Use of Principal Components for Scaling Up Topographic Models to Map Soil Redistribution and Soil Organic Carbon

This method can help answer key questions in the agricultural field, such as how landscape topography can affect soil erosion and soil organic matter dynamics. The main advantage of this technique is that it is applicable to sites with limited observations, and it provides a cost-effective estimation of soil organic carbon stocks and soil redistribution processes. First, collect data from the GeoTREE Light Detection and Ranging Mapping Project website.

Select boundary type and region to zoom in to a specific area. Then, draw a polygon to download light detection and ranging tiles for the selected study area. Convert the raw light detection and ranging data to a LAS file using the geographic information system mapping tool.

Next, generate digital elevation models, or DEMS, with a three-meter spacial resolution using inverse distance-weighted interpolation. Filter the three-meter DEMS twice with a three-kernel low-pass filter to reduce noises associated with local variation. To generate topographic metrics, first click Import Raster in the Import/Export section to import the filtered three-meter DEMS into SAGA.

Next, click the Slope, Aspect, Curvature module of SAGA with the default settings to generate the slope, the curvature-related metric, and the general curvature metric using the filtered DEMs. Click the Flow Accumulation Top-Down module of SAGA, and select Deterministic Infinity as the method to generate the flow accumulation metric using the filtered DEMs. Following this, click the SAGA Topographic Openness module with the default settings to generate the positive openness metric using a filtered z-axis amplified image.

The enlargement of the vertical distances in the digital elevation models improves the distinguishability of positive openness at sites with a relatively flat surface. Click the LS-Factor Field Based module of SAGA with the default settings to generate the upslope slope and slope length factor metrics using the filtered DEMs. Next, click the Flow Path Length module of SAGA with the default settings to generate the flow path length metric using the filtered DEMs.

Click the Downslope Distance Gradient module of SAGA with the default settings to generate the downslope index metric using the filtered DEMs. Now, click the SAGA Wetness Index module, and select absolute catchment area as the type of area to generate the catchment area and topographic wetness index metrics using the filtered DEMs. Click the Stream Power Index module of SAGA, and select pseudo specific catchment area as the area conversion to generate the stream power index metric using the filtered DEMs.

Following this, generate maximum elevation maps with multiple radiuses. Filter the maximum elevation maps twice through a three-kernel low-pass filter. Subtract the filtered three-meter DEM from the filtered maximum elevation maps to obtain a series of relief maps.

Extract a series of relief variables to a number of locations. Perform principal component analysis on the relief variables to convert the reliefs into topographic relief components. Select principal components that explain more than 90%variance of the relief dataset as the topographic relief metrics.

Standardize the seven relief maps using mean and standard deviation. Create relief principal components by sum of the standardized topographic relief weighted by the corresponding loadings. While creating the relief metric, it's important to generate relief images at various spatial scale to limit uncertainties associated with the arbitrary selection of radius because controls of relief on soil properties could be influenced by spatial scales of relief.

Select a number of cropland field locations that can adequately represent the landscape characteristics of the study area and several representative small-scale cropland fields that can be intensively sampled. Upload all the sample location coordinates to a code-based geographic positioning system, and physically locate them in the field. Next, collect three samples for each sampling location from the top 30-centimeter soil layer using a push probe.

Record geographic coordinate information of the sampling locations using the geographic positioning system. Following this, sieve the soil sample with a two-millimeter screen. Weigh the soil samples after drying.

Calculate the soil density using the total sample volumes at sampling locations and weights. Mix the three samples from the same location to get a composite soil sample. Grind a 10-gram subsample of the sieved soil to a very fine powder with a roller mill.

Now, measure the soil total carbon content in the roller-milled sample through combustion on a CN elemental analyzer at a temperature of 1350 degrees Celsius. After baking the soil organic matter in a furnace, estimate calcium carbonate carbon content by analyzing the remaining carbon. Now, place the bulk two-millimeter sieved soil samples in Marinelli beakers, and seal them.

Place the beaker in the detector, and measure the cesium concentration of each sample through gamma-ray analysis using a spectroscopy system that receives inputs from three high-purity coaxial germanium crystals into 8, 192-channel analyzers. Record the cesium concentration output. Finally, calculate the soil redistribution rate using cesium inventory by applying the Mass Balance Model II in a spreadsheet add-in program.

460 crop field locations were randomly selected to derive topographic information in the Walnut Creek Watershed in Iowa. Results of correlation analyses between topographic metrics and soil organic carbon density, soil redistribution are presented here. The topographic wetness index and large-scale topographic relief showed the highest correlations with density and soil redistribution rates, respectively.

Spatial patterns of the two metrics showed high values in depressional area and low values in sloping and ridge areas. However, differences between the two metrics occurred in ditch areas, where the topographic wetness index exhibited extremely high values, but the values of large-scale topographic relief were not different from adjacent areas. Five topographic principal components that were selected to build topography-based models are listed here.

Over 70 and 65%of variability in soil organic carbon density and soil redistribution rates were explained by the stepwise ordinary least square regression model with whole variables, respectively. For the models with collinear covariate removed, simulation efficiencies were slightly lower than the stepwise ordinary least square regression model with whole variables model. For SPCR models, similar simulation efficiencies as the stepwise ordinary least square regression model with collinear covariate removed are observed.

The soil redistribution and soil organic carbon density maps generated from SPCR models revealed consistent patterns between model simulations and field measurements. This technique paves the way for researchers in the field of agriculture to explore soil redistribution and organic matter patterns at watershed and regional scales. The technique could be improved with further refinement of light detection and ranging data and inclusion of additional topographic metrics.