Study On Precision Management Zones Based On Multi-source Data In Field Scale

With the continuous improvement of grain production, the research, selling and application of agricultural chemical fertilizer is also constantly strengthened. At the same time, there are more and more problems appears, the excessive use of fertilizer is one of the most outstanding problems. Excessive fertilization can not only cause the waste of fertilizer resource, but also cause a lot of pollution to the agricultural ecological environment, resulting in the reduce of economic benefits in agricultural and even debasing the production quality of crops. Therefore, it is of great and practical significance to the development of agricultural ecology and agricultural economy. Accurate demarcation management is one of the most effective methods to improve the efficiency of fertilizer utilization. It means dividing the specific areas into small units with the same characteristics, and then carry out variable fertilization to the unit with different attributes. Most of the traditional accurate demarcation management is mainly based on the soil grid sampling in t he field, by analyzing the physical and chemical properties of soil effecting crop production or partition production data to divide the precision management zones. Soil grid sampling in the field needs to consume a large amount of manpower, material resources, financial resources. At the same time its timeliness is poor and it is difficult to large-scale promotion. Its practicality is weak as well. While applying remote sensing technology to classify precise management partitions has a greater advantage than field grid sampling. It does not need to consume a lot of time in the field and could reduce the cost of variable fertilization. Its timeliness is higher and is suitable for large-scale promotion test.This paper took the Red Star Farm in Bei'an City of Heilongjiang Province as the experimental area, according to the crop yield records of previous years, the soybean planting plots of significantly different crop growth were studied. Carry out field grid sampling on the bare soil plots in the early stage of soybean sowing. Then took soil sample and analyzed the physical and chemical properties of the soil. Take the physical and chemical properties of soil and remote sensing images as the data source for precision management on field partition. Firstly, too k the physical and chemical properties of soil samples as the data source, combined with ISODATA partitioning algorithm, divide precision management zones; Secondly, taking into account the large topography of plot, combining physical and chemical properties of soil and elevation, slope, aspect conduct precise management division with ISODATA partitioning algorithm;Then, use GF-1 remote sensing image of bare soil in single period and the growing season period to conduct the object-oriented segmentation respectively; Finaly, conduct object segmentation using multi temporal vegetation index image to divide precision management zones. The main contents and conclusions are as follows:(1) Using the physical and chemical properties of soil to divide precision managem ent zones have better effect. Based on the physical and chemical properties of soil, use ISODATA partitioning algorithm to divide precision management zones. The spatial variability of soil physical and chemical properties was basis to judge the spatial variability of precision management zones. After dividing precision management zones, the coefficient of variation of soil physical and chemical properties were between 2.3% to 70%. Relative to the entire land, soil physical and chemical properties were decreased in different degrees, and there was a certain spatial heterogeneity between the various units of precision management zones. Therefore, using the physical and chemical properties of soil to divide precision management zones has a certain effect.(2) Combined with the terrain data to divide precision management zones can improve the accuracy of partitions. The topography of the experimental area is larger. Different topography has a certain effect on the flow of fertilizer application. Therefore, in the case of a comprehensive consideration of the terrain data, use ISODATA partitioning algorithm to divide precision management zones again. The results showed that the spatial variability of soil physical and chemical properties also get dropped, and the precision management zones partition results combined with terrain data was slightly better as well.(3) Precision management zones based on multi-scale segmentation and remote sensing images of bare soil period has the best accuracy, more suitable for vari able rate fertilization. Conduct multi-scale segmentation with GF-1 remote sensing images of bare soil period, and evaluate the optimal segmentation scale. The results show ed that the variability of soil physical and chemical properties have decreased greatly, and the spatial heterogeneity of each precision management zones was promoted, better than the precision management zones based on the physical and chemical properties of soil.(4) Dividing precision management zones based on remote sensing images of bare soil has batter effect than that based on multi-temporal vegetation index. After comparative analyzed the multi-scale segmentation based on the remote sensing images of bare soil on May 1st, remote sensing images of crop growth period on July 16 th, and the multi-temporal vegetation index on April 1st, May 20 th, July 16 th and September 4th, the results showed that, remote sensing images of bare soil were the best period of partition, followed by crop growing remote sensing images partition and the multi-temporal vegetation index partition was not ideal.