[Insert Figure 2 about here]
Figure 2: Flowchart of MODIS LST product processing procedure.
2.4 Statistical analysis of LWST
The mean LWST was obtained by averaging daytime and nighttime LST for all pixels within composited images over a specific period and for a given lake. The mean annual LWST for daytime or nighttime was computed by aggregating the 8-day LWST composite data for the corresponding year (Ke & Song, 2014). Diurnal temperature differences (DTDs) of LWST represented the difference between daytime and nighttime temperatures. The starting point of the freeze-over month was determined by referencing to day-night averaged LWSTs of lakes < 0 °C. The end of the frozen period was determined by referencing to day-night averaged LWSTs of lakes at > 0 °C.
Intra-annual variation in LWST was analyzed by averaging daytime and nighttime LWST during ice-free seasons over the 16 years study period. For each year, we mapped the annual average LST that was calculated using 8-day LWST time-series on a pixel-by-pixel basis. Additionally, since ice break-up date, freeze-up date and ice-free duration are directly related to LWST, these three variables were calculated by counting the accumulated days of mean temperature (i.e., [daytime + nighttime]/2) above and below 0 °C for each lake; for all these calculations, the median Julian day for each 8-day composite was used.

3. Results

3.1. Patterns of Eurasian lake surface temperatures
As shown in Figure 3(a), there was a discernable contrasting pattern in the annual average daytime LWST of Eurasian lakes derived from MODIS products. Lakes with high WST were mainly distributed in the Iran Plateau, the South Asian subcontinent, the Indo-China Peninsula, and Southeast Asia. As expected, lakes at higher latitudes, especially those situated in the boreal or the Taiga biome, exhibited low WST; lakes in the Tibetan Plateau also featured low WST (Figure 3(a)). It should also be noted that lakes with WST in the ranges 17–23 °C, 23–30 °C, and >30 °C intermingled in the Indo-China Peninsula, as well in the South Asian subcontinent (Figure 3(a)). Thus, besides solar radiation and air temperature, other factors such as water depth, volume and source likely regulate the variations in LWST. With respect to nighttime MODIS measurements (range: –11 °C to 26 °C; Figure 3(b)), the spatial pattern of LWST was generally similar to that of the daytime measurements. However, the pattern of nighttime LWST was more homogeneous in its distribution, more closely resembling that of air temperature at the Eurasian continental scale.
The 1,098 large Eurasian lakes examined in the present study exhibited identifiable spatial variations in their diurnal temperature difference (DTD). Evidence of a latitudinal pattern in DTD is apparent (Figure 3(c)). Lakes with larger DTDs were mainly distributed in the arid and semi-arid climatic regions (Figure 3(c)), although some lakes situated in high elevation areas also exhibited large DTDs. Careful examination of the data also revealed that large and deep lakes (e.g., Lakes Baikal, Issyk, Qinghai on the Tibetan Plateau; Table 2) showed lower DTDs compared to small and shallow lakes which exhibited large DTDs. Therefore, it may be inferred that large and deep lakes with huge water storage capacity possess more stable thermal regime and smaller DTDs.