• Energy Research

Les méthodes de réduction d'échelle statistiques sont des outils de post-traitement populaires qui sont largement utilisés dans de nombreux secteurs pour adapter les résultats biaisés à résolution grossière des simulations climatiques mondiales à l'échelle régionale à locale généralement requise par les utilisateurs. Ils vont de méthodes simples et pragmatiques de correction de biais (BC), qui ajustent directement les sorties du modèle d'intérêt (par exemple les précipitations) en fonction des observations locales disponibles, à des méthodes plus complexes de pronostic parfait (PP), qui dérivent indirectement des prédictions locales (par exemple les précipitations) à partir de variables appropriées du modèle à grande échelle de l'air supérieur (prédicteurs). Les méthodes de réduction d'échelle statistiques ont été largement utilisées et évaluées de manière critique dans les applications du changement climatique ; cependant, leurs avantages et leurs limites dans les prévisions saisonnières ne sont pas encore bien compris. En particulier, un problème clé dans ce contexte est de savoir s'ils servent à améliorer la qualité/compétence prévisionnelle des résultats des modèles bruts au-delà de l'ajustement de leurs biais systématiques. Dans cet article, nous analysons ce problème en appliquant deux méthodes BC et deux méthodes PP à la pointe de la technologie pour réduire les précipitations à partir d'un rétroprojecteur saisonnier multimodèle dans une région tropicale difficile, les Philippines. Pour évaluer correctement la valeur ajoutée potentielle au-delà de la réduction des biais du modèle, nous considérons deux scores de validation qui ne sont pas sensibles aux changements de la moyenne (catégories de corrélation et de fiabilité). Nos résultats montrent que, alors que les méthodes BC maintiennent ou aggravent la compétence des prévisions du modèle brut, les méthodes PP peuvent apporter une amélioration significative des compétences (aggravation) dans les cas où les variables prédictives à grande échelle considérées sont meilleures (pires) prédites par le modèle que les précipitations. Par exemple, les méthodes PP augmentent (diminuent) la fiabilité du modèle dans près de 40 % des stations considérées en été boréal (automne). Par conséquent, le choix d'une approche pratique de réduction d'échelle (BC ou PP) dépend de la région et de la saison. Los métodos de reducción de escala estadística son herramientas populares de posprocesamiento que se utilizan ampliamente en muchos sectores para adaptar los resultados sesgados de resolución gruesa de las simulaciones climáticas globales a la escala regional a local que generalmente requieren los usuarios. Van desde métodos de corrección de sesgo (BC) simples y pragmáticos, que ajustan directamente los resultados del modelo de interés (por ejemplo, precipitación) de acuerdo con las observaciones locales disponibles, hasta métodos de pronóstico perfecto (PP) más complejos, que derivan indirectamente predicciones locales (por ejemplo, precipitación) de variables apropiadas del modelo a gran escala del aire superior (predictores). Los métodos estadísticos de reducción de escala se han utilizado ampliamente y se han evaluado críticamente en aplicaciones de cambio climático; sin embargo, sus ventajas y limitaciones en el pronóstico estacional aún no se comprenden bien. En particular, un problema clave en este contexto es si sirven para mejorar la calidad/habilidad de pronóstico de los resultados del modelo bruto más allá del ajuste de sus sesgos sistemáticos. En este documento analizamos este problema aplicando dos métodos BC y dos PP de última generación para reducir la precipitación de un retroceso estacional multimodelo en una región tropical desafiante, Filipinas. Para evaluar adecuadamente el valor añadido potencial más allá de la reducción de los sesgos del modelo, consideramos dos puntuaciones de validación que no son sensibles a los cambios en la media (categorías de correlación y fiabilidad). Nuestros resultados muestran que, mientras que los métodos BC mantienen o empeoran la habilidad de los pronósticos del modelo en bruto, los métodos PP pueden