Browse
Diagnosing Seasonal Forecast Skill of the Indian Ocean Dipole Mode Using Model AnalogsAbstract
A retrospective tropical Indian Ocean dipole mode (IOD) hindcast for 1958–2014 was conducted using 20 models from the sixth phase of the Coupled Model Intercomparison Project (CMIP6), with a model-based analog forecast (MAF) method. In the MAF approach, forecast ensembles are extracted from preexisting model simulations by finding the states that initially best match an observed anomaly and tracking their subsequent evolution, with no additional model integrations. By optimizing the key factors in the MAF method, we suggest that the optimal domain for the analog criteria should be concentrated in the tropical Indian Ocean region for IOD predictions. Including external forcing trends improves the skills of the east and west poles of the IOD, but not the IOD prediction itself. The MAF IOD prediction showed comparable skills to the assimilation-initialized hindcast, with skillful predictions corresponding to a 4- and 3-month lead, respectively. The IOD forecast skill had significant decadal variations during the 55-yr period, with low skill after the early 2000s and before 1985 and high skill during 1985–2000. This work offers a computational efficient and practical approach for seasonal prediction of the tropical Indian Ocean sea surface temperature.
Abstract
A retrospective tropical Indian Ocean dipole mode (IOD) hindcast for 1958–2014 was conducted using 20 models from the sixth phase of the Coupled Model Intercomparison Project (CMIP6), with a model-based analog forecast (MAF) method. In the MAF approach, forecast ensembles are extracted from preexisting model simulations by finding the states that initially best match an observed anomaly and tracking their subsequent evolution, with no additional model integrations. By optimizing the key factors in the MAF method, we suggest that the optimal domain for the analog criteria should be concentrated in the tropical Indian Ocean region for IOD predictions. Including external forcing trends improves the skills of the east and west poles of the IOD, but not the IOD prediction itself. The MAF IOD prediction showed comparable skills to the assimilation-initialized hindcast, with skillful predictions corresponding to a 4- and 3-month lead, respectively. The IOD forecast skill had significant decadal variations during the 55-yr period, with low skill after the early 2000s and before 1985 and high skill during 1985–2000. This work offers a computational efficient and practical approach for seasonal prediction of the tropical Indian Ocean sea surface temperature.
Abstract
For the tropical country of Sri Lanka, subseasonal variability in precipitation is both ecologically and societally relevant, influencing agricultural yields, natural hazard risk, energy production, and disease incidence. The primary driver of this subseasonal precipitation variability is the Madden–Julian oscillation (MJO). Here we investigate this influence on Sri Lankan precipitation across seasons, describing MJO-associated precipitation patterns and exploring the potential for MJO-informed subseasonal forecasts. We do so using 40-yr satellite-derived records of precipitation with high spatial resolution (from CHIRPS v2.0) and related meteorological and atmospheric fields (from ERA5 and MERRA-2). We find a direct MJO influence on precipitation corresponding to propagation of the MJO’s convectively active region and suppressed region near Sri Lanka, with the strength and spatial patterns of this influence differing across seasons. There are particularly strong impacts in the second intermonsoon (SIM; October–November) and southwest monsoon (SWM; May–September) seasons. During SIM the impacts are island-wide, but strongest in the northeast. During the SWM the absolute impacts are localized to the southwest, but the relative impacts (i.e., relative to precipitation climatology) are fairly uniform across the island. Moreover, we find significant associations between MJO phase and Sri Lankan precipitation at time scales of up to several weeks. Notably, these associations are stronger when using the OLR-based MJO index (OMI) rather than the more commonly used real-time multivariate MJO index (RMM). While the MJO associations we describe here arise from a highly simplified forecasting scheme, they provide a foundation and impetus for developing a more complete, MJO-informed precipitation forecast method.
Significance Statement
Rainfall variability at the subseasonal (weeks–months) time scale is critical to societal well-being, given its fundamental importance for agriculture, flood risk, hydropower generation, and disease incidence. Our work describes how such rainfall variability in Sri Lanka is impacted by the Madden–Julian oscillation, in which a region of enhanced rainfall and cloudiness, paired with a region of decreased rainfall and cloudiness, circles the globe every 30–60 days. Our results suggest that its influence on Sri Lankan rainfall may be strong enough that incorporating knowledge of the Madden–Julian oscillation into forecasts can improve the accuracy of rainfall prediction for Sri Lanka. Future work should develop a more comprehensive forecast method to assess viability in real-world forecasting scenarios.
