| Land resources are important natural resources for the subsistence and development of the mankind. "Protection of arable land is to protect our lifeline", the stock and quality of cultivated land can determine the degree of national security. Compare with the number reduction of cultivated land, land quality decline of is hidden, but it’s as serious as the impact of number decline. The decline of farmland quality can serious influence our country’s social economy development and ecological environment constitution. The classifications of agricultural land are the important way to protect our country farming resources quality security and help to sustainable use farmland.In recent years, land evaluation developed mainly in quantifying, precise and informationization. Land classification as an important part of land evaluation developed1 rapidly in recent years. There are grading theory frame and the method system initially of land classification and the application of land classification was also clear. From now on, the research of land classification will turn to choose specific parameter, and enhance the precision of result. Soil physics and chemistry is main impact factor of land nature quality. The topsoil’s organic carbon (SOC) content is one of important soil attributes, and it’s also the necessary factor of agricultural land classification. Therefore, in order to save the cost of land classification, it should be improve the accuracy of sampling method through efficient sample point layout and optimization of spatial interpolation methods. The efficient sampling methods are also the important aspect of intensive study in land classification.Scale is the inherent characteristics of nature. Land evaluation also has the scale-effect problem similarly with spatial position, and the soil attributes are the most important indicator of land quality evaluating. Scale effect of soil attribute in land classification is also critical. Present study of optimal sampling method in land evaluation mainly in county level or field level, and lacks locally administered level. There are none research about optimal sampling method and optimized space interpolation model in different scales.Therefore, study SOC spatial variation in different scales; research optimization sampling methods and spatial prediction method have theoretical and practical value. They can help to enhance the efficiency and accuracy of land classification.This study chooses three different size scales in three typical areas which is Zhangzhou (Prefecture-level city)、Longhai (County-level city) and Chengxi Town (Township-level), and use grid based mode to set investigation points. Grid based mode is considered a more accurately method to characterize spatial variability of soil property, we design four different grid densities and base on consideration all of the topography、land use patterns and soil types to set investigation points, there are 1743 investigation points and 473 verification points in prefecture-level, and 1133 investigation points and 259 verification points in county-level, and 140 points in township-level.This study is divided into two parts. Firstly, on the basis of variation coefficient which is an important index to characterize soil organic carbon (SOC) variability and under every grid density we research the influences of 5 classifications (unclassified grid based mode、 topography type based mode、land use based mode、soil type based mode、topography type pattern-soil type based mode and land use pattern-soil type based mode) on characterize soil organic carbon variability, then find which density can characterize soil organic carbon variability efficiently.Secondly, upon the Land Statistics Method which is believed to be one of the best and also the most frequently used spatial prediction method on soil property, we design 6 methods under different scales, they are KDM、KDL、KTR、KDMTR、KDLTR and KDLTR. Then under every size scale and every density, we predict in space and carry out comparative analysis of predicted value and measured value of verification points. Through the research of the spatial predict effects of all kinds of spatial interpolation method within a grid and the behavior of a spatial interpolation method among different grid density, we can draw the optimized spatial interpolation mode and the required sampling densities. Results are as follows:There are 62 times differences between maximum value and minimum value of SOC content in Zhangzhou (Prefecture-level), it shows that the variation of SOC content in southern hilly soil areas is large, and the variation coefficient is 49.34%. After classification, the characterization of SOC spatial variability has better effects than unclassified. In addition, the topography type pattern-soil type based mode is the best, because it can reduce the variation coefficient of SOC (nearly 20%).The difference between maximum value and minimum value of SOC content in Longhai (County-level) is nearly 23 times, and the variation coefficient is 43.35%, so the variation of SOC content is also large. Compare topography type based mode and land use based mode with unclassified grid based mode, the variations of SOC only reduce 1% and 3%, respectively, so it is not good enough to characterize SOC spatial variability when only use land use based mode. In soil type based mode, when classify to soil genus, the SOC spatial variability is reduced by 14%. Combine the land use based mode and soil type based mode, then the variation coefficient of land use based mode reduce by 11%, and nearly by 12% for soil type based mode. It is indicated that the more effective way to assess the spatial distribution of SOC is to allocate sampling points based on land use pattern-soil type based mode.The minimum value and the maximum value differ more than 7 times for SOC content in Chengxi (Township-level), so even if to township-level, there are still big variations of SOC content in southern hilly soil areas. Results shows:"When study scale down to township-level, compare land use based mode and topography type based mode with unclassified grid based mode, the SOC spatial variability increase by 2.5% and 0.5% respectively. It is similar to soil type based mode, topography type pattern-soil type based mode and land use pattern-soil type based mode. Along the reduction of study scale, there is no discrepancy between unclassified grid based mode and classified grid based mode, however, the variation coefficient is lower when set sampling point directly by grid, so it is better to choose the latter.