Browse
Abstract
Understanding climate variability from millennial to glacial–interglacial time scales remains challenging due to the complex and nonlinear feedbacks between ice, ocean, sediments, biosphere, and atmosphere. Complex climate models generally struggle to dynamically and comprehensively simulate such long time periods as a result of the large computational costs. Here, we therefore coupled a dynamical ice sheet model to the Bern3D Earth system model of intermediate complexity, which allows for simulating multiple glacial–interglacial cycles. The performance of the model is first validated against modern observations and its response to abrupt perturbations, such as atmospheric CO2 changes and North Atlantic freshwater hosing, is investigated. To further test the fully coupled model, the climate evolution over the entire last glacial cycle is explored in a transient simulation forced by variations in the orbital configuration and greenhouse gases and aerosols. The model simulates global mean surface temperature in fair agreement with reconstructions, exhibiting a gradual cooling trend since the last interglacial that is interrupted by two more rapid cooling events during the early Marine Isotope Stage (MIS) 4 and Last Glacial Maximum (LGM). Simulated Northern Hemispheric ice sheets show pronounced variability on orbital time scales, and ice volume more than doubles from MIS3 to the LGM in good agreement with recent sea level reconstructions. At the LGM, the Atlantic overturning has a strength of about 14 Sv (1 Sv ≡ 106 m3 s−1), which is a reduction by about one-quarter compared to the preindustrial. We thus demonstrate that the new coupled model is able to simulate large-scale aspects of glacial–interglacial cycles.
Abstract
Understanding climate variability from millennial to glacial–interglacial time scales remains challenging due to the complex and nonlinear feedbacks between ice, ocean, sediments, biosphere, and atmosphere. Complex climate models generally struggle to dynamically and comprehensively simulate such long time periods as a result of the large computational costs. Here, we therefore coupled a dynamical ice sheet model to the Bern3D Earth system model of intermediate complexity, which allows for simulating multiple glacial–interglacial cycles. The performance of the model is first validated against modern observations and its response to abrupt perturbations, such as atmospheric CO2 changes and North Atlantic freshwater hosing, is investigated. To further test the fully coupled model, the climate evolution over the entire last glacial cycle is explored in a transient simulation forced by variations in the orbital configuration and greenhouse gases and aerosols. The model simulates global mean surface temperature in fair agreement with reconstructions, exhibiting a gradual cooling trend since the last interglacial that is interrupted by two more rapid cooling events during the early Marine Isotope Stage (MIS) 4 and Last Glacial Maximum (LGM). Simulated Northern Hemispheric ice sheets show pronounced variability on orbital time scales, and ice volume more than doubles from MIS3 to the LGM in good agreement with recent sea level reconstructions. At the LGM, the Atlantic overturning has a strength of about 14 Sv (1 Sv ≡ 106 m3 s−1), which is a reduction by about one-quarter compared to the preindustrial. We thus demonstrate that the new coupled model is able to simulate large-scale aspects of glacial–interglacial cycles.
Abstract
In this study, we analyze drivers of non–El Niño–Southern Oscillation (ENSO) precipitation variability in the Southwest United States (SWUS) and the influence of the atmospheric basic state, using atmosphere-only and ocean–atmosphere coupled simulations from the Community Earth System Model version 2 (CESM2) large ensemble. A cluster analysis identifies three main wave trains associated with non-ENSO SWUS precipitation in the experiments: a meridional ENSO-type wave train, an arching Pacific–North American-type (PNA) wave train, and a circumglobal zonal wave train. The zonal wave train cluster frequency differs between models and ENSO phase, with decreased frequency during El Niño and the coupled runs, and increased frequency during La Niña and the atmosphere-only runs. This is consistent with an El Niño–like bias of the atmospheric circulation in the coupled model, with strengthened subtropical westerlies in the central and eastern North Pacific that cause a retraction of the waveguide in the midlatitude eastern North Pacific. As such, zonal wave trains from the East Asian jet stream (EAJS) are more likely to be diverted southward in the east Pacific in the coupled large ensemble, with a consequently smaller role in driving SWUS precipitation variability. This study illustrates the need to reduce model biases in the background flow, particularly relating to the jet stream, in order to accurately capture the role of large-scale teleconnections in driving SWUS precipitation variability and improve future forecasting capabilities.
Abstract
In this study, we analyze drivers of non–El Niño–Southern Oscillation (ENSO) precipitation variability in the Southwest United States (SWUS) and the influence of the atmospheric basic state, using atmosphere-only and ocean–atmosphere coupled simulations from the Community Earth System Model version 2 (CESM2) large ensemble. A cluster analysis identifies three main wave trains associated with non-ENSO SWUS precipitation in the experiments: a meridional ENSO-type wave train, an arching Pacific–North American-type (PNA) wave train, and a circumglobal zonal wave train. The zonal wave train cluster frequency differs between models and ENSO phase, with decreased frequency during El Niño and the coupled runs, and increased frequency during La Niña and the atmosphere-only runs. This is consistent with an El Niño–like bias of the atmospheric circulation in the coupled model, with strengthened subtropical westerlies in the central and eastern North Pacific that cause a retraction of the waveguide in the midlatitude eastern North Pacific. As such, zonal wave trains from the East Asian jet stream (EAJS) are more likely to be diverted southward in the east Pacific in the coupled large ensemble, with a consequently smaller role in driving SWUS precipitation variability. This study illustrates the need to reduce model biases in the background flow, particularly relating to the jet stream, in order to accurately capture the role of large-scale teleconnections in driving SWUS precipitation variability and improve future forecasting capabilities.
