Using Drones to Measure Cover Crop Biomass as a Predictor of Soil Nitrogen And Corn Emergence Issues

Cover Crop Planting Rate Curves

Cover crops were planted at their recommended NRCS rates for rye (100 lbs/acre), clover (15 lbs/acre), and their rye & clover mixture (40 and 10 lbs/acre), as well as 0, 25, and 75% of the maximum rate. The biomass produced by the planting rates had a logarithmic best fit, with R2 from 0.88 to 0.97 (Figure 1). Although rates were planted at 25% of their maximum, they still produced significant biomass, enough that a linear relationship couldn’t be established. This is positive news for cover crop planting, as lower rates still perform well, but the variability at each treatment rate could be significant (Figure 1). Biomass for all three treatments averaged 2146 lbs/acre, reaching a maximum of 3995 lbs/acre (Table 1). The C:N ranged from 12 to 43, which is within expected values for rye, clover, and their mixtures. The average nitrogen (N) that could be contributed to corn production was 47 lbs/acre.

Both rye and mixed cover crops had higher biomass than control plots, but produced similar amounts between 25-100% of the recommended planting rate while no differences were observed across clover (Table 2). The lack of differences in clover is probably related to field variability in the plots, with wet spots and winter kill reducing plot consistency. The total N in the cover crop biomass (lbs N/acre) did not vary among planting rates, and was only higher than the control (weed) plots in rye and mix plots. Clover did have greater total N, but mixing with rye produced similar total N, but a  higher C:N ratio (12 vs 20). Soil nitrate and ammonium did not vary by any of the cover crop rates and is not shown here, besides overall averages in Table 1.

Stand counts were made to determine whether cover crop biomass reduced corn emergence and growth, however no differences were observed. This may be due to the fact that total biomass was also similar between rates, outside of the control plots.

Rate Curves and Drone Indexes

Drone indexes observed similar differences in biomass, with NDVI being greater for plots with cover crop in rye and clover plots (Table 2), but not mixes. Differences weren’t detected by NDRE or VARI, but the webVARI was able to differentiate between any rye rate and the control plots.

Since there was a lot of variability in cover crop survival and biomass per treatment, data was combined to see if drone indexes would correlate to all biomass samples (Table 3). Besides the Pix4D Phantom4 VARI index, the other indexes had significant positive correlations to all plot biomass, with WebVARI having the strongest correlation. When biomass was turned into total N (lb/acre) in the cover crops, all indexes were significantly correlated, with NDRE, NDVI, and WebVARI all above 0.8 (Table 3). This probably represents the measurement of plant productivity that imagery picks up, versus leaf area index and plant height that biomass represents. As more farmers will be interested in the N availability in cover crops, and not just biomass, this is a positive result.

Drone derived indexes also correlated to soil N and C measurements. As reflectance from cover crops makes up most of the drone imagery, this may be more of a sign of overall soil fertility. Soils with higher N or C may be more productive, increasing cover crop biomass, and therefore reflectance in each drone index. This was in opposition to cover crop tissue C and N, which did not correlate to drone indexes very well until they were converted to C:N, with each correlation being negative. So that as drone index values increase, C:N, and therefore mineralizable N availability, decreases.

Another example of how N content of the tissue increases drone index predictive power can be found in Figure 2. In this case the webVARI index is being compared to all biomass measurements in each cover crop treatment. While biomass has respectable R2 with a linear fit, the conversion of biomass to N content (lbs/acre) improved all relationships. This figure also displays the difference each cover crop type may have in curve fit and the algorithm produced to predict biomass in other fields, which may not be repeatable year to year.

Cover Crops Over the Winter Growing Season

Another useful aspect of using a drone is the ease and frequency at which data can be collected. The dataset in Figure 3 represents NDVI from fall planting until biomass harvest. Although treatment differences for rye were not observed at biomass harvest, there was consistent separation between rye treatments and the control until winter temperatures (and possible soil saturation) decreased the highest cover crop rate. Past February 2020, all rye plots appear to grow at the same rate, and the standard rate never pulled above the 25-75% rates. Clover, on the other hand, never had significant growth until late March, at least which NDVI could detect. This may be useful information to Delaware farmers, who could possibly cut their application rates 25%, but only represents one year of data.

Farmer Field Results

Most farmers and consultants will have access to more affordable consumer drones and web-based software, so predictive curves were compared using the Phantom4 and the VARI and webVARI data (Figure 4). For the research plots and the two rye (Georgetown, Harbeson) and one rye clover mix (Laurel), a different relationship was found in every field. Biomass from the research plots could not reliably predict biomass in any of the farmers’ fields, due to variability in cover crop planted, timing of sampling and varying field properties. However, flights on each farmer field still had a good relationship to biomass sampled (Figure 4), with R² ranging from 0.78 to 0.90. While one year of rate curve plots at our research station did not yield a reliable prediction, field variability was uncovered through drone images. Flights could be performed prior to burn-down to predict cover crop biomass, although it would still require image stitching and biomass sampling to provide a realistic baseline. Still, a drone (or satellite) map could provide sampling points, similar to zone soil sampling, that help predict potential benefits (e.g. nitrogen) to the following corn or soybean crop.

Additional analyses of correlation coefficients between drone derived indexes and cover crop biomass were also strong for all three fields. The webVARI index had correlations from 0.64 to 0.91, while the VARI index ranged from 0.89 to 0.95 (data not shown). Similar results were observed when comparing N (lbs/acre), with values of N prediction only dropping for Laurel vs biomass, possibly because it was a mix of rye and clover, with uneven clover growth.