Comparison of lake surface water temperature retrieval algorithms:A case study of Lake Qiandaohu
Lake surface water temperature is an important indicator of water quality,lake physical environment,and climate change.Monitoring lake surface water temperature and understanding its spatiotemporal variations are critical for local governments to protect lake ecosystems.Remote sensing is an effective method to monitor lake surface water temperature,and many algorithms have been developed and applied to retrieve lake surface water temperature.However,the suitability of these algorithms varies in different lakes.Especially,the suitability of these algorithms in deep,oligotrophic-to-mesotrophic lakes still needs to be discussed.Thus,taking Lake Qiandaohu,China as the study area,we attempt to validate the performance of various land surface temperature retrieval algorithms,analyze the sensitivity of the parameters in each algorithm,and map the spatiotemporal distribution of lake surface water temperature.In this study,six land surface temperature retrieval algorithms(i.e.,radiative transfer equation algorithm,monowindow algorithm,generalized single-channel algorithm,practical single-channel algorithm,and two split-window algorithms)were selected to retrieve lake surface water temperature using Landsat 8 data in Lake Qiandaohu.The performance of these algorithms and the Landsat 8 Collection 2 Level-2(C2L2)temperature product were validated with in-situ buoy data.By applying the best performing algorithm to 37 cloud-free Landsat 8 data collected from 2013 to 2021,the spatial and temporal distribution of lake surface water temperature in Lake Qiandaohu were mapped.Furthermore,the sensitivity of the relevant parameters(i.e.,water surface emissivity,effective mean atmospheric temperature,atmospheric water vapor content,upwelling radiance,downwelling radiance,and atmospheric transmittance)in each algorithm were explored.The results showed the following:(1)For bands 10 and 11 of Landsat 8 data,the most suitable water surface emissivity in Lake Qiandaohu is 0.9926 and 0.9877,respectively.(2)The accuracy of the split-window algorithm is better than that of the single-channel algorithm,and the estimation accuracy of Landsat temperature product is moderate.The split-window algorithm,SWA_G,showed the optimal performance,with a mean absolute percentage error(MAPE)of 7.61％and a root mean square error(RMSE)of 2.0 ℃.The MAPE and RMSE values of the Landsat 8 C2L2 temperature product were 9.33％and 2.08 ℃,respectively.(3)Lake surface water temperature in Lake Qiandaohu has considerable spatial and temporal variation.Seasonally,the lake surface water temperature of Lake Qiandaohu has the lowest value(14.2±0.6 ℃)and highest value(31.0±0.5 ℃)in winter and summer,respectively.Spatially,the lake surface water temperature was higher in the northwest segment(23.0±0.3 ℃)and southwest segment(22.8±0.2 ℃)and lower in the northeast segment(22.2±0.3 ℃).(4)The radiative transfer equation algorithm is sensitive to the upwelling radiance and atmospheric transmittance.The monowindow algorithm shows less sensitivity to the effective mean atmospheric temperature and atmospheric transmittance.In conclusion,in Lake Qiandaohu,the Split Window algorithm has the best performance and the least dependency with atmospheric parameters.By contrast,the single-channel algorithm is suitable for retrieving long-term lake surface water temperature utilizing Landsat series data.Our study validated the performance of various land surface temperature retrieval algorithms in a deep,oligotrophic-to-mesotrophic lake and provided a reference for the remote estimation of lake surface water temperature in other similar lakes.
remote sensinglake surface water temperatureLandsatLake Qiandaohuwater surface emissivitysensitivity analysis
NETL
NSTL
万方数据
