Remote Sensing Method On Potato Planting Area And Progress Mapping In Southern China

The planting area information of crop is a prerequisite for crop monitoring and further study. There are two traditional approaches to obtain planting area and spatial distribution of cropland: agricultural statistics and land survey. It often takes time and resource to carry out agricultural census data, which are not accurate and available to the public in the same year. Space-borne remote sensing technology provides an alternative and independent approach for mapping cropland area.The recent research about remote sensing detection of cropland was always concentrated on staple crops, like wheat, rice, corn, soybean and so on. And research region is usually located in northern, eastern or northeastern China, or partly located in northwest and southwest. At present, there are few studies about monitoring potato planting area using HJ-1 A/B CCD data in our country especially in southern China, where is usually cloudy and rainy, and the parcels are very fragmental.Potato is a special kind of winter-planted crop in Guangdong Province, and relevant parcels which are in three rotations pattern(early rice – late rice – winter potato) have their unique annual time series curve of vegetation index(with three peaks and three troughs). Therefore, we can distinguish the target pixels from pixels of other land cover. Based on this, the paper puts forward a cropland extraction method based on time series data of vegetation index(VI) and Spectral Angle Mapping(SAM). Considering the cloudy and rainy weather in Southern China, it is difficult to obtain cloud-free images of all phenological time nodes. Hence, the paper explored a method to simplify time nodes of time series data stacks with the aim of utilizing less time nodes and still maintaining suitable extraction accuracy. Based on the extraction result and EVI time series data, the paper also monitored and mapped the cultivation progress. The paper takes Renping Peninsula in Huizhou City, Guangdong Province as research region and planting area and progress of winter potato as research object. The vegetation indexes used in this paper are NDVI(Normalized Difference Vegetation Index) and EVI(Enhanced Vegetation Index).The main methods and conclusions are as follows:(1) Time series curve of vegetation index(NDVI and EVI) reflecting the phenological rule and cultivation progressLayer stack the vegetation index(NDVI / EVI) images to construct a time series stack, and obtain the vegetation index time series curves of every training sample points from the time series stack. At each time node, we calculate the average values without the maximum and minimum. The combination of average values of all the time nodes is taken as a vegetation index(NDVI / EVI) time series reference curve representing phenology of research region. They reflect not only the “First Rice-Second Rice-Winter Potato” cropping patterns, but also the difference of potato planting progress.(2) A VI-SAM-based cropland extraction methodVegetation index time series curve of each training sample points is considered as a vector. After calculating the angles between a reference curve and training sample points’ curves respectively, we calculated the mean value and standard deviation of the training sample angle dataset. We took the sum of the mean value and standard deviation(or three time standard deviation) as threshold of rule image for extracting potato planting area.(3) A VI-SD-based time nodes simplification methodAt each time node, based on vegetation index time series curves of training sample points, we calculated vegetation index standard deviation values at each time node based on the vegetation index time series curves of training sample points. On the basis of the principal which is to reflect both the phenological character of potato planting area and the consistency of the cultivation progress(that is the choice of low vegetation index standard deviation values), the simplified time nodes are selected, including the 8th, 105 th, 144 th, 214 th, 261 th and 312 th. After simplifying the time series stack and using the cropland extraction method based on vegetation index time series data and SAM, the correct rate of the result has increased 11.54%-20.19% and the error rate has decreased 9.1%-18.2%.(4)The accuracy assessment of extraction resultsExtraction results include:(1)extraction result based on pre-simplified NDVI time series data and SAM;(2)extraction result based on pre-simplified EVI time series data and SAM;(3)extraction result based on simplified NDVI time series data and SAM;(4)extraction result based on simplified EVI time series data and SAM. Four accuracy assessments include: Ⅰcorrect rate assessment based on potato ground-truth sample points;Ⅱerror rate assessment based on non-potato ground-truth sample points;Ⅲrelative error assessment based on statistics data;Ⅳsupplementary assessment based on spatial distribution. No matter the result based on NDVI data or EVI data, those derived from simplified data are more accurate than those derived from pre-simplified data. And the cropland extraction method based on simplified EVI time series data and SAM is the most effective one for extracting winter potato planting area in the research region. The overall correct rate is 92.31%, error rate is 9.1%, and relative error rate is 9.47%.(5)A multi-EVI-based method for monitoring and mapping the cultivation progressThe result, extracted by the method based on simplified EVI time series data and SAM, is taken as region of interest, and the temporal EVI images are taken as data to be processed. After intersecting, the EVI data of the region of interest are obtained. Each pixel of them is symbolized in different colors according to their EVI value, and different colors correspond with vegetation growth stages. So the colored images can intuitively reflect cultivation progresses in different places from a spatio-temporal view, and are the result of monitoring and mapping the cultivation progress.