Non-crossing Quantile Regression Neural Network as a Calibration Tool for Ensemble Weather Forecasts

Abstract

Despite the maturity of ensemble numerical weather prediction (NWP), the resulting forecasts are still, more often than not, under-dispersed. As such, forecast calibration tools have become popular. Among those tools, quantile regression (QR) is highly competitive in terms of both flexibility and predictive performance. Nevertheless, a long-standing problem of QR is quantile crossing, which greatly limits the interpretability of QR-calibrated forecasts. On this point, this study proposes a non-crossing quantile regression neural network (NCQRNN), for calibrating ensemble NWP forecasts into a set of reliable quantile forecasts without crossing. The overarching design principle of NCQRNN is to add on top of the conventional QRNN structure another hidden layer, which imposes a non-decreasing mapping between the combined output from nodes of the last hidden layer to the nodes of the output layer, through a triangular weight matrix with positive entries. The empirical part of the work considers a solar irradiance case study, in which four years of ensemble irradiance forecasts at seven locations, issued by the European Centre for Medium-Range Weather Forecasts, are calibrated via NCQRNN, as well as via an eclectic mix of benchmarking models, ranging from the naïve climatology to the state-of-the-art deep-learning and other non-crossing models. Formal and stringent forecast verification suggests that the forecasts post-processed via NCQRNN attain the maximum sharpness subject to calibration, amongst all competitors. Furthermore, the proposed conception to resolve quantile crossing is remarkably simple yet general, and thus has broad applicability as it can be integrated with many shallow- and deep-learning-based neural networks.

High-resolution projections of outdoor thermal stress in the twenty-first century: a Tasmanian case study

Abstract

To adapt to Earth’s rapidly changing climate, detailed modelling of thermal stress is needed. Dangerous stress levels are becoming more frequent, longer, and more severe. While traditional measurements of thermal stress have focused on air temperature and humidity, modern measures including radiation and wind speed are becoming widespread. However, projecting such indices has presented a challenging problem, due to the need for appropriate bias correction of multiple variables that vary on hourly timescales. In this paper, we aim to provide a detailed understanding of changing thermal stress patterns incorporating modern measurements, bias correction techniques, and hourly projections to assess the impact of climate change on thermal stress at human scales. To achieve these aims, we conduct a case study of projected thermal stress in central Hobart, Australia for 2040–2059, compared to the historical period 1990–2005. We present the first hourly metre-scale projections of thermal stress driven by multivariate bias-corrected data. We bias correct four variables from six dynamically downscaled General Circulation Models. These outputs drive the Solar and LongWave Environmental Irradiance Geometry model at metre scale, calculating mean radiant temperature and the Universal Thermal Climate Index. We demonstrate that multivariate bias correction can correct means on multiple time scales while accurately preserving mean seasonal trends. Changes in mean air temperature and UTCI by hour of the day and month of the year reveal diurnal and annual patterns in both temporal trends and model agreement. We present plots of future median stress values in the context of historical percentiles, revealing trends and patterns not evident in mean data. Our modelling illustrates a future Hobart that experiences higher and more consistent numbers of hours of heat stress arriving earlier in the year and extending further throughout the day.

High-resolution projections of outdoor thermal stress in the twenty-first century: a Tasmanian case study

Abstract

To adapt to Earth’s rapidly changing climate, detailed modelling of thermal stress is needed. Dangerous stress levels are becoming more frequent, longer, and more severe. While traditional measurements of thermal stress have focused on air temperature and humidity, modern measures including radiation and wind speed are becoming widespread. However, projecting such indices has presented a challenging problem, due to the need for appropriate bias correction of multiple variables that vary on hourly timescales. In this paper, we aim to provide a detailed understanding of changing thermal stress patterns incorporating modern measurements, bias correction techniques, and hourly projections to assess the impact of climate change on thermal stress at human scales. To achieve these aims, we conduct a case study of projected thermal stress in central Hobart, Australia for 2040–2059, compared to the historical period 1990–2005. We present the first hourly metre-scale projections of thermal stress driven by multivariate bias-corrected data. We bias correct four variables from six dynamically downscaled General Circulation Models. These outputs drive the Solar and LongWave Environmental Irradiance Geometry model at metre scale, calculating mean radiant temperature and the Universal Thermal Climate Index. We demonstrate that multivariate bias correction can correct means on multiple time scales while accurately preserving mean seasonal trends. Changes in mean air temperature and UTCI by hour of the day and month of the year reveal diurnal and annual patterns in both temporal trends and model agreement. We present plots of future median stress values in the context of historical percentiles, revealing trends and patterns not evident in mean data. Our modelling illustrates a future Hobart that experiences higher and more consistent numbers of hours of heat stress arriving earlier in the year and extending further throughout the day.

