A novel approach to mapping corn chlorophyll content using a spectral-based airborne sensor

Authors: TIAN, FENGKAI - University Of Missouri; ZHOU, JIANFENG - University Of Missouri; Ransom, Curtis

Submitted to: ASABE Annual International Meeting
Publication Type: Proceedings
Publication Acceptance Date: 7/28/2024
Publication Date: 7/28/2024
Citation: Tian, F., Zhou, J., Ransom, C.J. 2024. A novel approach to mapping corn chlorophyll content using a spectral-based airborne sensor. Proceedings of the 2024 American Society of Agricultural and Biological Engineers (ASABE) Annual International Meeeting, July 28-31, 2024, Anaheim, California. Paper No. 2400094. https://doi.org/10.13031/aim.202400094
DOI: https://doi.org/10.13031/aim.202400094

Interpretive Summary: Corn is a commonly grown staple crop worldwide, and its health status directly determines food productivity. One key aspect of corn plants’ health status is chlorophyll, which helps the plant convert sunlight energy through the photosynthesis process. The amount of chlorophyll in corn plants is often related to how well the corn is growing and how much fertilizer it needs. However, determining the corn chlorophyll content is labor intensive and requires either sending tissue samples to the laboratory for analysis or taking readings from corn leaves. Our goal was to test alternative ways to measure corn chlorophyll. We utilized drones with special cameras that capture visible and near-infrared light reflected from the corn canopy during different growth stages. We used this data to accurately estimate the amount of chlorophyll in the corn. Our method was very similar to taking measurements directly from the plant leaves in terms of accuracy, being able to represent over 87% of the variability in corn chlorophyll content. This was made possible by using machine learning that considers several factors when extracting information from the images. Our discovery can help farmers by providing a nondestructive, accurate, and fast way of measuring corn chlorophyll. This could mean a better way of determining how much fertilizer to apply and when. It would mean healthier corn and would contribute to the better sustainable resource management.

Technical Abstract: Chlorophyll is a critical component for corn (Zea mays L.) as it plays a key role in photosynthesis, which directly affects plant growth and yield. Rapid, non-destructive estimation of chlorophyll concentration is needed to more accurately manage corn nitrogen needs. Our objective in this study is to use airborne multispectral imaging to estimate Soil Plant Analysis Development (SPAD) chlorophyll values in corn across multiple vegetative stages. In this study, 12 different nitrogen treatments ranging from 0 to 255 lb/acre were broadcast at the V4 corn growth stage in three replicates. UAV-based multispectral cameras equipped with five spectral bands (red, blue, green, NIR, red-edge) collected data at different corn growth stages (V8, V9, V11, V12). Chlorophyll content was estimated using SPAD meter readings. To account for soil variability, bulk soil electrical conductivity was included as an independent environmental variable. Efforts to relate image-based indices to SPAD readings showed that a sequential cross-validated forward selection combined with epsilon support vector regression outperformed other machine learning models such as PLSR, Elastic-Net, and Random Forest, achieving an R² of 0.87, with MAE and RMSE values of 1.80 and 2.26, respectively. The more important features used by the machine learning models included summary statistics other than the mean value. The study highlighted the importance of using diverse summary statistics beyond mean values to improve modeling performance—especially when extracting features from an image. The refined models have potential applications in mapping corn leaf SPAD readings across vegetation stages to help growers make timely management decisions prior to the reproductive stage.