producir una mejora significativa de la habilidad (empeoramiento) en los casos en que las variables predictoras a gran escala consideradas son mejores (peores) predichas por el modelo que la precipitación. Por ejemplo, se encuentra que los métodos PP aumentan (disminuyen) la confiabilidad del modelo en casi el 40% de las estaciones consideradas en el verano boreal (otoño). Por lo tanto, la elección de un enfoque de reducción de escala conveniente (ya sea BC o PP) depende de la región y la temporada. Statistical downscaling methods are popular post-processing tools which are widely used in many sectors to adapt the coarse-resolution biased outputs from global climate simulations to the regional-to-local scale typically required by users. They range from simple and pragmatic Bias Correction (BC) methods, which directly adjust the model outputs of interest (e.g. precipitation) according to the available local observations, to more complex Perfect Prognosis (PP) ones, which indirectly derive local predictions (e.g. precipitation) from appropriate upper-air large-scale model variables (predictors). Statistical downscaling methods have been extensively used and critically assessed in climate change applications; however, their advantages and limitations in seasonal forecasting are not well understood yet. In particular, a key problem in this context is whether they serve to improve the forecast quality/skill of raw model outputs beyond the adjustment of their systematic biases. In this paper we analyze this issue by applying two state-of-the-art BC and two PP methods to downscale precipitation from a multimodel seasonal hindcast in a challenging tropical region, the Philippines. To properly assess the potential added value beyond the reduction of model biases, we consider two validation scores which are not sensitive to changes in the mean (correlation and reliability categories). Our results show that, whereas BC methods maintain or worsen the skill of the raw model forecasts, PP methods can yield significant skill improvement (worsening) in cases for which the large-scale predictor variables considered are better (worse) predicted by the model than precipitation. For instance, PP methods are found to increase (decrease) model reliability in nearly 40% of the stations considered in boreal summer (autumn). Therefore, the choice of a convenient downscaling approach (either BC or PP) depends on the region and the season. طرق تقليص النطاق الإحصائي هي أدوات شائعة لما بعد المعالجة تستخدم على نطاق واسع في العديد من القطاعات لتكييف المخرجات المتحيزة ذات الدقة الخشنة من محاكاة المناخ العالمي إلى النطاق الإقليمي إلى المحلي المطلوب عادةً من قبل المستخدمين. وهي تتراوح من طرق بسيطة وعملية لتصحيح التحيز (BC)، والتي تعدل بشكل مباشر مخرجات النموذج محل الاهتمام (مثل هطول الأمطار) وفقًا للملاحظات المحلية المتاحة، إلى طرق التكهن المثالي (PP) الأكثر تعقيدًا، والتي تستمد بشكل غير مباشر التنبؤات المحلية (مثل هطول الأمطار) من متغيرات النموذج المناسبة واسعة النطاق في الهواء العلوي (التنبؤات). تم استخدام طرق تقليص النطاق الإحصائي على نطاق واسع وتقييمها بشكل نقدي في تطبيقات تغير المناخ ؛ ومع ذلك، فإن مزاياها وقيودها في التنبؤ الموسمي ليست مفهومة جيدًا بعد. على وجه الخصوص، تتمثل المشكلة الرئيسية في هذا السياق في ما إذا كانت تعمل على تحسين جودة/مهارة التنبؤ لمخرجات النموذج الخام بما يتجاوز تعديل تحيزاتها المنهجية. في هذه الورقة، نقوم بتحليل هذه المشكلة من خلال تطبيق طريقتين حديثتين قبل الميلاد وطريقتين للبولي بروبلين لتقليل هطول الأمطار من توقعات موسمية متعددة النماذج في منطقة استوائية صعبة، الفلبين. لتقييم القيمة المضافة المحتملة بشكل صحيح بما يتجاوز الحد من تحيزات النموذج، نأخذ في الاعتبار درجتي التحقق غير الحساستين للتغيرات في المتوسط (فئتي الارتباط والموثوقية). تظهر نتائجنا أنه في حين أن طرق BC تحافظ على مهارة تنبؤات النموذج الخام أو تزيدها سوءًا، فإن طرق PP يمكن أن تسفر عن تحسن كبير في المهارات (تدهور) في الحالات التي تكون فيها المتغيرات التنبؤية واسعة النطاق التي يعتبرها النموذج أفضل (أسوأ) من هطول الأمطار. على سبيل المثال، تم العثور على طرق PP لزيادة (تقليل) موثوقية النموذج في ما يقرب من 40 ٪ من المحطات التي يتم النظر فيها في الصيف الشمالي (الخريف). لذلك، يعتمد اختيار نهج تصغير النطاق المناسب (إما BC أو PP) على المنطقة والموسم.