Abstract
For the tropical country of Sri Lanka, subseasonal variability in precipitation is both ecologically and societally relevant, influencing agricultural yields, natural hazard risk, energy production, and disease incidence. The primary driver of this subseasonal precipitation variability is the Madden–Julian oscillation (MJO). Here we investigate this influence on Sri Lankan precipitation across seasons, describing MJO-associated precipitation patterns and exploring the potential for MJO-informed subseasonal forecasts. We do so using 40-yr satellite-derived records of precipitation with high spatial resolution (from CHIRPS v2.0) and related meteorological and atmospheric fields (from ERA5 and MERRA-2). We find a direct MJO influence on precipitation corresponding to propagation of the MJO’s convectively active region and suppressed region near Sri Lanka, with the strength and spatial patterns of this influence differing across seasons. There are particularly strong impacts in the second intermonsoon (SIM; October–November) and southwest monsoon (SWM; May–September) seasons. During SIM the impacts are island-wide, but strongest in the northeast. During the SWM the absolute impacts are localized to the southwest, but the relative impacts (i.e., relative to precipitation climatology) are fairly uniform across the island. Moreover, we find significant associations between MJO phase and Sri Lankan precipitation at time scales of up to several weeks. Notably, these associations are stronger when using the OLR-based MJO index (OMI) rather than the more commonly used real-time multivariate MJO index (RMM). While the MJO associations we describe here arise from a highly simplified forecasting scheme, they provide a foundation and impetus for developing a more complete, MJO-informed precipitation forecast method.
Significance Statement
Rainfall variability at the subseasonal (weeks–months) time scale is critical to societal well-being, given its fundamental importance for agriculture, flood risk, hydropower generation, and disease incidence. Our work describes how such rainfall variability in Sri Lanka is impacted by the Madden–Julian oscillation, in which a region of enhanced rainfall and cloudiness, paired with a region of decreased rainfall and cloudiness, circles the globe every 30–60 days. Our results suggest that its influence on Sri Lankan rainfall may be strong enough that incorporating knowledge of the Madden–Julian oscillation into forecasts can improve the accuracy of rainfall prediction for Sri Lanka. Future work should develop a more comprehensive forecast method to assess viability in real-world forecasting scenarios.
Abstract
Dynamical downscaling is a crucial process for providing regional climate information for broad uses, using coarser-resolution global models to drive higher-resolution regional climate simulations. The pool of global climate models (GCMs) providing the fields needed for dynamical downscaling has increased from the previous generations of the Coupled Model Intercomparison Project (CMIP). However, with limited computational resources, the need for prioritizing the GCMs for subsequent downscaling studies remains. GCM selection for dynamical downscaling should focus on evaluating processes relevant for providing boundary conditions to the regional models and be inspired by regional uses such as the response of extremes to changes in the boundary conditions. This leads to the need for metrics representing processes of relevance to diverse stakeholders and subregions of a domain. Procedures to account for metric redundancy and the statistical distinguishability of GCM rankings are required. Further, procedures for selecting realizations from ensembles of top-performing GCM simulations can be used to span the range of climate change signals in multiple ways. As a result, distinct weighting of metrics and prioritization of particular realizations may depend on user needs. We provide high-level guidelines for such region-specific evaluations and address how CMIP7 might enable dynamical downscaling of a representative sample of high-quality models across representative shared socioeconomic pathways (SSPs).
Abstract
Dynamical downscaling is a crucial process for providing regional climate information for broad uses, using coarser-resolution global models to drive higher-resolution regional climate simulations. The pool of global climate models (GCMs) providing the fields needed for dynamical downscaling has increased from the previous generations of the Coupled Model Intercomparison Project (CMIP). However, with limited computational resources, the need for prioritizing the GCMs for subsequent downscaling studies remains. GCM selection for dynamical downscaling should focus on evaluating processes relevant for providing boundary conditions to the regional models and be inspired by regional uses such as the response of extremes to changes in the boundary conditions. This leads to the need for metrics representing processes of relevance to diverse stakeholders and subregions of a domain. Procedures to account for metric redundancy and the statistical distinguishability of GCM rankings are required. Further, procedures for selecting realizations from ensembles of top-performing GCM simulations can be used to span the range of climate change signals in multiple ways. As a result, distinct weighting of metrics and prioritization of particular realizations may depend on user needs. We provide high-level guidelines for such region-specific evaluations and address how CMIP7 might enable dynamical downscaling of a representative sample of high-quality models across representative shared socioeconomic pathways (SSPs).
