Comparing estimates of climate change impacts from process-based and statistical crop models
By David B. Lobell, Senthold Asseng. Environmental Research Letters, 2017
331 citations recorded by the source corpus.
The potential impacts of climate change on crop productivity are of widespread interest to those concerned with addressing climate change and improving global food security. Two common approaches to assess these impacts are process-based simulation models, which attempt to represent key dynamic processes affecting crop yields, and statistical models, which estimate functional relationships between historical observations of weather and yields. Examples of both approaches are increasingly found in the scientific literature, although often published in different disciplinary journals. Here we compare published sensitivities to changes in temperature, precipitation, carbon dioxide (CO2), and ozone from each approach for the subset of crops, locations, and climate scenarios for which both have been applied. Despite a common perception that statistical models are more pessimistic, we find no systematic differences between the predicted sensitivities to warming from process-based and statistical models up to +2 °C, with limited evidence at higher levels of warming. For precipitation, there are many reasons why estimates could be expected to differ, but few estimates exist to develop robust comparisons, and precipitation changes are rarely the dominant factor for predicting impacts given the prominent role of temperature, CO2, and ozone changes. A common difference between process-based and statistical studies is that the former tend to include the effects of CO2 increases that accompany warming, whereas statistical models typically do not. Major needs moving forward include incorporating CO2 effects into statistical studies, improving both approaches’ treatment of ozone, and increasing the use of both methods within the same study. At the same time, those who fund or use crop model projections should understand that in the short-term, both approaches when done well are likely to provide similar estimates of warming impacts, with statistical models generally requiring fewer resources to produce robust estimates, especially when applied to crops beyond the major grains.
Subfields: Ecology, Evolution, Behavior and Systematics, Global and Planetary Change, Plant Science
View Paper: https://www.semanticscholar.org/paper/3b19266b1901fe948401c1a7c76cb25b90ca0fa3
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Ecology, Evolution, Behavior and Systematics
In Global and Planetary Change
In Plant Science
Citations in the corpus
- Machine learning methods for crop yield prediction and climate change impact assessment in agriculture (2018, background): “However, they are generally outperformed by statistical models in prediction over larger spatial scales (Lobell and Burke 2010, Lobell and Asseng 2017).”
- Climate change has likely already affected global food production (2019, background): “Empirical (statistical) global estimates of recent climate change impacts, which some studies have shown to perform equally well as process-based modeling in assessing future impacts on crop yields [11, 12], is available, but only at the coarse national scale [13, 14] or at selected locations, and o”
- The 2022 report of the Lancet Countdown on health and climate change: health at the mercy of fossil fuels (2022, neutral): “141,142 Changing environmental conditions affect the spread of crop and livestock pests and diseases, resulting in production losses.”
- Does rainfall matter for economic growth? Evidence from global sub-national data (1990–2014) (2019, neutral): “We demonstrate that highly aggregated statistical models may underestimate the impact of rainfall due to measurement error, since rainfall is spatially heterogenous, and much more so than temperature (Lobell and Asseng, 2017).”