Southern Illinois University Carbondale Projects

At SIU my focus was on alternate farming practices for corn and soy production. I started out in the fields applying fertilizer and other treatments, taking soil samples, and analyzing them in the lab. When the research team realized that I could use GIS systems my main task shifted and I began making maps and contributing to papers.

Topography, Tillage, and Cover Crop Effects on Corn and Soybean Yield

Prior to this study, my GIS work generated either climate statistics or relatively simple maps of experiment sites. I was given full rein over this one, though; I wrote most of the material, performed most of the analysis, and made sure to include as much mapping of the project as I could, which wasn't very common for these treatment studies. It involved cover cropping and no-till practices as part of a broader examination into alternate farming practices in light of severe agricultural run-off flowing out of the mid-west. By no-till we meant a little tilling, actually, but just for the seeds. If we could show that yields respond positively to nitrogen-fixing cover crops, and that no-till practices are as productive as traditional tilling, an argument for less fertilizer use and runoff generating tillage practice could have been made. It was a randomized, split-plot treatment design with several types of cover crops that combined with the no-till treatments. It ran for two years: soy in 2014 and corn in 2015.

Harvest Yield Interpolation

After assigning harvest subplot values to each subplot centroid, I interpolated yields for the using kriging, because literature cited this as the most accurate for dense concentrations of sample points and ours had about 175 points per hectare. After trying several paramter combinations I settled on universal kriging with a linear drift semi-variogram model at 12 points per search radius. This is displayed with ten value ranges for interpretation. Though you could see it in the subplot yield graphic, it was when the surface was plotted that we first noticed the effects of topography.

Corn vs Soy Yields

The first year, 2014, was particularly dry with about 40 mm less rainfall than the average growing season. In contrast, 2015 was incredibly wet with 170 mm above the average growing season. You could say it was a good thing that we decided to map this rather than run the typical regressions of yields by treatments, though we would've had a much easier time of it without doing so. It was clear that water accumulated in the troughs lead to higher yields in the dry year and lower yields in the soggy wet year. This trough contained all no-till treatment plots. Without taking into account topographic effects, the conclusion from would have likely been that no-till is the best management practice in dry years and the worst for wet years.

Topographical Position Index

We ran an exploratory regression analysis in ArcGIS with a suite of soil characteristics to try and find some relationship with treatment. We also included categorical variables for topographical positions (summit, trough, slope) to try and control for topography, but with only two years of data, and completely inverse precipiation years, we were unable to mete out a solid treatment signal. We did see value in the study because of the topographical problem and its importance to precision agriculture, so we submitted to Agronomy Journal. However, after submission, I moved to Colorado, the PI moved to North Carolina, and we never found the time to tackle the revisions. We did come away with a decent lessen, though: consider topography when designing agricultural treatment blocks! Randomized treatment designs might only be appropriate for fully flat fields.