Soil sampling is an important way to estimate regional soil properties statistical parameters and spatial variability analysis models. Sampling design as an important part of the sampling strategy, including the sampling method (such as random sampling or hierarchical sampling) and the number of samples, is a key factor in determining the sampling cost and estimation accuracy, so the sampling design has been studied by scholars at home and abroad. Hot spot. The general sampling design is based on the prior knowledge of known soil characteristics statistical parameters and spatial variability characteristics, but obtaining this prior knowledge is usually based on a certain number of sample data analysis, sampling scale and sampling space layout. It is the two key factors that determine the quality of the sample data. It also directly affects the sampling design results of the next step. So what sampling scale can meet this need? This is an important issue that needs to be solved based on soil sampling estimation regional soil characteristics parameters or spatial variability modeling analysis, but the current sampling number research focuses on the calculation of reasonable sample sizes based on classical statistics, or through comparative analysis of different sampling densities. The results of spatial variability estimation of soil characteristics determine the appropriate number of sampling or sampling intervals, and little is known about the effect of sampling scale on spatial variability analysis results of soil characteristics. Using high-density soil nutrient sampling data as the data source, many soil testing instruments such as soil nutrient fastness testers have a complete soil sampling system. Random sampling generates sample data sets at different sampling densities to study sampling density on soil. The estimation of nutrient statistical parameters, the characteristics of spatial variability and the effects of geostatistical interpolation analysis are intended to provide the basis for the analysis of the uncertainty of the analysis results by the sampling scale.
Soil sampling can currently be done in three ways:
1. Outlier sample detection:
Analyze and eliminate the outliers in the soil total nitrogen sample data set, and generate data sets that remove outliers as the original sample data. Most of the sampling outliers in soil samples are based on domain values, such as the sample mean plus and minus n times standard deviation method, the normal quantile map method, and the box line map method. The study uses an quartile method based on the cumulative frequency of sampling point data to set outlier thresholds, analyzes and removes outliers in the original sample data set, and generates 3 data sets with removed outliers.
2. Sample data set generation for different sampling scales:
The original sample data after eliminating outliers were used to calculate the optimal number of samples using the calculation method of reasonable sample size in classical statistics. Then the optimal number of samples was randomly selected based on the initial data set after removing outliers. A certain number of sample data constitutes 8 new sample data sets with different sampling scales. Among them, low multiple sample data sets are randomly selected from adjacent high multiple sample data sets.
3. Sample data analysis of different sampling scales:
Based on the above sample data sets, the regional statistical parameters, spatial variation characteristics, and geostatistical interpolation analysis accuracy of total nitrogen under different sampling scales were compared and analyzed, and the effect of sampling scale on spatial variation analysis of total nitrogen was examined. The above spatial data processing and spatial variability analysis were implemented in ArcGIS 9.2 software, and general statistical analysis of data was implemented in Excel 2003.
Through the above analysis and experiment you can draw conclusions:
(1) When the sampling scale is large and the autocorrelation between samples is weak, relatively few sample data can meet the needs of regional statistical estimates such as nutrient mean, coefficient of variation, and histogram distribution, but it cannot be used. For spatial variability and interpolation analysis, when the average sampling distance is greater than half of the correlation distance, and the number of samples is less than the best sample number, when the sampling points are relatively concentrated, the autocorrelation between samples is very strong, and the information redundancy is large. The data also cannot meet the needs of regional nutrient statistics, spatial variability, and interpolation analysis.
(2) With the increase of sampling scale, the global trend of spatial distribution of soil nutrients increases, but it does not affect the semivariogram model of soil nutrients; sampling scale has a great influence on the correlation distance, but the spatial distribution of samples has a greater impact on correlation distance than sampling. The scale itself is more pronounced.
(3) When the number of samples is larger than the optimal number of samples, the statistical characteristics of nutrients, spatial variability, and interpolation analysis increase with the decrease of sampling scale, while the sample size is less than 0.2, the sample data can satisfy the medium spatial variation of soil. The need for spatial interpolation analysis of nutrients; sample space layout has a greater impact on spatial prediction analysis than sampling scale.
Soil sampling can currently be done in three ways:
1. Outlier sample detection:
Analyze and eliminate the outliers in the soil total nitrogen sample data set, and generate data sets that remove outliers as the original sample data. Most of the sampling outliers in soil samples are based on domain values, such as the sample mean plus and minus n times standard deviation method, the normal quantile map method, and the box line map method. The study uses an quartile method based on the cumulative frequency of sampling point data to set outlier thresholds, analyzes and removes outliers in the original sample data set, and generates 3 data sets with removed outliers.
2. Sample data set generation for different sampling scales:
The original sample data after eliminating outliers were used to calculate the optimal number of samples using the calculation method of reasonable sample size in classical statistics. Then the optimal number of samples was randomly selected based on the initial data set after removing outliers. A certain number of sample data constitutes 8 new sample data sets with different sampling scales. Among them, low multiple sample data sets are randomly selected from adjacent high multiple sample data sets.
3. Sample data analysis of different sampling scales:
Based on the above sample data sets, the regional statistical parameters, spatial variation characteristics, and geostatistical interpolation analysis accuracy of total nitrogen under different sampling scales were compared and analyzed, and the effect of sampling scale on spatial variation analysis of total nitrogen was examined. The above spatial data processing and spatial variability analysis were implemented in ArcGIS 9.2 software, and general statistical analysis of data was implemented in Excel 2003.
Through the above analysis and experiment you can draw conclusions:
(1) When the sampling scale is large and the autocorrelation between samples is weak, relatively few sample data can meet the needs of regional statistical estimates such as nutrient mean, coefficient of variation, and histogram distribution, but it cannot be used. For spatial variability and interpolation analysis, when the average sampling distance is greater than half of the correlation distance, and the number of samples is less than the best sample number, when the sampling points are relatively concentrated, the autocorrelation between samples is very strong, and the information redundancy is large. The data also cannot meet the needs of regional nutrient statistics, spatial variability, and interpolation analysis.
(2) With the increase of sampling scale, the global trend of spatial distribution of soil nutrients increases, but it does not affect the semivariogram model of soil nutrients; sampling scale has a great influence on the correlation distance, but the spatial distribution of samples has a greater impact on correlation distance than sampling. The scale itself is more pronounced.
(3) When the number of samples is larger than the optimal number of samples, the statistical characteristics of nutrients, spatial variability, and interpolation analysis increase with the decrease of sampling scale, while the sample size is less than 0.2, the sample data can satisfy the medium spatial variation of soil. The need for spatial interpolation analysis of nutrients; sample space layout has a greater impact on spatial prediction analysis than sampling scale.
Spider Crane,Spider Crank Truck,5Section Spider Crank Truck,5.0Tons Spider Crane Truck
Hubei ruiyate Automobile Co.,Ltd , https://www.ruiyate-automobile.com