In-season Corn Nitrogen Status Diagnosis and Precision Management with Proximal and Remote Sensing

Abstract

Nitrogen (N) is one of the most intensively used resources in corn (Zea Mays L.) production worldwide, however, it is often the most limiting nutrient for plant growth. This discrepancy is largely due to low N use efficiency (NUE) caused by poor application timing and inadequate estimation of native soil nutrient supply. Precision N management (PNM) seeks to supplement available soil nutrient supply using timely seasonal applications to match crop N demand spatially and temporally. The development of strategies to improve N fertilizer management and enhance in-season crop N diagnosis is essential to mitigate over-application which can result in loss in grower economics and environmental degradation. The objectives of this research were to 1) evaluate the potential of an innovative Crop Circle Phenom multiparameter proximal sensing system to directly estimate in-season N plant status metrics and 2) develop a practical remote sensing and calibration strip-based PNM strategy for on-farm applications. To evaluate the newly available Crop Circle Phenom multiparameter sensor system was tested in Wells, Minnesota during the 2018 and 2019 growing seasons on a split-split plot N experiment composed of 144 plots planted with corn. The treatments consisted of drained and undrained main blocks, three tillage treatments representing conventional tillage, strip-tillage, and no-tillage as the split plot, and pre-plant N (PPN) urea (46-0-0, N-P-K) with N-(n-butyl) thiophosphoric triamide inhibitor ranging from 0 to 225 kg ha-1 in 45 kg increments as the split-split plot. Plant samples were collected at V8 growth stage to determine aboveground biomass (AGB), plant N concentration (PNC), plant N uptake (PNU), and N nutrition index (NNI). Regression analysis for the multiparameter Crop Circle Phenom measurements alongside N rates was conducted integrating the tillage and drainage variables. Both simple and eXtreme Gradient Boosting (XGB) machine learning regression models were evaluated and compared for accuracy of N status diagnosis. The results indicated that simple regression modelling using normalized difference vegetation index (NDVI) would be sufficient for estimating AGB compared to more complex machine learning method. Conversely, PNC, PNU, and NNI all benefited from XGB modeling based on multiple inputs. Among different approaches of XGB modeling, combining management information and Crop Circle Phenom measurements together increased model performance for predicting each of the four plant N metrics compared with solely using sensing data. The PPN rate was the most important management variable for all models compared to drainage and tillage information. To develop a practical remote sensing and calibration strip-based PNM strategy for on-farm application, three on-farm experiments located in central and western Minnesota were conducted during the 2019 and 2020 growing seasons. Each on-farm trial was composed of five replicated PPN rates which stretched the length of the field and were implemented at pre-plant using granular urea fertilizer. The selected urea rates were 0%, 35%, 70%, 100%, and 130% of the traditional farmer’s N rate. These N strips were used to predict site-specific optimal N rates before side-dress fertilizer application. In-season N variability was measured with unmanned aerial vehicle (UAV), airplane, and satellite platforms at approximately the V6 to V8 corn growth stage. The NDVI data was calculated from images obtained using AeroVironment Quantix Mapper UAV, Ceres Imaging airplane, and Planet Labs PlanetScope satellite platforms. Using the NDVI response curves generated from the imagery measurements, site-specific optimal N rates were estimated for each grid by selecting the best performing PPN rate. The results indicated that all three remote sensing platforms were able to predict the site-specific N rates, which approximately coincided with the economic or agronomic optimal N rates. Across the three on-farm trials, 65-90% N rates determined using any two remote sensing platforms were within 10 kg ha-1 difference. Furthermore, the results showed an overall increase in partial factor productivity (PFP) from all three experimental trials, which was attributed to split-application management. The calibration strip-based PNM strategy is a practical and viable strategy for improving NUE of corn. However, further evaluation is needed to improve the proposed calibration strip-based PNM strategy under diverse on-farm conditions and in climates outside of Minnesota.