Claim · model-written summary
Machine learning explained 93% of observed yield variability.
Not reportedTaxonMaterialGeneCompoundOrgan / stageInteractorInteractor roleSecond interactorSecond interactor roleTreatmentComparatorSettingDirectionQualifier
Verbatim
e0“machine learning can explain 93% of observed yield variability”
Title
Guoyong Leng; Jim W Hall Environmental Research Letters 2020. doi:10.1088/1748-9326/ab7b24
Other claims from this paper (9)
- Method / resource Regression and machine learning reproduced observed yield-average patterns.
- Method / resource Process-based crop models had large bias in yield averages despite harmonized parameters.
- Method / resource Machine learning had the best skill for yield distributions, followed by regression and process-based models.
- Method / resource Regression explained 51% of observed yield variability.
- Method / resource Process-based models explained 42% of observed yield variability.
- Observation US maize yield was projected to decrease 13.5% under 2 °C warming.
- Observation Baseline low-yield years were projected in 19% of years at 1.5 °C warming.
- Observation Baseline low-yield years were projected in 25% of years at 2 °C warming.
- Method / resource Machine learning and regression were computationally more efficient than process-based models.