By the end of the lab work you should have an estimate of soil organic carbon (SOC) stock as kg per m² to a depth of 400kg mineral soil (using the cumulative mass approach) or to a fixed depth (using the cumulative depth approach) for each field sample location. The aims of data analysis are then to:

1. Turn this raw data into the information products you set out to generate.
2. Spot any other patterns that could be important.

The statistical methods needed depend mainly on the products you are aiming for.  The methods to be used determine:

• The software needed: Estimates of total landscape SOC stock or change in total can be done in spreadsheets. Mapping of results and estimation of effects of interventions require blending SOC data with remote sensing imagery and use of statistical and spatial analysis software ﹘ we recommend and provide guides for R.
• The expertise needed: Estimates of total landscape SOC stock or change in total can be done by anyone with a feel for numbers and competence in using Excel.
• Mapping of results and estimation of effects of interventions will require input from someone with suitable statistical skills, as judgments have to be made.

 

Principles of data analysis

Four principles of data analysis should be followed in all work. These are not specific to estimation of SOC, but your data analysis will not go well if they are ignored.

1. Focus on meeting the output objectives, but be alert for unanticipated results that might be important.

2. Every estimated quantity that you produce should have a measure of precision (e.g. standard error or confidence interval) attached to it.

3. Methods used should be repeatable and objective and not based on arbitrary decisions. However, only the simplest methods can be fully automated.

4. Keep records of all the data and code used along with details of analysis decisions, so that analysis can be repeated or updated at any time.

 

Steps in data analysis

1. Prepare data files in formats that will make analysis efficient. The standard formats assumed by all code and examples are described here.

2. Run error detection procedures for all data. These will include confirming: that the numbers of observations are as expected, that the ranges of all variables are realistic and that relations between variables are realistic. Details and example are provided here. To estimate total SOC stock Calculate SOC stock for each plot HERE and for a project area assuming simple random sampling and two-stages sampling HERE 

3. Calculate the average SOC for each stratum multiply by stratum area and sum for the project area (Eq. 1,4).

4. Calculate the standard error of the total SOC using the method appropriate to the sampling scheme used. To estimate change in total SOC stock (Eq. 3, 5).

5. Estimate the change in total C per stratum, using the method appropriate to the first and second phase sampling scheme you used, and total for the project area.

6. Calculate the standard error of the change in total using the method appropriate to the sampling schemes used. 

 

To map current SOC stocks.

7. Mapping is done using a combination of two statistical ideas:

a. Exploiting the fact that SOC is typically spatially correlated, with locations close to each other having more similar SOC than points far apart. This allows us to interpolate SOC for points not measured.

b. Exploiting the fact that SOC is typically related to quantities measured in remotely sensed images. This allows us to use regression to relate SOC at measured points to data from images and then use image data to predict SOC for points not measured.

8. Methods become more complex when modeling change with depth simultaneously with change in 2D position. To map change in C stocks.

9. Mapping changes is done using similar methods as mapping SOC at one time. 

Check on the formulas