On The Monitoring Method Of The Songnen Plain Wetland Flood Dynamics

Wetland remote sensing monitoring technology has been widely used in quantitative analysis,extraction of wetland information,water resources management,flood disaster monitoring and wetland protection.The Songnen Plain is a typical inland wetland ecosystem in northeastern China.With the disturbance of human activities and extreme hydrological conditions,the wetland hydrological pattern has changed significantly,resulting in a significant reduction in the wetland area,severe fragmentation of the wetland landscape,and lack of Related basic data research.Therefore,real-time and accurate control of large-scale wetland hydrological changes in the Songnen Plain provides reliable data sources and technical support for flood disaster monitoring,regional hydrological distributed models,and flooding of large-scale wetlands.Dynamic monitoring of scope is of great significance.In this paper,an adaptive threshold selection method is proposed to extract the flooded area of Songnen Plain based on the adarsat-2 and MODIS images of time series.Firstly,the radar image of the region of interest is selected as the reference data,and the object-oriented classification method accurately extracts the flooded area of the region of interest in different time phases.Then,six kinds of water body indices are extracted based on MODIS image as classification characteristic variables,including Normalized Difference Water Index(NDWI)?Normalized Difference Vegetation Index(NDVI)?Modified Normalized Difference Water Index(MNDWI)?Normalized Difference Pond Index(NDPI)?Normalized Difference Turbidity Index(NDTI)and Normalized Difference Moisture Index(NDMI)were used to extract wetland flooding area.The classification results of 8 m spatial resolution Radarsat-2 images in the region of interest were matched with the feature variables of 500 m resolution MODIS images.Finally,six water body indices of MODIS images were used to extract the flooding area of simultaneous phase radar images in the interest region which were used to validate the data.It were calculated that the Commission rate and Omission rate of the flooded area classification results of MODIS images under different segmentation thresholds,so that selected the optimal combination of classification features and thresholds to extract the flooded area of Songnen Plain.The accuracy of the flooded area extraction results was validated based on the measured water level data and high resolution image data,respectively.The main research results show that:(1)Object-oriented classification method was used to extract the wetland flooding area of 8 Radarsat-2 images in the region of interest from April to November 2015.The flooding area of Zhalong nature reserve in the interest region reached the maximum area of 1500 km~2 in late June and early September,and the flooding area was less than 900km~2 in late April and early November.The accuracy classification results was verified by field data,and the overall accuracy was above 80%,the classification accuracy of open water surface is the highest,and the average user accuracy and mapping accuracy are all above 95%.Therefore,four types of flooding types,namely,open water surface,submerged vegetation,non-submerged vegetation and bare soil,can be effectively extracted in wetlands,and the higher recognition accuracy can be used as reference data to select the threshold of MODIS images.(2)An adaptive threshold selection method is proposed to extract large-scale wetland flooding area in Songnen Plain by combining Radarsat-2 image with MODIS image.The scale of statistical unit of submerged area of radar image is increased to 500m by scale conversion,and the scale of MODIS image is matched.Then,the optimal combination of water body index and threshold is selected from six water body indexes by adaptive threshold selection method to extract the submerged area of Songnen Plain.According to the self-adaptive threshold,the Normalized Difference Moisture Index(NDMI)is more sensitive to the water under vegetation canopy,which is suitable for the extraction of flooded area in the low water level period of April,May and early November in the study area.The improved Modified Normalized Difference Water Index(MNDWI)has a higher recognition accuracy for open water surface,and is suitable for flooding in the high water level period of June,July,August and September.(3)Based on the measured water level data and high resolution image data,the applicability of the extraction results and methods of wetland flooding area were verified quantitatively.Based on the digital elevation model,a validated small watershed is generated.The measured water level data at the entrance of the watershed are compared with the trend of the flooding range extracted from the remote sensing images of watershed downstream.The trend is consistent.It showed that the dynamic monitoring results of the flooding range with adaptive threshold can reflect the changes of the large-scale hydrological situation in the study area.The flooding range extracted from the high-resolution Sentinel-1A image seen as a validation sample to verify the accuracy of the classification results.The results showed that the accuracy of flooding range extraction in different seasons ranged from 69.8%to 75.7%.The two validation results showed that the adaptive threshold selection method is suitable for large-scale floodplain flooding area extraction,and has certain guiding significance for other areas.