Abstract
Machine learning algorithms are able to capture complex, nonlinear, interacting relationships and are increasingly used to predict agricultural yield variability at regional and national scales. Using explainable artificial intelligence (XAI) methods applied to such algorithms may enable better scientific understanding of drivers of yield variability. However, XAI methods may provide misleading results when applied to spatiotemporal correlated datasets. In this study, machine learning models are trained to predict simulated crop yield from climate indices, and the impact of cross-validation strategy on the interpretation and performance of the resulting models is assessed. Using data from a process-based crop model allows us to then comment on the plausibility of the “explanations” provided by XAI methods. Our results show that the choice of evaluation strategy has an impact on (i) interpretations of the model and (ii) model skill on held-out years and regions, after the evaluation strategy is used for hyperparameter tuning and feature selection. We find that use of a cross-validation strategy based on clustering in feature space achieves the most plausible interpretations as well as the best model performance on held-out years and regions. Our results provide the first steps toward identifying domain-specific “best practices” for the use of XAI tools on spatiotemporal agricultural or climatic data.
Significance Statement
“Explainable” or “interpretable” machine learning (XAI) methods have been increasingly used in scientific research to study complex relationships between climatic and biogeoscientific variables (such as crop yield). However, these methods can return contradictory, implausible, or ambiguous results. In this study, we train machine learning models to predict maize yield anomalies and vary the model evaluation method used. We find that the evaluation (cross validation) method used has an effect on model interpretation results and on the skill of resulting models in held-out years and regions. These results have implications for the methodological design of studies that aim to use XAI tools to identify drivers of, for example, crop yield variability.
Abstract
Machine learning algorithms are able to capture complex, nonlinear, interacting relationships and are increasingly used to predict agricultural yield variability at regional and national scales. Using explainable artificial intelligence (XAI) methods applied to such algorithms may enable better scientific understanding of drivers of yield variability. However, XAI methods may provide misleading results when applied to spatiotemporal correlated datasets. In this study, machine learning models are trained to predict simulated crop yield from climate indices, and the impact of cross-validation strategy on the interpretation and performance of the resulting models is assessed. Using data from a process-based crop model allows us to then comment on the plausibility of the “explanations” provided by XAI methods. Our results show that the choice of evaluation strategy has an impact on (i) interpretations of the model and (ii) model skill on held-out years and regions, after the evaluation strategy is used for hyperparameter tuning and feature selection. We find that use of a cross-validation strategy based on clustering in feature space achieves the most plausible interpretations as well as the best model performance on held-out years and regions. Our results provide the first steps toward identifying domain-specific “best practices” for the use of XAI tools on spatiotemporal agricultural or climatic data.
Significance Statement
“Explainable” or “interpretable” machine learning (XAI) methods have been increasingly used in scientific research to study complex relationships between climatic and biogeoscientific variables (such as crop yield). However, these methods can return contradictory, implausible, or ambiguous results. In this study, we train machine learning models to predict maize yield anomalies and vary the model evaluation method used. We find that the evaluation (cross validation) method used has an effect on model interpretation results and on the skill of resulting models in held-out years and regions. These results have implications for the methodological design of studies that aim to use XAI tools to identify drivers of, for example, crop yield variability.
Abstract
This paper presents the Thunderstorm Nowcasting Tool (ThunderCast), a 24-hour, year round model for predicting the location of convection that is likely to initiate or remain a thunderstorm in the next 0-60 minutes in the continental United States, adapted from existing deep learning convection applications. ThunderCast utilizes a U-Net convolutional neural network for semantic segmentation trained on 320 km by 320 km data patches with four inputs and one target dataset. The inputs are satellite bands from the Geostationary Operational Environmental Satellite (GOES-16) Advanced Baseline Imager (ABI) in the visible, shortwave infrared, and longwave infrared spectrum, and the target is Multi-Radar Multi-Sensor (MRMS) radar reflectivity at the - 10°C isothermin the atmosphere. On a pixel-by-pixel basis, ThunderCast has high accuracy, recall, and specificity but is subject to false positive predictions resulting in low precision. However, the number of false positives decreases when buffering the target values with a 15×15 km centered window indicating ThunderCast’s predictions are useful within a buffered area. To demonstrate the initial prediction capabilities of ThunderCast, three case studies are presented: a mesoscale convective vortex, sea breeze convection, and monsoonal convection in the southwestern United States. The case studies illustrate that the ThunderCast model effectively nowcasts the location of newly initiated and ongoing active convection, within the next 60 minutes, under a variety of geographic and meteorological conditions.
Abstract
This paper presents the Thunderstorm Nowcasting Tool (ThunderCast), a 24-hour, year round model for predicting the location of convection that is likely to initiate or remain a thunderstorm in the next 0-60 minutes in the continental United States, adapted from existing deep learning convection applications. ThunderCast utilizes a U-Net convolutional neural network for semantic segmentation trained on 320 km by 320 km data patches with four inputs and one target dataset. The inputs are satellite bands from the Geostationary Operational Environmental Satellite (GOES-16) Advanced Baseline Imager (ABI) in the visible, shortwave infrared, and longwave infrared spectrum, and the target is Multi-Radar Multi-Sensor (MRMS) radar reflectivity at the - 10°C isothermin the atmosphere. On a pixel-by-pixel basis, ThunderCast has high accuracy, recall, and specificity but is subject to false positive predictions resulting in low precision. However, the number of false positives decreases when buffering the target values with a 15×15 km centered window indicating ThunderCast’s predictions are useful within a buffered area. To demonstrate the initial prediction capabilities of ThunderCast, three case studies are presented: a mesoscale convective vortex, sea breeze convection, and monsoonal convection in the southwestern United States. The case studies illustrate that the ThunderCast model effectively nowcasts the location of newly initiated and ongoing active convection, within the next 60 minutes, under a variety of geographic and meteorological conditions.
