Research On Spatio-temporal Change Monitoring Of Cropland Quantity And Quality By Remote Sensing

The quantity and quality of cropland are important basic informations that determines agricultural production,utilization of agricultural resources and formulation of agricultural policies.The high-accuracy spatial distribution of cropland,the multi-level characteristics of cropland utilization pattern and its change,and the dynamic change of cropland quality are of great significance to the monitoring of agricultural conditions,field management and food security.Remote sensing technology has been widely used in cropland distribution mapping,cropland utilization pattern analysis and cropland quality spatio-temporal change monitoring because of its advantages of large scope and high timeliness.However,there are still some problems and challenges in the selection of cropland synergy mapping approaches,establishing cropland utilization pattern indicator system and the dynamic extraction of cropland quality indicator.This study focused on the cropland in China and Hubei province.In addition,multi-source remote sensing data,meteorological data,topographic data,cropland and soil sample data,and agricultural statistical data were used to conduct remote sensing monitoring research on spatio-temporal change around the cropland quantity and quality in large sacle..Based on the research of cropland quantity,the cropland quality is studied deeply,and the relationship between them is analyzed comprehensively.The main contents and conclusions of this study are as follows:(1)Multi-source remote sensing data synergy of cropland mapping.Based on the geographic weighted regression(GWR)and the modified fuzzy agreement scoring(MFAS),seven remote sensing land cover datasets were used to map the cropland in China.Different scenario combinations were set up from the three aspects of training samples quantity,input satellite-based datasets quality and landscapes.The advantages,disadvantages and regional suitability of different data synergy approaches were analyzed.The strategy of fusion algorithm selection was proposed and the optimal cropland distribution map in China was obtained.The results showed that the number of training samples,the quality of input satellite-based datasets and landscapes were the three main influencing factors to determine the accuracy of multi-source remote sensing data synergy approaches.The selection of approaches depended on the input dataset,landscapes and application purpose.The number of training samples was the key to determine whether the overall accuracy of GWR was higher than that of MFAS.The MFAS method was more sensitive to the quality of input satellite-based datasets and landscape changes.MFAS was the optimal choice from the perspective of producing a global or regional cropland spatial distribution map for global economy,biophysics and other land use models.From the perspective of producing high-accuracy large-scale cropland percentage maps and spatial distribution maps,GWR was the optimal choice.(2)Establishing multi-level cropland utilization pattern indicator system and analyzing cropland spatio-temporal changes.Based on the GlobeLand30 data with a spatial resolution of 30 meters,a series of indicators such as the cropland spatial distribution and its change,the cropland area and its change,the conversions between cropland and other land use types,the multi-cropping index and its change,and the fragmentation index and its change were calculated to establish a multi-level and high-accuracy indicator system for systematical and comprehensive analyzing cropland utilization pattern and its change in China.The results showed that from 2000 to 2010,China's cropland area decreased and accompanied by the increasing trend of fragmentation.Existing cropland protection policies have not effectively alleviated this problem.Urbanization was the first reason of cropland loss and fragmentation,and “Green for Grain Project” was the second reason of cropland loss.Compared with the negative state of cropland area and fragmentation,the intensive degree of cropland use in China has been improved positively and significantly.(3)Remote sensing monitoring method of cropland quality spatio-temporal change based on machine learning.Soil organic matter was selected as the representative of cropland quality index to explore the method.Time-series MOD09A1 data,meteorological data and topographic data were used as the basic data.Slope,aspect,topographic wetness index,the annual mean and annual maximum values of MODIS surface reflectance and vegetation indices were calculated.Four machine learning algorithms were used to predict the spatial distribution of soil organic matter in cropland for multi-year in Hubei,China.The spatio-temporal change of soil organic matter in cropland was analyzed.The relationship between the quantity and quality of cropland in Hubei province was analyzed.The results showed that the surface reflectance and vegetation indices of MODIS data were the key factors to predict the soil organic matter content.Gradient boosting regression trees model had the best ability on predicting spatial distribution of soil organic matter content in cropland in Hubei province.The soil organic matter content of cropland in Hubei province is “high in the south and low in the north”,and showed a slight increase from 2000 to 2017.A large amount of land with decreasing soil organic matter content in northern Hubei has been reclaimed to cropland in the past 18 years,while a large portion of high-quality land in eastern Hubei was lost due to land use change(such as urbanization).