AbstractInternal variability, multiple emission scenarios, and different model responses to anthropogenic forcing are ultimately behind a wide range of uncertainties that arise in climate change projections. Model weighting approaches are generally used to reduce the uncertainty related to the choice of the climate model. This study compares three multi-model combination approaches: a simple arithmetic mean and two recently developed weighting-based alternatives. One method takes into account models’ performance only and the other accounts for models’ performance and independence. The effect of these three multi-model approaches is assessed for projected changes of mean precipitation and temperature as well as four extreme indices over northern Morocco. We analyze different widely used high-resolution ensembles issued from statistical (NEXGDDP) and dynamical (Euro-CORDEX and bias-adjusted Euro-CORDEX) downscaling. For the latter, we also investigate the potential added value that bias adjustment may have over the raw dynamical simulations. Results show that model weighting can significantly reduce the spread of the future projections increasing their reliability. Nearly all model ensembles project a significant warming over the studied region (more intense inland than near the coasts), together with longer and more severe dry periods. In most cases, the different weighting methods lead to almost identical spatial patterns of climate change, indicating that the uncertainty due to the choice of multi-model combination strategy is nearly negligible.

AbstractThe Coordinated Regional Climate Downscaling Experiment (CORDEX) initiative has made available an enormous amount of regional climate projections in different domains worldwide. This information is crucial for the development of adaptation strategies and policy‐making. A relevant open issue in this context is assessing the potential multidomain conflicts that may result in overlapping regions and developing appropriate ensemble methods trying to make the most of all available information. This work addresses this timely topic by focusing on precipitation over the Mediterranean region, a first illustrative case study that is encompassed by both the Euro‐ and Africa‐CORDEX domains. We focus on several mean, extreme, and temporal indices and use variance decomposition to assess the separate contribution of the domain and models to the climate change signal, concluding that the contribution of the domain alone is nearly negligible (below in all cases). Nevertheless, for some cases, the combined model/domain effect triggers up to of the total variance.

Abstract. Several sets of reference regions have been proposed in the literature for the regional synthesis of observed and model-projected climate change information. A popular example is the set of reference regions introduced in the IPCC Special Report on Managing the Risks of Extreme Events and Disasters to Advance Climate Adaptation (SREX) based on a prior coarser selection and then slightly modified for the 5th Assessment Report of the IPCC. This set was developed for reporting sub-continental observed and projected changes over a reduced number (33) of climatologically consistent regions encompassing a representative number of grid boxes (the typical resolution of the 5th Climate Model Intercomparison Projection, CMIP5, climate models was around 2º). These regions have been used as the basis for several popular spatially aggregated datasets, such as the seasonal mean temperature and precipitation in IPCC regions for CMIP5. Here we present an updated version of the reference regions for the analysis of new observed and simulated datasets (including CMIP6) which offer an opportunity for refinement due to the higher model resolution (around 1º for CMIP6). As a result, the number of regions increased to 43 land plus 12 open ocean, better representing consistent regional climate features. The paper describes the rationale followed for the definition of the new regions and analyses their homogeneity. The regions are defined as polygons and are provided as coordinates and shapefile together with companion R and Python notebooks to illustrate their use in practical problems (trimming data, etc.). We also describe the generation of a new dataset with monthly temperature and precipitation spatially aggregated in the new regions, currently for CMIP5 (for backwards consistency) and CMIP6, to be extended to other datasets in the future (including observations). The use of these reference regions, dataset and code is illustrated through a worked example using scatter diagrams to offer guidance on the likely range of future climate change at the scale of reference regions. The regions, datasets and code (R and Python notebooks) are freely available at the ATLAS GitHub repository; https://github.com/SantanderMetGroup/ATLAS, doi:10.5281/zenodo.3688072 (Iturbide et al., 2020).

This is the second in a pair of papers in which the performance of statistical downscaling methods (SDMs) is critically reassessed with respect to their robust applicability in climate change studies. Whereas the companion paper focused on temperatures, the present manuscript deals with precipitation and considers an ensemble of 12 SDMs from the analog, weather typing, and regression families. First, the performance of the methods is cross-validated considering reanalysis predictors, screening different geographical domains and predictor sets. Standard accuracy and distributional similarity scores and a test for extrapolation capability are considered. The results are highly dependent on the predictor sets, with optimum configurations including information from midtropospheric humidity. Second, a reduced ensemble of well-performing SDMs is applied to four GCMs to properly assess the uncertainty of downscaled future climate projections. The results are compared with an ensemble of regional climate models (RCMs) produced in the ENSEMBLES project. Generally, the mean signal is similar with both methodologies (with the exception of summer, which is drier for the RCMs) but the uncertainty (spread) is larger for the SDM ensemble. Finally, the spread contribution of the GCM- and SDM-derived components is assessed using a simple analysis of variance previously applied to the RCMs, obtaining larger interaction terms. Results show that the main contributor to the spread is the choice of the GCM, although the SDM dominates the uncertainty in some cases during autumn and summer due to the diverging projections from different families.