| The process of snow accumulation and ablation continues to affect the global energy balance,regional hydrological cycle and climate change.Temporal and spatial variation of snow cover may also affect freshwater resources in high-altitude areas and water resources management in arid areas.With the rapid development of earth observation technology,remote sensing satellite image has become an important means to study the spatio-temporal variation of snow cover.Thick clouds make it difficult for optical remote sensing to identify snow under clouds,which hinders accurate recognition of snow based on optical remote sensing images.SAR remote sensing data’s high sensitivity to physical state of snow and its feature of penetrating cloud and fog overcome the vulnerability of optical remote sensing to weather.Significant progress has been made in snow recognition based on SAR remote sensing data.However,it is found that the accuracy of snow remote sensing inversion varies greatly in different regions.Therefore,it is still necessary to conduct in-depth research on the accuracy of snow recognition in different underlying surfaces and different snow cover periods.Based on 94 sentinel-1A SAR radar images from 2017 to 2021,snow cover identification experiment was carried out in the Caiertes River Basin,and the "true value" snow cover map obtained by visual interpretation of 12 Landsat8 OLI remote sensing images with close date and cloud cover meeting the requirements was selected as far as possible to evaluate its accuracy.The composition of uncertainty in snow cover recognition based on multi-temporal SAR data is discussed,and the spatiotemporal variation of snow cover in the study area from April to June 2017 and from January to June 2018 to 2021 is further studied.The main results are as follows:(1)Based on multi-temporal C-band Sentinel-1A SAR radar data and Global Land30 global land cover data,this study considers the differences of backscattering process of SAR data in different land cover,and proposes to add land cover factor into change detection method and Sigmoid function snow inversion algorithm.The wet snow discriminant thresholds and dry snow discriminant altitude calibration factors of different land cover in the study area were given in the stable snow accumulation period(January-February),alternating snow accumulation and ablation period(March-April),and snow ablation period(May-June).(2)The average overall accuracy of snow retrieval algorithm proposed in this study is85.4%,87% and 87.4%,and the average F1-Score value is 91.1%,91.8% and 69.1%,respectively,in the stable snow accumulation period(January-February),alternating snow accumulation and ablation period(March-to-April)and snow ablation period(May-June).It shows that the overall snow recognition accuracy is high,and the snow recognition effect of the algorithm is the best in the alternate snow accumulation and ablation period.The average overall accuracy of the algorithm is the highest in the snow ablation period,but its average F1-Score value is lower than the other two snow periods,indicating that the consistency of the snow recognition in the snow ablation period is relatively poor.(3)According to the inversion results of different underlying surfaces,the algorithm has high accuracy in snow recognition of glacier,bare land,shrub and forest land in the stable snow accumulation period and alternating snow accumulation and ablation period,but poor in snow recognition of artificial surface and cultivated land.The snow recognition errors mainly come from the low altitude areas mainly covered by grassland in the southwest and central part of the study area,and the concentrated covered areas of cultivated land and artificial surface in the south of the study area.During the snow ablation period,the snow recognition errors mainly come from the low altitude area in the middle of the study area(below 2500m)and the concentrated distribution area of the cultivated land in the south of the study area,and are mostly spot-like distribution.The snow recognition accuracy of the glacier is high,but the snow recognition effect of the cultivated land and bare land is poor.(4)The average snow cover rate in the study area during the alternating snow accumulation and ablation period from 2017 to 2021 was 68.9%,68.3%,66.5%,68.7%and 63.8%,respectively,and the average snow cover rate in the snow ablation period is 25.5%,41.4%,26.1%,22.0% and 20.3%,respectively.The average snow cover rate in the stable snow accumulation period from 2018 to 2021 is 79.4%,81.7%,79.8%,and 82.2%,indicating that the snow cover in the study area has little change in the alternating period between the stable snow accumulation period and the snow accumulation and ablation period,and the snow cover rate in the snow ablation period from 2018 to 2021 shows a fluctuation change year by year.(5)The change rates of snow cover of shrubs,water bodies,bare land and glaciers were less than 1‰ from January to June in the study area from 2018 to 2021,but the change rates of snow cover of forestland fluctuated the most from May to June,increasing by 10.8% from 2017 to 2018,and decreasing year by year from 2018 to2021.(6)from the perspective of different altitudes,the variation of snow cover in januaryfebruary of 2018-2021 is mainly concentrated below 875 m,in march-to-april of 2017-2021 is mainly concentrated below 1075 m,and in may-june is mainly concentrated at900-1900 m with complex variation.From the perspective of different slopes,the variation of snow cover from January to April is concentrated below 40°,and the fluctuation of snow cover from May to June is large.From the perspective of different aspects,the variation of snow cover on the north slope,northeast slope and east slope was great from January to February of 2018 to 2019;from March to April of 2017 to2019,the variation of snow cover on the north slope,northwest slope,west slope,east slope and southwest was great;from May to June of 2017,the variation of snow cover on each slope direction was great. |
PDF Full Text Request