| The wetland water body and quality information extraction is a basic job to study the wetland structure function and develop the wetland scientifically. Due to the high spatial resolution and multi-spectral characteristics, GF-1 data has been widely applied in many fields. However, the limited time resolution as well as the cloud’s influence brings out few applicable data in the practice, which restricted the application of some times series analysis. The high time solution of MODIS data can be more applicable for time series analysis but with limited spatial resolution and property, which can only be applicable for large-scale researches. Based on the time-spatial information reconstruction method of GF-1 and MDIS data, this paper integrated the high spatial solution with the high time resolution effectively to obtain the information with both characteristics so as to conduct the research on temporal variation of high time resolution.This paper adopted STDFA model (Spatial and Temporal Data Fusion Model) to explore the reconstruction method of high spatial resolution data with the missing GF-1 data. The STDFA model aims to extract the temporal variation information in the time series MODIS data and then construct the images with MODIS time resolution and GF-1 spatial resolution by combing the texture information of GF-1 image aiming to make up for the missing data during the GF-1 image research. The reconstructed high spatial resolution data was adopted to detect the multi-temporal variations on water information and water quality information. The main research contents and research results are as follows:(1) The NDWI data with 16m resolution was reconstructed based on GF-1 and MODIS; the neighboring analysis was conducted on GF-1 data to obtain 16*16 pixel window so as to construct the precision of abundance map. The abundance data was introduced in STDFA model to reconstruct MODIS data. Compared with the water body data extracted by GF-1, the precision of data extracting water body information was 79.8% after the reconstruction, which provides high spatial resolution data for the follow-up multi-temporal variation.(2) Establishment of GF-1 data suspended matter concentration inversion model:the GF-1 data on May,1,2014 was selected to analyze the correlation coefficient between various bands and band combination of GF-1 data and suspended matter concentration of sub-region of Lake District. According to the actual suspended matter concentration measured by water quality sampling, the linear model, exponential model and exponential model were established to reverse the suspended matter concentration. Compared with the fitting results, the exponential model Css= 0.94e7.195B3/B2 had the maximum reversion accuracy with R2= 0.789.(3) Establishment of cyanobacterial bloom abundance relation model based on MODIS NDVI:due to the low MODIS resolution, the water bloom information extracted by GF-1 with high resolution was taken as the basis to establish the cyanobacterial bloom information extraction model concerning the relation between cyanobacterial bloom coverage ratio and MODIS NDVI. This model can be used to extract the cyanobacterial bloom coverage ratio, which was 15.49% different from that extracted by GF-1. Hence, this model can improve MODIS data’s cyanobacterial bloom information extraction accuracy and acquire high-precision and multi-temporal area during the cyanobacterial bloom changes.(4) Lake water body and quality change detection:the reconstructed high spatial resolution remote data was adopted to detect the wetland water body and quality change via trajectory analysis and time series analysis etc. The water area variation curve was consistent with that calculated by historical water level data. The cyanobacterial bloom area change curve was highly fitted with that extracted by GF-1. |
PDF Full Text Request