Cotton Information Extraction And Management Zoning Based On RS And GIS At County Scale

Quickly and timely monitoring planting area, growth and spatial distribution of cottonusing remote sensing technology has extremely significance to timely understand the state ofcotton production, guide production and management. Precision management of cotton fieldis objective need to improve the management of cotton production and digital agriculturaldevelopment.This article choose Xiajin County in Shandong Province as the study area and selectedremote sensing images of HJ1B satellite in five different cotton growing seasons in2009and2011. The best phase to identify cotton was determined by analyzing phenological calendarand spectral characteristics of major crops. In this study, the cotton planting area informationwas extracted by decision tree extraction model established combined with the spectralcharacteristic values and vegetation index. Growth monitoring model was established toanalyze the dynamic changes of cotton growth.Combining spectral characteristic indexestablished at flower and boll stage with five kinds of main soil nutrients measured includingorganic matter, total nitrogen, available nitrogen, phosphorus and potassium, cotton fieldmanagement zones were divided by correlation analysis, principal component analysisand K-means cluster analysis. The main conclusions of the study were as follows:(1) To determine the best phase to identify cotton planting area information, combinedwith spectral characteristics analysis of remote sensing images and phenology calendaranalysis of major crops in Xiajin, the results showed that: the best phase to extract cottonplanting area is bud stage from late June to early July in Xiajin.(2) This paper selected HJ1B satellite remote sensing images in July1,2009and June28,2011, evaluated the separability of the training area chosen after image pre-processing,including radiation correction, geometric accurate correction and masking. The decision tree classification model was established according to statistical spectral characteristic value andthe normalized difference vegetation index extracted of each class to classify the remotesensing images.The results showed that: The established classification model was quite accurate.Cotton space classification accuracy reached95%. Cotton planting areas extraction precision was93.81%and92.86%in2009and2011respectively. The land use type overall classification accuracy was93%.Kappa coefficient was0.91. Cotton acreage in2009was39700hm~2, reduced5900hm2in2011than2009,a decrease of13.8%.(3) Using HJ1B satellite remote sensing images after preprocessed in July1, August7,August30and September20, this paper calculate optimize the soil adjusted vegetation index(OSAVI) in cotton growing areas. The results showed that: Optimized soil adjusted vegetationindex (OSAVI) can well reflect the cotton growth. OSAVI shows regular changes in cottongrowing process. The average was0.618in July1, up to0.834in August7,0.792in August30, dropped to0.579in September20. From early July and early August, the cotton in astrong growing phase. Growth tends to be stable after entering the flower and boll stage.Cotton lied in bud stage in early July, when a portion of cotton seedling state was worse grewslow. The proportion of growth poor in prophase was higher than late stage. Cotton growthrate slowed down after mid-August.Cotton lied in the middle of boll opening stage in mid-lateSeptember, which originally growth well entered into boll opening stage earlier, but OSAVIvalue decreased, the proportion of growth well reduced.(4) This paper used ordinary kriging interpolation make GPS sampling points Vector databecame grid soil nutrients spatial distribution map, analyzed correlation between optimizedsoil adjusted vegetation index (OSAVI) reflect the cotton growth vigor information and soilnutrient, principal component analysis method, obtained the management district indicatorsby principal component analysis method, used K-means clustering algorithm and attemptedrepeatedly fast clustering. A reasonable number of districts were determined by F statistic andANOVA analysis results.The cotton field in the study area was divided into three managementzones. The results showed that, the OSAVI was significantly correlated with soil nutrientsdata. The first management zone accounted for24.67%of total area with the highest contentof soil organic matter, nitrogen, phosphorus and potassium, cotton growth well. The secondmanagement zone accounted for47.02%with medium soil nutrients and general cotton growth. The third management zone accounted for28.31%with the lowest soil fertility andpoor cotton growth. The entire study area after partitioned, in each management partition,coefficient variation of soil nutrients decreased significantly. There were significantlystatistical differences between mean soil nutrients data and spectral data in different definedmanagement zones, which showed the partition’s reasonable. It’s suitable to take samecultivation and management measures in the same management zone.This study provides accurate and convenient method for cotton information extraction atsmall-scale in Shandong Province using remote sensing, provide scientific basis of real-timeand accurate cotton cultivation management for decision-making, has great significance toscientific guide cotton cultivation. Combined with soil nutrients and remote sensing data forcotton field management district is an important means for the precision agriculture in futureand sustainable agricultural development and be of great significance for realizing China’sdigital agricultural strategy.