| The spatial pattern and spatio-temporal dynamic change of surface cover information are vital to the study of global and regional changes,sustainable development and ecological protection.Among them,the planting information and evolution pattern of cropland(paddy,dry land)cover elements have important practical significance for food security risk assessment,ecological function research,water resources management and so on.The development of satellite remote sensing has led to the emergence of multiple sets of land cover products using remote sensing images as data sources.However,there are great differences in data source selection and mapping methods for different products,which makes the application value of land cover products affected to a certain extent.Therefore,it is extremely important to evaluate and analyze these land cover products.Analysis of relevant research,the evaluation and analysis of different products have been carried out in existing research,but there are few studies on the consistency and distribution of medium and high-resolution products from a qualitative and quantitative comprehensive perspective for different geographical partitions.In addition,the existing single-element cropland products can be used as auxiliary data for remote sensing interpretation of cropland and its fine second-class types.However,at present,few studies have been done to integrate these products into land cover mapping,resulting in a great waste of resources.In response to the above-mentioned research deficiencies and problems,this paper proposes a set of "land cover product analysis-element information extraction and time-series data product production-analysis of elements spatio-temporal evolution" systematic process supported by multi-source information big data and extraction and spatio-temporal analysis of paddy planting area in Cambodia as an example,in order to make full use of existing land cover products and auxiliary information to guide the development of more accurate surface cover data production.The main research contents of this paper are as follows:1.For the first time,a multi-level and multi-perspective analysis and research on the consistency evaluation of all element types of existing medium-resolution land cover products was carried out.This paper comprehensively analyzed the consistency and influence factors of land cover products,including Globeland30-2010,GLC?FCS30-2015,FROM?GLC,SERVIR MEKONG and RLC in South and Southeast Asia from qualitative and quantitative perspective by using theoretical methods such as area,spatial superposition,confusion matrix and landscape index.The results showed that when studying ecosystem functions through landscape patterns,land cover data should be selected reasonably according to the research purpose,and the scale effect should be emphasized.The consistency of the above-mentioned land cover products was relatively high in regions with low surface heterogeneity and obvious dominant land types.On the contrary,the consistency was relatively low in regions with high surface heterogeneity and obvious fragmentation of landscape.The research methods and conclusions of this paper can provide guidance for the subsequent improvement of the quality of land cover mapping,and can also provide an important reference for selecting suitable land cover products for research on global climate change and global ecological environment changes.2.For the first time,a more detailed and comprehensive consistency evaluation study on single-element cropland and its fine types of the existing medium-resolution land cover products,including GFSAD30 SEACE,GLC?FCS30-2015,FROM?GLC2015 and SERVIR MEKONG was carried out.The evaluation results of the research region in Cambodia showed that:(1)the consistency of the first-class cropland for different products was higher,but the consistency of the second-class cropland was lower.Especially,the difference of paddy and orchards was more obvious.(2)In terms of topographic influence,the consistency of elevation in the region of 50-200 m was low.The main reason is that the spectral characteristics of cropland and forest,grassland and shrub types are similar in this region,and it is difficult to distinguish accurately in remote sensing images.The factors of elevation and slope are significantly related to the inconsistency of cropland,and its impacts on the classification accuracy should be considered in the subsequent mapping of cropland and its fine types.(3)The overall accuracy of the first-class cropland type of FROM?GLC2015 products was low,while that of the other three products was higher,with an overall accuracy of more than 82%.The precision of second-class cropland types was generally low,especially for GFSAD30 SEACE,GLC?FCS30-2015 and FROM?GLC2015 products.The research results of this paper can not only provide reference for land cover product users to select appropriate data in cropland resource monitoring and cropland quality evaluation,but also provide guidance for land cover product producers how to quickly and accurately obtain cropland and its fine type.3.Under the guidance of the aforementioned consistency analysis,with the help of remote sensing big data processing platform GEE(Google Earth Engine),a large-scale and highprecision rice planting area extraction and time-series data product production model with multi-source information cooperation was constructed,which solved the problems of low utilization rate of existing land cover products,rapid and high-precision extraction of largescale rice planting area in tropical cloudy areas.Taking the study area of Cambodia as an example,(1)semi-automatic high-precision sample acquisition was realized by using different land cover products to assist training sample selection.On this basis,rice planting area extraction with an overall accuracy of 89.90% was completed in cooperation with rice phenological features,Sentinel remote sensing images and topographic features.(2)Based on the time-series medium-resolution Landsat images provided by GEE platform,a method was proposed to extract NDVI from four quarters of a year to characterize the phenological characteristics of vegetation,and then effectively distinguished different vegetation types.The production of time-series rice data products in the study area from 2000 to 2010 was made,and the overall accuracy was higher than 88%.(3)Compared the results with SERVIR MEKONG and ESACCI-LC products,the rice planting area mapping method in this paper was better in accuracy and spatial expression effect.This study revealed the potential of GEE for large-scale rice planting area mapping in cloudy areas and provided a reference for rapid acquisition of rice data products in other regions of Southeast Asia that are vulnerable to cloud effects.4.Based on the time-series rice data products extracted by this method,the analysis and study on the spatio-temporal evolution pattern of rice planting area and its resource utilization in Cambodia for 15 years was carried out,which filled in gaps in the mapping analysis information in the long-term rice planting area of Cambodia in Southeast Asia.The evolution trajectory,pattern change and utilization degree of rice planting area were comprehensively analyzed from the perspectives of land use dynamic degree,gravity center transfer,landscape index and multiple-crop index.The results showed that the rice planting area in Cambodia expanded rapidly for 15 years,and the rice planting center of gravity moved to the northeast.For 15 years,the fragmentation degree and shape complexity of rice landscape were higher and higher,and the connectivity between patches was lower and lower.The multiple-crop index of Cambodian rice field was at a low level in each period,which has great potential for development.The possible reason was that the population growth and socio-economic development in Cambodia have promoted the pattern change of rice planting areas,but the government policies,regulations and policy guidance were the dominant mechanisms that affect the change of rice planting and utilization structure. |
PDF Full Text Request