| The Tibet Plateau is the birthplace of many rivers and lakes,and the changes of its lake area play an important role in the regional water cycle.In order to accurately understand the spatial and temporal changes of different types of lakes on the Tibet Plateau in the past 15 years from 2003 to 2018,this paper introduces the Deeplab v3+algorithm with the best effect in the field of semantic segmentation for the extraction of plateau lakes,and proposes a lake extraction method based on Deeplab v3+,compared with the conventional water body index method,the the feasibility of the method was verified,combined with the Landsat series of multi-phase remote sensing images,we respectively obtained lake distribution data that each lake area is more than1 km2on the Tibetan Plateau in 2003,2008,2013,and 2018.After extracting the lake distribution data in the 4 periods,the overall changes of the lake and the changes of 24 individual lakes were analyzed,which provides the latest data reference for the study of the typical lakes in the Tibet Plateau,and has important guiding significance for the regional environmental monitoring and the rational use of water resources.The main conclusions of the study are as follows:(1)The lake extraction method based on Deeplab v3+is suitable for the extraction of plateau lakes.Compared with the commonly used water body indexes:NDWI,MNDWI,AWEInsh,its accuracy can fully meet the requirements of water body extraction,it has a good distinction between ice,snow and lake.and has a good adaptability in the extraction of plateau lakes.(2)During the 15 years from 2003 to 2018,the number of lakes and the total area of lakes have kept increasing.The total area of lakes has continued to increase for 15years,but the expansion rate of the total area has continued to decline,with an average annual growth rate of 377.66km2yr-1.Among them,the lake area has the fastest growth rate between 2003 and 2008,and the lowest area growth rate between 2013 and2018.The total number lakes has grown steadily.The total number of expansions and new lakes has always been higher than the total number of shrinking and drying lakes.The total number of lakes has grown the fastest between 2013 and 2018.The number of smaller lakes(10km2)on the Tibet Plateau increased most significantly from 2003 to 2018.Lakes with an area between 10.0 and 50km2were relatively stable in number.The annual average fluctuation of lakes with larger areas(50km2)is small during these 15years.(3)From 2003 to 2008,most of the expanding lakes were distributed in the northern part of the inner watershed.The further south,the more stable the lake.From2003 to 2018,the lakes in most areas of the Tibet Plateau were in a state of rapid expansion,with an average annual change rate of about 2.41%,and stable lakes were mainly distributed on both sides of the Tanggula Mountains.From 2008 to 2013,the expansion rate of the total lake area slowed down as a whole,and the annual average rate of change of the lake showed a characteristic of decreasing from the northeast to the southwest of the study area.The lakes in the northeast area of the inner watershed still maintained a rapid expansion trend,the further southwest,the annual change of the lake becomes more and more stable.From 2013 to 2018,the lakes on the Tibet Plateau began to shrink in size.The originally rapidly expanding lakes began to slowly shrink.The large lakes in the south of the inner watershed also changed from a stable state to a slowly shrinking state.The number of stable lakes decreased sharply.The violent change of the lake is more intense than in previous years.Statistics are based on watershed as a unit:During the 15 years,a total of 322 lakes in the watershed expanded,58 lakes shrank,and the number of stable watersheds was 206.Among them,the inner watershed of the Tibet Plateau is the area with the most dramatic changes.(4)From the perspectives of the geographical location of the lake,the climate of lake area,and the way of lake water replenishment,the analysis of the representative lake area change direction in 24 watersheds shows that the lakes in the same watershed generally have the same change trend,but are affected by regional characteristics.Environmental factors have a greater impact. |