Abstract
Verif is an open-source tool for verifying weather predictions against a ground truth. It is suitable for a range of applications and designed for iterative product development involving fine-tuning of algorithms, comparing methods, and addressing scientific issues with the product. The tool generates verification plots based on user-supplied input files containing predictions and observations for multiple point-locations, forecast lead times, and forecast initialization times. It supports over 90 verification metrics and diagrams and can evaluate deterministic and probabilistic predictions. An extensive set of command-line flags control how the input data are aggregated, filtered, stratified, and visualized. The broad range of metrics and data manipulation options allows the user to gain insight from both summary scores and detailed time series of individual weather events. Verif is suitable for many applications, including assessing numerical weather prediction models, climate models, reanalyses, machine learning models, and even the fidelity of emerging observational sources. The tool has matured through long-term development at the Norwegian Meteorological Institute and the University of British Columbia. Verif comes with an extensive wiki page and example input files covering a wide range of prediction applications, allowing students and researchers interested in verification to get hands-on experience with real-life datasets. This article describes the functionality of Verif version 1.3 and shows how the tool can be used for effective product development.
Abstract
Verif is an open-source tool for verifying weather predictions against a ground truth. It is suitable for a range of applications and designed for iterative product development involving fine-tuning of algorithms, comparing methods, and addressing scientific issues with the product. The tool generates verification plots based on user-supplied input files containing predictions and observations for multiple point-locations, forecast lead times, and forecast initialization times. It supports over 90 verification metrics and diagrams and can evaluate deterministic and probabilistic predictions. An extensive set of command-line flags control how the input data are aggregated, filtered, stratified, and visualized. The broad range of metrics and data manipulation options allows the user to gain insight from both summary scores and detailed time series of individual weather events. Verif is suitable for many applications, including assessing numerical weather prediction models, climate models, reanalyses, machine learning models, and even the fidelity of emerging observational sources. The tool has matured through long-term development at the Norwegian Meteorological Institute and the University of British Columbia. Verif comes with an extensive wiki page and example input files covering a wide range of prediction applications, allowing students and researchers interested in verification to get hands-on experience with real-life datasets. This article describes the functionality of Verif version 1.3 and shows how the tool can be used for effective product development.
Abstract
It is common when speaking colloquially to describe climate as the average weather, which implies weather is the driver and climatic averages are a passive by-product of it, but it is useful to reframe this toward weather being the “expression” of climate. That is, a region’s climate defines the range of weather it might experience (including the extent and frequency of extremes). In this framing, weather is driven by a region’s climate. A changing climate then, necessarily, is experienced as a change in local weather events—often most visibly through changes in the extent or frequency of extreme weather.
Abstract
It is common when speaking colloquially to describe climate as the average weather, which implies weather is the driver and climatic averages are a passive by-product of it, but it is useful to reframe this toward weather being the “expression” of climate. That is, a region’s climate defines the range of weather it might experience (including the extent and frequency of extremes). In this framing, weather is driven by a region’s climate. A changing climate then, necessarily, is experienced as a change in local weather events—often most visibly through changes in the extent or frequency of extreme weather.
Abstract
This paper explores the application of emerging machine learning methods from image super-resolution (SR) to the task of statistical downscaling. We specifically focus on convolutional neural network-based Generative Adversarial Networks (GANs). Our GANs are conditioned on low-resolution (LR) inputs to generate high-resolution (HR) surface winds emulating Weather Research and Forecasting (WRF) model simulations over North America. Unlike traditional SR models, where LR inputs are idealized coarsened versions of the HR images, WRF emulation involves using non-idealized LR and HR pairs resulting in shared-scale mismatches due to internal variability. Our study builds upon current SR-based statistical downscaling by experimenting with a novel frequency-separation (FS) approach from the computer vision field. To assess the skill of SR models, we carefully select evaluation metrics, and focus on performance measures based on spatial power spectra. Our analyses reveal how GAN configurations influence spatial structures in the generated fields, particularly biases in spatial variability spectra. Using power spectra to evaluate the FS experiments reveals that successful applications of FS in computer vision do not translate to climate fields. However, the FS experiments demonstrate the sensitivity of power spectra to a commonly used GAN-based SR objective function, which helps interpret and understand its role in determining spatial structures. This result motivates the development of a novel partial frequency-separation scheme as a promising configuration option. We also quantify the influence on GAN performance of non-idealized LR fields resulting from internal variability. Furthermore, we conduct a spectra-based feature-importance experiment allowing us to explore the dependence of the spatial structure of generated fields on different physically relevant LR covariates.