CMIP6 precipitation and temperature projections for Chile

Abstract

Precipitation and near-surface temperature from an ensemble of 36 new state‐of‐the‐art climate models under the Coupled Model Inter‐comparison Project phase 6 (CMIP6) are evaluated over Chile’s climate. The analysis is focused on four distinct climatic subregions: Northern Chile, Central Chile, Northern Patagonia, and Southern Patagonia. Over each of the subregions, first, we evaluate the performance of individual global climate models (GCMs) against a suit of precipitation and temperature observation-based gridded datasets over the historical period (1986–2014) and then we analyze the models’ projections for the end of the century (2080–2099) for four different shared socioeconomic pathways scenarios (SSP). Although the models are characterized by general wet and warm mean bias, they reproduce realistically the main spatiotemporal climatic variability over different subregions. However, none of the models is best across all subregions for both precipitation and temperature. Moreover, among the best performing models defined based on the Taylor skill score, one finds the so-called “hot models” likely exhibiting an overestimated climate sensitivity, which suggests caution in using these models for accessing future climate change in Chile. We found robust (90% of models agree in the direction of change) projected end-of-the-century reductions in mean annual precipitation for Central Chile (~ − 20 to ~ − 40%) and Northern Patagonia (~ − 10 to ~ − 30%) under scenario SSP585, but changes are strong from scenario SSP245 onwards, where precipitation is reduced by 10–20%. Northern Chile and Southern Patagonia show non-robust changes in precipitation across the models. Yet, future near-surface temperature warming presented high inter-model agreement across subregions, where the greatest increments occurred along the Andes Mountains. Northern Chile displays the strongest increment of up to ~ 6 °C in SSP585, followed by Central Chile (up to ~ 5 °C). Both Northern and Southern Patagonia show a corresponding increment by up to ~ 4 °C. We also briefly discuss about the environmental and socio-economic implications of these future changes for Chile.

Uncertainty assessment of future climate change using bias-corrected high-resolution multi-regional climate model datasets over East Asia

Abstract

The quantitative assessment of the uncertainty components of future climate projections is critical for decision-makers and organizations to establish climate change adaptation and mitigation strategies at regional or local scales. This is the first study in which the changes in the uncertainty components of future temperature and precipitation projections are quantitatively evaluated using multiple regional climate models over East Asia, vulnerable to future climate change. For temperature, internal variability and model uncertainty were the main factors affecting the near-term projections. The scenario uncertainty continued to increase and was estimated to be the dominant factor affecting the uncertainty after the mid-term projections. Although precipitation has the same main uncertainty factors as the temperature in the near-term projections, it considerably differs from temperature because the internal variability notably contributes to the fraction to the total variance, even in the long-term projections. The internal variability of the temperature and precipitation in the near-term projections were predicted to be larger in Korea than that in East Asia. This was confirmed by regional climate models as well as previous studies using global climate models as to the importance of internal variability at smaller regional scales during the near-term projections. This study is meaningful because it provides new possibilities with respect to the consideration of climate uncertainties to the establishment of climate change policies in more detail on the regional scale.

Assessment of the wind power density over South America simulated by CMIP6 models in the present and future climate

Abstract

Expanding the South American renewable energy matrix to ensure more sustainable socio-economic development, mitigate the climate change effects, and meet the targets set in the Paris Agreement is crucial. Hence, this study sought to estimate South America’s wind speed and wind power density alterations projected by eight global climate models (GCMs) from the Coupled Model Intercomparison Project—Phase 6 (CMIP6). To this end, we applied statistical downscaling and bias correction to the GCMs outputs through the Quantile Delta Mapping method and assessed the projected changes in wind power in future climate under the Shared Socioeconomic Pathways (SSPs) SSP2-4.5 and SSP5-8.5 emission scenarios. ERA5 reanalysis data from 1995 to 2014 validated the models’ historical simulations. The CMIP6 multi-model ensemble indicated an approximate 25–50% increase in wind power density in sectors such as Northeast and South Brazil and growing wind power in regions such as Argentine Patagonia, northern Venezuela, and portions of Uruguay, Bolivia, and Paraguay. Estimates of the wind power growth for the twenty-first century in those regions reiterated their potential performance in the historical period. For the SSP5-8.5 emission scenario, the ensemble projections indicated even more favorable wind power conditions in the sectors mentioned. However, individual projections of wind intensity anomalies obtained by each ensemble member showed a large spread among the GCMs, evidencing the uncertainties associated with the prospects of change in wind power on the continent. Furthermore, this study has presented a first analysis of CMIP6 projections for South American wind power generation, providing relevant information to the energy sector decision-makers.

