Home

Program Information

Research Groups

Related Information

CASMGS

 

Field Monitoring
    of Soil Carbon

Project Goal
The main task of our group is to develop general methodologies for monitoring soil carbon changes over time. The latter is of great importance for implementing verifiable C sequestration programs. We are also interested in how our measurements of soil C changes compare with those estimated by the net ecosystem exchange of CO2-C (Atmospheric CO2 Flux group) or those predicted by simulation models (Soil and Crop Modeling).

Project Description
Our group has two major goals: (i) to monitor annual changes in soil carbon stock at each of the three project sites and (ii) to develop general methodologies for mapping soil carbon and estimating its changes over time. In both, we focus on measuring soil carbon at the scale of a typical production field in the North American Midwest, i.e., fields that are about 40 to 60 ha in size. We believe, however, that many of the methodologies we develop at the field scale are also applicable to monitoring soil carbon at the regional scale.

Our principle challenge is that these methods must be cost-efficient and enable us to measure small changes in soil C against a large background of carbon that already exists in the upper soil layers (Fig. 1).

Figure 1. Typical amounts of soil carbon in a deep loess soil in relation to changes that may occur over short time periods of carbon sequestration in agricultural fields.

Progress
We have developed a general methodology for mapping of soil carbon that is based on the integrated use of secondary information and measured values of soil carbon (Fig. 2).

Figure 2. General methodology for mapping of soil carbon.

Our assumption is that numerous spatial information sources exist that provide indirect information about the spatial distribution of soil carbon and that this information is often available at fine spatial resolution and little cost. Our proposed methodology involves five major steps:

1. Acquire the available secondary information, evaluate its usefulness for C mapping, and process it to a common spatial resolutions (e.g., 5 m x 5 m or 10 m x 10 m pixels). Variables we find most useful for soil C mapping include soil series, elevation, slope, electrical conductivity (measured with a Veris-3000 system), and soil surface reflectance measured by remote sensing. The latter works best if crop residue coverage is small. In our studies, soil EC measured on-the-go with the Veris system had the highest correlations with measured soil organic carbon (r = 0.7 to 0.8).

2. Conduct a spatial classification to delineate homogeneous areas within a field (strata or patches). For this, we have developed a new cluster analysis algorithm that utilizes both ordinal and continuous attributes and results in spatially contiguous clusters. The user can also assign different weights to the categorical variable(s) used. In most cases, the latter is a ‘soil type’.

3. The stratification obtained from the cluster analysis is then used to optimize sampling. Specifically, we use a technique called ‘spatial simulated annealing’, in which sampling locations are distributed within a field to honor the proportional area of each classified patch as well as to optimize the estimation of semivariograms and achieve minimum distance among locations.

4. Sampling to measure soil C and bulk density in the top 0 to 30 cm of soil. In our research, we further split those samples in 0-5, 5-15, and 15-30 cm depths ranges. Multiplying soil organic carbon mass fraction with bulk density and depth results in an estimate of soil organic carbon stock for a certain soil volume. We usually express this as Mg C ha-1 (Mg = megagrams = 1 metric ton = 1000 kg) for a depth of 0 to 30 cm.

5. Use the sampled soil carbon stock values in combination with selected secondary variables to obtain a detailed map of soil organic carbon stock. The secondary variables we use must be correlated with soil carbon, but, unlike the few destructive carbon samples collected, the secondary information is exhaustive in a sense that it is available for a fine grid (see step 1). The technique we propose is called ‘Simple kriging with varying local means (SKLM)’ or ‘Regression kriging’. It involves modeling the relationship between soil carbon stock and the secondary variables by regression, using that regression to predict soil C for all pixels in a map, and modifying those predictions by kriging interpolation of the regression residuals obtained for the locations at which soil C was measured.

We have evaluated these procedures at all three sites (Fig. 3).

Figure 3. Experimental sites of the UNL Carbon Sequestration Program at Mead, Nebraska. Each field is a ¼ section of 800 m x 800 m size (64 ha).

Figure 4. Secondary variables used to improve mapping of soil organic carbon stock at site 1.

In general, the utilization of secondary information greatly improved the accuracy of soil carbon maps. With intensive sampling, relative improvements in map precision ranged from about 20 to 35% over a commonly used interpolation method such as ordinary kriging of soil carbon alone (Fig. 5). When sampling density was reduced to half the original density, map precision decreased significantly with ordinary kriging, but not with the proposed SKLM interpolation method (Fig. 5). In other words, high precision can be maintained with much reduced sampling cost, provided that secondary information is available and correlated with soil carbon.

Those results were very consistent across all three sites and they indicate substantial potential for precise soil C mapping. Maps are important for research studies and site-specific soil and crop management, which, ultimately, could also lead to greater crop yields and, therefore, increased rates of carbon sequestration. It remains to be seen whether similar methods can be used to improve carbon mapping in larger regions.

Our current work focuses on testing similar approaches for estimating the mean soil C stock in a field, which is of more interest to carbon sequestration programs than creating maps. In other words, the objective is to obtain a precise estimate of the overall (global) field mean and its precision by collecting an appropriate number of samples from the right locations. These project sample locations are called intensive management zones (IMZs).

Figure 5. Maps of soil organic carbon stock at site 1, based on intensive soil sampling (TOP row) or reduced sampling (BOTTOM row). The root mean square error (RMSE) is a measure of map precision and should be as small as possible. Relative improvement is calculated as the decrease or increase in RMSE compared to intensive sampling ordinary kriging (the map on the top left).

Sampling

Relevant Web Pages and Citations

Antle, J.M., S.M. Capalbo, S. Mooney, E.T. Elliott, and K. Paustian. 2003. Spatial heterogeneity, contract design, and the efficiency of carbon sequestraton policies for agriculture. Journal of Environmental Economics and Management 46:231-250. http://www.casmgs.colostate.edu/pubs/files/304_file.pdf

CASMGS subtask 4.1 http://www.casmgs.colostate.edu/insider/vigview.asp?action=2&titleid=448

Recommended procedures for soil C sampling http://www.casmgs.colostate.edu/vignette/448_8___recommended%20procedures.pdf

Staff
Achim Dobermann is professor of soil science and nutrient management at the University of Nebraska-Lincoln. From 1992 to 2000, he was a soil scientist at the International Rice Research Institute and led a multi-national research program on developing new concepts and tools for site-specific nutrient management in irrigated rice systems. Prof. Dobermann conducts research on nutrient cycling, soil variability, geospatial and crop modeling, soil greenhouse gas emissions, and approaches for site-specific nutrient management in major cereal production systems of Asia and North America. He has published two books and more than 60 papers in international scientific journals.

Michelle Haddix (Research Technologist; Soil and plant measurements)

Darren Binder (Research Technologist; Soil and plant measurements)

Arlene Adviento

Dan Walters


 Home   Program Info   Research Groups   Related Information   CASMGS 


 Atm CO2 Flux   Plant Carbon   Surface GHG Fluxes   Litter Decomposition   Monitoring 


 Soil Moisture   Remote Sensing   Energy Costs   Modeling   Adoption & Market