|CHANDEL, ABHILASH - Washington State University|
|KHOT, LAV - Washington State University|
Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/9/2021
Publication Date: 2/15/2021
Citation: Chandel, A., Khot, L., Yu, L. 2021. Alfalfa (Medicago sativa L.) crop vigor and yield characterization using high resolution aerial multispectral and thermal infrared imaging technique. Computers and Electronics in Agriculture. 182. Article 105999. https://doi.org/10.1016/j.compag.2021.105999.
Interpretive Summary: Alfalfa is an important forage crop grown worldwide for animal feed, green manure, and as a cover crop during crop rotation. However, very few rapid approaches exist for assessing alfalfa yield, forage quality, or response to stress. We conducted a study to assess the suitability of using small unmanned aerial systems (UAS), commonly known as “drones”, coupled with cameras to assess crop stress, vigor, and yield across two cutting cycles. Such characterizations may be useful for breeders and researchers to quantify forage quality and yield across different spatial scales including rows, plots, and entire fields. Such assessments may also help growers for precision management of crop inputs, harvest and storage facilities, crop insurance and cash budgeting.
Technical Abstract: Alfalfa (Medicago sativa L.) is an important forage crop grown worldwide for animal feed, green manure, and cover cropping during crop rotation. However, very few approaches exist for timely and field scale mapping of forage quality and yield, as well as for assisting precision management of crop inputs and harvest and storage resources. The objective of this study was to apply low altitude high-resolution aerial remote sensing to characterize above crop attributers. A small unmanned aerial system (UAS) integrated with multispectral and thermal infrared imaging sensors was used to image an alfalfa field at a spatial resolution of 7 cm/pixel. Two cutting cycles were imaged during the 2018 season. Eight crop vigor index (VI) and a Crop Water Stress Index (CWSI) features were derived from collected imagery data. Fresh crop yield data was collected as ground-reference by weighing a total of 2050 alfalfa plots during the harvest cycles. Modified Non-Linear Index (MNLI), Modified Simple Ratio (MSR) and CWSI were found to reliably evaluate spatial variations in crop vigor and stress traits (Coefficient of variation [CV] ranged from of 24–69%). Distinct yield prediction models were also developed from a randomized training dataset (70% of the plot samples) with indices as the predictor and yield as the response variable. These included nine simple linear regression (LRs), seven multiple linear regression (MLRs), a stepwise linear regression (SLR), a partial least square regression (PLSR), and a least absolute shrinkage and selection operator (LASSO) models. Performance of these models was evaluated by a test dataset of actual yield (30% of the plot samples). MLR-4 with MNLI and CWSI features performed the best (Root mean square error [RMSE] = 0.45, R2 = 0.64) and LR-5 with MNLI feature was the second-best model (RMSE = 0.51, R2 = 0.54). The complex regression models; SLR, PLSR and LASSO predicted yield with accuracy similar to the best MLR model (RMSE in the ranges of 0.45– 0.46, R2 in the ranges of 0.63–0.64). MNLI (canopy vigor) and CWSI (stress) were found to be key indices among the best yield prediction models for their non-saturation and non-linearity features. Overall, low altitude high-resolution aerial remote sensing in the visible-NIR and thermal infrared domain demonstrated potential to effectively monitor alfalfa crop condition and predict yield.