Hydrologic Extremes in a Changing Climate: a Review of Extremes in East Africa

Abstract

Purpose

Eastern Africa has a complex hydroclimate and socio-economic context, making it vulnerable to climate change-induced hydrological extremes. This review presents recent research on drivers and typologies of extremes across different geographies and highlights challenges and improvements in forecasting hydrological extremes at various timescales.

Recent Findings

Droughts and floods remain the major challenges of the region. Recently, frequent alterations between droughts and floods have been a common occurrence and concern. Research underlines the heterogeneity of extremes and the impact of climate change as increased intensity and duration of extremes. Moreover, the importance of local and antecedent conditions in changing the characteristics of extremes is emphasized.

Summary

A better understanding of these drivers and how they interact is required. Observational and modeling tools must capture these relationships and extremes on short timescales. Although there are improvements in forecasting these extremes, providing relevant information beyond meteorological variables requires further research.

Downscaling Seasonal Precipitation Forecasts over East Africa with Deep Convolutional Neural Networks

Abstract

This study assesses the suitability of convolutional neural networks (CNNs) for downscaling precipitation over East Africa in the context of seasonal forecasting. To achieve this, we design a set of experiments that compare different CNN configurations and deployed the best-performing architecture to downscale one-month lead seasonal forecasts of June–July–August–September (JJAS) precipitation from the Nanjing University of Information Science and Technology Climate Forecast System version 1.0 (NUIST-CFS1.0) for 1982–2020. We also perform hyper-parameter optimization and introduce predictors over a larger area to include information about the main large-scale circulations that drive precipitation over the East Africa region, which improves the downscaling results. Finally, we validate the raw model and downscaled forecasts in terms of both deterministic and probabilistic verification metrics, as well as their ability to reproduce the observed precipitation extreme and spell indicator indices. The results show that the CNN-based downscaling consistently improves the raw model forecasts, with lower bias and more accurate representations of the observed mean and extreme precipitation spatial patterns. Besides, CNN-based downscaling yields a much more accurate forecast of extreme and spell indicators and reduces the significant relative biases exhibited by the raw model predictions. Moreover, our results show that CNN-based downscaling yields better skill scores than the raw model forecasts over most portions of East Africa. The results demonstrate the potential usefulness of CNN in downscaling seasonal precipitation predictions over East Africa, particularly in providing improved forecast products which are essential for end users.

Multivariate Modeling of Precipitation-Induced Home Insurance Risks Using Data Depth

Abstract

While political debates on climate change become increasingly heated, our houses and city infrastructure continue to suffer from an increasing trend of damages due to adverse atmospheric events, from heavier-than-usual rainfalls to heat waves, droughts, and floods. Adapting our homes and critical infrastructure to sustain the effects of climate dynamics requires novel data-driven interdisciplinary approaches for efficient risk mitigation. We develop a new systematic framework based on the machinery of statistical and machine learning tools to evaluate water-related home insurance risks and quantify uncertainty due to varying climate model projections. Furthermore, we introduce the concept of data depth to the analysis of weather and climate ensembles, which remains a novel territory for statistical depth methodology as well as the field of environmental risk and ensemble forecasting in general. We illustrate the new data-driven methodology for risk analysis in application to rainfall-related home insurance in the Canadian Prairies over 2002–2011.

Supplementary materials accompanying this paper appear online.

Probabilistic seasonal precipitation forecasts using quantiles of ensemble forecasts

Abstract

Seasonal precipitation forecasting is vital for weather-sensitive sectors. Global Circulation Models (GCM) routinely produce ensemble Seasonal Climate Forecasts (SCFs) but suffer from issues like low forecast resolution and skills. To address these issues in this study, we introduce a post-processing method, Quantile Ensemble Bayesian Model Averaging (QEBMA). It utilises quantiles from a GCM ensemble forecast to create a pseudo-ensemble forecast. Through their reasonable linear relationships with observations, each pseudo-member connects a hurdle distribution with a point mass at zero for dry months and a gamma distribution for wet months. These distributions are mixed to construct a forecast probability distribution with their weights, proportional to the quantiles’ historical forecast performance. QEBMA is applied to three GCMs, including GloSea5 from the United Kingdom, ECMWF from Europe and ACCESS-S1 from Australia, for monthly precipitation forecasts in 32 locations across four climate zones in Australia. Leave-one-month-out cross-validation results illustrate that QEBMA enhances forecast skills compared to raw GCMs and other post-processing techniques, including quantile mapping and Extended Copula Post-Processing (ECPP), for forecast lead time of 0 to 2 months, based on five metrics. The skill improvements achieved by QEBMA are often statistically significant, particularly when compared to raw GCM forecasts across the 32 study locations. Among these post-processing models, only QEBMA consistently outperforms the SCF benchmark climatology, offering a promising alternative for improving seasonal precipitation forecasts.