateau.

3.4 Analysis of Factors Influencing Sayram Lake Ice Phenology

We analyzed relationships between lake ice phenology and various factors including lake area, mean annual temperature, accumulated negative temperature, days with temperature < 0°C, shoreline length, and lake morphometric factors (shoreline length to lake area ratio), with significance testing applied \cite{}. Temperature data came from Wenquan Meteorological Station, lake area was derived as the maximum value of ice and open water areas from MODIS data, and shoreline length was obtained from the lake dataset through interpolation (Table 3).

Freeze-up start is primarily influenced by lake area, days with temperature < 0°C, shoreline length, and lake morphometric factors. Larger lake area (greater heat capacity) correlates with later freeze-up start, while more days with temperature < 0°C and more complex shorelines correlate with earlier freezing. Complete freeze-up is most strongly affected by mean annual temperature, days with temperature < 0°C, and shoreline length, with mean annual temperature showing significant negative correlation—related to decreasing January temperatures in recent years. Break-up start and complete break-up are mainly influenced by lake area, shoreline length, and lake morphometric factors, with larger lakes and more complex shorelines melting earlier. Mean annual temperature also significantly affects complete break-up.

Accumulated negative temperature and days with temperature < 0°C are the main factors affecting the freezing period, showing positive correlation with freeze duration. Lake morphometric factors and shoreline complexity also influence ice phenology, while lake ice conditions conversely affect local climate and water-atmosphere mass and energy exchange.

4. Discussion

This analysis of Sayram Lake ice phenology from 2000-2019 differs from previous studies on other lakes. Under global warming, many researchers \cite{} have reported delayed freezing and advanced melting dates for lakes in the Northern Hemisphere, contrasting with our findings of advanced complete freeze-up at Sayram Lake. However, studies on the Tibetan Plateau \cite{} showing earlier freezing in northern areas and earlier complete melting in southern areas align with our spatial pattern results.

The advanced freeze-up timing at Sayram Lake may relate to: (1) decreasing January mean temperatures, and (2) differential lagged responses of ice phenology to climate factors, manifesting as regional and topographic variations. While this study provides valuable data for data-scarce arid/semi-arid regions, limitations include: (1) lack of consideration for salinity, wind speed, and geological structure effects; (2) need for field verification of some conclusions; (3) absence of comparative multi-scale data; and (4) relatively short 20-year study period. Future research should examine longer time series and extend to the entire Tianshan region, while investigating feedback effects of ice phenology on regional climate.

5. Conclusions

Based on MODIS imagery, Chinese lake datasets, and meteorological data, this study investigated ice phenology changes and freezing/thawing patterns in Sayram Lake—a uniquely alpine, closed-basin lake—and analyzed influencing factors:

1) Average freeze-up start occurred on November 2, with complete freeze-up by January 18 after 77 days of freezing. Break-up started on April 26, with complete break-up by May 17 after 21 days of melting. Average complete ice duration was 99 days and average ice duration was 196 days.

2) From 2000-2019, complete freeze-up dates advanced at -0.25 d·a⁻¹, associated with decreasing January temperatures. Both break-up start and complete break-up advanced at -0.03 d·a⁻¹ and -0.44 d·a⁻¹, respectively, related to increasing April-May and May-June temperatures. Complete ice duration extended (0.01 d·a⁻¹) while overall ice duration shortened (-0.43 d·a⁻¹).

3) Spatial freezing and thawing patterns were identical, beginning at shorelines and expanding toward the center. Ice breakup began in late April from the northeastern shore, with melting lasting only 40 days—considerably faster than freezing.

4) Ice phenology variations resulted from combined intrinsic and climatic factors. Days with temperature < 0°C and shoreline complexity were key for freeze-up timing, while lake area, shoreline length, and morphometric factors primarily controlled break-up timing. Temperature was the dominant factor overall.

This study provides supplemental data for data-scarce regions and scientific basis for understanding climate change impacts on Sayram Lake and the broader Tianshan region.

References

[References section preserved as provided in the original text, with all citations maintained in their original format.]