Near surface temperature inversions are periods in which a low layer of warm air is trapped between cooler air higher up in the atmosphere and dense cooler air below it near the surface level. By causing cooler air to pool near the surface level, inversions can have detrimental effects for crop growers, including frost, increased moisture, and pesticide drift. As a result, predicting the occurrence and magnitude of these inversions yields substantial benefits for growers. We introduce a Long Short-Term Memory (LSTM) model for temperature inversion forecasting that is able to effectively predict localized, near surface temperature inversions in advance such that growers can take actions to mitigate the detrimental effects. We show a substantial performance gain over a deployed temperature inversion forecasting system, and include a series of ablations that show the benefit of using publicly available terrain-specific feature information when modeling inversions at this scale.

Conventional methods for crop lodging assessments need accurate ground observations and tend to be laborious. Lodging assessment methods and accuracy can thus be improved using remote sensing data from small unmanned aerial systems (UASs) and low orbiting satellites (LOSs). With such aim, imagery to assess spearmint crop lodging was acquired using a small UAS at two ground sample distances (GSDs) of 0.01 and 0.03 m. Crop surface model (CSM) and six image color features were extracted from small UAS-based data. These features were then classified into not lodged (NL), partially lodged (PL), and lodged (L) groups. Mean and majority feature classes were obtained for 50 regions of interest (ROI) of size 1 m2each. Features were compared with visual crop lodging ratings using Pearson correlation (r) and Cohen's kappa (CK) coefficients. CSM showed higher assessment accuracy with r ≈ 0.85 and CK > 0.60. Mean percentage red (%R) was observed to have the strongest correlation with visual ratings (r = 0.75 and CK = 0.40-0.59) followed by mean percentage blue (%B), both at 0.01-m GSD. The percentage of lodging calculated from %R and %B maps was also contrasted with similar estimates from LOS-based imagery at 3.00-m GSD with no statistical differences found at 5% level.