Search

Satellite remote sensing offers a cost-effective means of generating long-term hindcasts of yield that can be used to understand how yield varies in time and space. This study investigated the use of remotely sensed phenology, climate data and machine learning for estimating yield at a resolution suitable for optimising crop management in fields. We used spatially weighted growth curve estimation to identify the timing of phenological events from sequences of Landsat NDVI and derive phenological and seasonal climate metrics. Using data from a 17,000 ha study area, we investigated the relationships between the metrics and yield over 17 years from 2003 to 2019. We compared six statistical and machine learning models for estimating yield: multiple linear regression, mixed effects models, generalised additive models, random forests, support vector regression using radial basis functions and deep learning neural networks. We used a 50-50 train-test split on paddock-years where 50% of paddock-year combinations were randomly selected and used to train each model and the remaining 50% of paddock-years were used to assess the model accuracy. Using only phenological metrics, accuracy was highest using a linear mixed model with a random effect that allowed the relationship between integrated NDVI and yield to vary by year (R 2 = 0.67, MAE = 0.25 t ha−1, RMSE = 0.33 t ha−1, NRMSE = 0.25). We quantified the improvements in accuracy when seasonal climate metrics were also used as predictors. We identified two optimal models using the combined phenological and seasonal climate metrics: support vector regression and deep learning models ( R 2 = 0.68, MAE = 0.25 t ha−1, RMSE = 0.32 t ha−1, NRMSE = 0.25). While the linear mixed model using only phenological metrics performed similarly to the nonlinear models that are also seasonal climate metrics, the nonlinear models can be more easily generalised to estimate yield in years for which training data are unavailable. We conclude that long-term hindcasts of wheat yield in fields, at 30 m spatial resolution, can be produced using remotely sensed phenology from Landsat NDVI, climate data and machine learning. [ABSTRACT FROM AUTHOR]

Seasonal climate is the main driver of crop growth and yield in broadacre grain cropping systems. With a 40-year record of 30 m resolution images and 16-day revisits, the Landsat satellite series is ideal for producing long-term records of remotely sensed phenology to build understanding of how climate affects crop growth. However, the time-series of Landsat images exhibits gaps caused by cloud cover, which is common in wet periods when crops reach maximum growth. We propose a novel spatial–temporal approach to gap-filling that avoids data fusion. Crop growth curve estimation is used to perform temporal smoothing and incorporation of spatial weights allows spatial smoothing. We tested our approach using Landsat NDVI data acquired for an 8000 ha study area in Western Australia using a train/test approach where 157 available Landsat-7 images between 2013 and 2019 were used to train the model, and 95 at least 80% cloud-free Landsat-8 images from the same period were used to test its performance. We found that compared to nonspatial estimation, use of spatial weights in growth curve estimation improved correlation between observed and predicted NDVI by 75%, MAE by 31% and RMSE by 75%. For cropland, the correlation is improved by 58%, the MAE by 36% and the RMSE by 76%. We conclude that spatially weighted estimation of crop growth curves can be used to fill spatial and temporal gaps in Landsat NDVI for the purpose of within-field monitoring. Our approach is also applicable to other data sources and vegetation indices. [ABSTRACT FROM AUTHOR]