(4) Under different scales every point of sample point density influenced by the way quasi-geoid concerned. The not classification sample point of Zhangzhou’s (prefecture-level city scale) soil organic carbon variations increases with the increase of grid. Locally administered level scales. Locally administratered level according to landform type scales with soil type classification method of sample points by quasi-geoid coefficient of variation of the soil organic carbon minimum, at the moment, when samples laid out a grid density is equal to or greater than the 6km × 6km, for the spatial variability of soil organic carbon content of the characterization of the same effect, but when the sample density of less than or equal 8km× 8km,the coefficient of variation increased significantly, indicate that under the density that sample point layout can not very well characterization soil organic carbon space variations.Longhai city (county scale) unclassified sample point grid in the coefficient of variation of density difference is less than 5%, among them 2 km x 2km and 4km×4km grid is slightly less than the coefficient of variation scale level cities. County by soil type classification scale characterization of soil organic carbon under the coefficient of variation of the smallest. In the 0.5km x 0.5km and 1km × 1km grid size variation was similar, the coefficient of variation began to increase when grid size increases to 2 km × 2km. To state clearly, Scales at the county level shall sample the soil type layout, but cannot use the current use of land evaluation in the common situation for sample layout. Meanwhile, if soil types are only divided into great soil group, the sample density required to achieve more than lkm x lkm.if soil types are divided into subclass and above soil genus, the sample density can be relaxed to 2km x 2km. If the sample density of 4km x 4km and below, the characterization of soil organic carbon content of the spatial variability of large display samples for characterization of soil organic carbon content greatly increased the uncertainty, the cell density appropriate to adopt the following research sample density.ChengXi town (township-level scale) of the situation and prefecture-level cities and county scale is completely different; sample the size of the grid layout of the township-level scale has little effect. Classify samples but increased the coefficient of variation after emplacement, which also indicates that with the reduced scale of the study area is not classified grid method is the most suitable method for sample layout.(5) In the hilly area south of prefecture-level city scale, the direct application of ordinary Kriging Soil organic carbon in the spatial interpolation, the prediction accuracy is not high, while their spatial distribution to predict the spatial interpolation method used sampling sites because of their density differences. In this scale, soil type and topography relative to the type of information in terms of land use on soil organic carbon content greater impact on the space predictions. When the sample density is more than or equal to 6 x 6km of the grid, in order to improve the prediction accuracy of spatial interpolation should be used with landscape-soil type information Kriging method of interpolation. Meanwhile, the spatial prediction accuracy as the sample density increases. Sample point on the scale layout density can not be less than or equal 8 x 8km of the grid density, if subject to conditions, only the density of the grid layout like the following point, the direct use of samples from the organic carbon content of the data predicted higher accuracy compared to other methods.(6) The research area of county scale is much smaller than cities scale, but the hilly area in southern China, topsoil organic carbon content is still existence a strong spatial variability. The study found that the agricultural land classification and grading at the county level according to the current land use when grid method combining layouts sample point and not in good characterization of soil organic carbon space mutation, at the same time different sample point density directly affect the choice of spatial interpolation method. When the study demand for moderate accuracy of the space prediction results,can press 2 × 2km grid density quasi-geoid investigation sample points and when sample points only consider the differences of soil types laid, When making spatial prediction Should adopt the soil is level with the soil type information for ordinary Kriging (KTR) interpolation; When the researchers asked for the results of high precision spatial prediction, it need by the grid density of 0.5 × 0.5km and above to Layout samples, and land use by combining type and soil type information, ordinary Kriging (KDLTR) spatial interpolation; the grid density is less than 4×4km, because too few sample points, in the county scale does not accurately reflect the topsoil organic carbon content spatial distribution, so that the land evaluation projects carried out in the county when the sample density can not be less than the grid density. If there are conditions, the grid density should be directly used ordinary Kriging (KYJZ) for spatial prediction. The results also show that the hilly area in the south county scale, the soil type and topography than the type of land use on topsoil organic carbon content of the spatial distribution of large. Any combination of soil type information related to Kriging method (KTR, KDLTR, KDMTR) spatial interpolation of the results obtained were significantly better than the combination of soil type information is not the method (KYJZ, KDL, KDM). Therefore, in the county scales research the spatial distribution of soil organic carbon must consider the influence of soil types. |
PDF Full Text Request