Font Size: a A A

Spatio-temporal Sampling Schemes Optimization For Aquatic Environment Based On Time-series Remote Sensing Data

Posted on:2019-06-07Degree:MasterType:Thesis
Country:ChinaCandidate:T T LiFull Text:PDF
GTID:2370330545492319Subject:Photogrammetry and Remote Sensing
Abstract/Summary:PDF Full Text Request
In-situ sampling is the basis for theoretical research,inversion modeling,and product authenticity testing of quantitative remote sensing,and it is also the core monitoring service of environmental protection and hydrology departments.In-situ sampling is a monitoring method with high consumption of manpower and material resources,so how to optimize the resource allocation and improve the decision-making level on the premise of ensuring the sampling accuracy is a scientific problem that needs to be solved.At present,the sampling method for lakes mainly relies on subjective methods(eg.expert experience),and it lacks objective and standardized theoretical basis and evaluation methods for sampling design,besides,the sampling process lacks consideration of prior information such as the temporal and spatial variation of water environment elements.The current researches for water environment monitoring using remote sensing technology are mainly focused on atmospheric correction,the retrieval of three elements of water body,hydrology and hydrodynamics and so on,studies on sampling methods and sampling layout optimization are still rare.Taking the water environment monitoring of the Poyang Lake as an example,this study established chlorophyll inversion algorithm for turbid waters of Poyang Lake.Based on the spatial-temporal remote sensing dataset,the sampling accuracy of commonly used traditional sampling methods(regular sampling and stratified random sampling)were evaluated systematically,and the priori analysis method of lake water environment variability considering both temporal and spatial dimension changes was proposed.The simulated annealing optimization algorithm was introduced to optimize the sampling layout of Poyang Lake and it will provide scientific and reasonable prior knowledge for lake sampling strategy optimization.The main conclusions are as follows:(1)Based on the theory of optical partition,the chlorophyll partition inversion model for turbid water in Poyang Lake was established,the three bands([1/Rrs(665)-1/Rrs(705)]*Rrs(740))model was the most accurate one for sediment-dominated area(Normalized Difference Chl-a Index(NDCI)?0.06)with the determination coefficient of 0.65 and MRE of 38.53%.The difference((Rrs(705)-Rrs(665))model was the most accurate one for non-sediment dominated area(Normalized Difference Chl-a Index(NDCI)>0.06)with the determination coefficient of 0.63 and MRE of 39.87%.(2)Based on the idea of spatio-temporal clustering,using the K-means clustering algorithm,the Sentinel satellite reflectivity datasets were divided into six clusters.Due to the spatio-temporal dynamic variation of water environment in Poyang Lake,the time-series variation of different clusters was significantly different.Combined with the semi-variogram,the results showed that the spatial variation in the central part of Poyang Lake was as high as 5,700 meters while the spatial variation in the southern and eastern part was as low as 1,200 meters.(3)In general,when the number of sampling points was reduced from 162 to 37,for the time-series 443nm remote sensing reflectance products,the prediction average absolute errors of regular sampling and stratified random sampling increased from 3.09*10-3 to 3.95*10-3,3.17*10-3 to 4.22*10-3(sr-12),respectively.For the time-series 560nm remote sensing reflectance products,the prediction average absolute errors of the two methods increased from 6.09*10-3 to 7.76*10-3,6.37*10"3 to 8.05*10-3(sr-1),respectively.For the time-series 665nm remote sensing reflectance products,the prediction average absolute errors of the two methods increased from 5.08*10-3 to 6.49*10-3,5.18*10-3to 6.83*10-3(sr-1),respectively;For the time-series 705nm remote sensing reflectance products,the prediction average absolute errors of the two methods increased from 5.87*10-3 to 7.55*10-3,6.18*10-3 to 7.83*10-3(sr-1),respectively.And the accuracy of regular sampling was better than that of the random sampling.(4)Based on the time-series clustering results of Sentinel satellite images and analysis of semi-variogram method,the number of sample sites suggested at Poyang Lake were 41,49,45,and 48 for the 443 nm,560 nm,665 nm and 705 nm band,respectively.Compared with the regular sampling method with 52 sampling sites,after using the simulated annealing algorithm to optimize the layouts,the average absolute sampling errors decreased by 46.75%.57?57%?35.54%and 43.17%for the four bands,respectively.The optimized sampling layouts of the four characteristic bands were applied to the temporal Poyang Lake chlorophyll products,the average relative errors of time-series chlorophyll concentration prediction were 17.93%,17.78%,17.79%,and 18.21%,respectively.
Keywords/Search Tags:Poyang Lake, Time-series remote sensing data, Sampling schemes optimization, Chlorophyll
PDF Full Text Request
Related items