Abstract
This paper explores the application of emerging machine learning methods from image super-resolution (SR) to the task of statistical downscaling. We specifically focus on convolutional neural network-based Generative Adversarial Networks (GANs). Our GANs are conditioned on low-resolution (LR) inputs to generate high-resolution (HR) surface winds emulating Weather Research and Forecasting (WRF) model simulations over North America. Unlike traditional SR models, where LR inputs are idealized coarsened versions of the HR images, WRF emulation involves using non-idealized LR and HR pairs resulting in shared-scale mismatches due to internal variability. Our study builds upon current SR-based statistical downscaling by experimenting with a novel frequency-separation (FS) approach from the computer vision field. To assess the skill of SR models, we carefully select evaluation metrics, and focus on performance measures based on spatial power spectra. Our analyses reveal how GAN configurations influence spatial structures in the generated fields, particularly biases in spatial variability spectra. Using power spectra to evaluate the FS experiments reveals that successful applications of FS in computer vision do not translate to climate fields. However, the FS experiments demonstrate the sensitivity of power spectra to a commonly used GAN-based SR objective function, which helps interpret and understand its role in determining spatial structures. This result motivates the development of a novel partial frequency-separation scheme as a promising configuration option. We also quantify the influence on GAN performance of non-idealized LR fields resulting from internal variability. Furthermore, we conduct a spectra-based feature-importance experiment allowing us to explore the dependence of the spatial structure of generated fields on different physically relevant LR covariates.
Abstract
Recent severe droughts, extreme floods, and increasing differences between seasonal high and low flows on the Amazon River may represent a twenty-first-century increase in the amplitude of the hydrologic cycle over the Amazon Basin. These precipitation and streamflow changes may have arisen from natural ocean–atmospheric variability, deforestation within the drainage basin of the Amazon River, or anthropogenic climate change. Tree-ring reconstructions of wet-season precipitation extremes, substantiated with historical accounts of climate and river levels on the Amazon River and in northeast Brazil found in the Brazilian Digital Library, indicate that the recent river-level extremes on the Amazon may have been equaled or possibly exceeded during the preinstrumental nineteenth century. The “Forgotten Drought” of 1865 was the lowest wet-season rainfall total reconstructed with tree-rings in the eastern Amazon from 1790 to 2016 and appears to have been one of the lowest stream levels observed on the Amazon River during the historical era according to first-hand descriptions by Louis Agassiz, his Brazilian colleague João Martins da Silva Coutinho, and others. Heavy rains and flooding are described during most of the tree-ring-reconstructed wet extremes, including the complete inundation of “First Street” in Santarem, Brazil, in 1859 and the overtopping of the Bittencourt Bridge in Manaus, Brazil, in 1892. These extremes in the tree-ring estimates and historical observations indicate that recent high and low flow anomalies on the Amazon River may not have exceeded the natural variability of precipitation and streamflow during the nineteenth century.
Significance Statement
Proxy tree-ring and historical evidence for precipitation extremes during the preinstrumental nineteenth century indicate that recent floods and droughts on the Amazon River may have not yet exceeded the range of natural hydroclimatic variability.
Abstract
Recent severe droughts, extreme floods, and increasing differences between seasonal high and low flows on the Amazon River may represent a twenty-first-century increase in the amplitude of the hydrologic cycle over the Amazon Basin. These precipitation and streamflow changes may have arisen from natural ocean–atmospheric variability, deforestation within the drainage basin of the Amazon River, or anthropogenic climate change. Tree-ring reconstructions of wet-season precipitation extremes, substantiated with historical accounts of climate and river levels on the Amazon River and in northeast Brazil found in the Brazilian Digital Library, indicate that the recent river-level extremes on the Amazon may have been equaled or possibly exceeded during the preinstrumental nineteenth century. The “Forgotten Drought” of 1865 was the lowest wet-season rainfall total reconstructed with tree-rings in the eastern Amazon from 1790 to 2016 and appears to have been one of the lowest stream levels observed on the Amazon River during the historical era according to first-hand descriptions by Louis Agassiz, his Brazilian colleague João Martins da Silva Coutinho, and others. Heavy rains and flooding are described during most of the tree-ring-reconstructed wet extremes, including the complete inundation of “First Street” in Santarem, Brazil, in 1859 and the overtopping of the Bittencourt Bridge in Manaus, Brazil, in 1892. These extremes in the tree-ring estimates and historical observations indicate that recent high and low flow anomalies on the Amazon River may not have exceeded the natural variability of precipitation and streamflow during the nineteenth century.
Significance Statement
Proxy tree-ring and historical evidence for precipitation extremes during the preinstrumental nineteenth century indicate that recent floods and droughts on the Amazon River may have not yet exceeded the range of natural hydroclimatic variability.
