Search

Ensemble forecasting has become a standard means to obtain information about forecast uncertainties in meteorological centres across the world. The large datasets generated by ensemble prediction systems carry much information that is difficult to analyse manually – here, techniques from the field of artificial intelligence can be beneficial to aid the analysis. Cluster analysis is one commonly used (unsupervised machine learning) approach to automatically determine distinct scenarios in numerical weather forecasting ensembles, both in atmospheric research and operational forecasting. Typically, a cluster analysis focusses on a selected meteorological forecast variable, a specific region, and time (or a time window). The dimensionality of the data is reduced by techniques like principal component analysis, and a clustering algorithm – typically k-means – is applied to the reduced data set. Challenges with such an approach arise through the determined clusters often being sensitive to factors including the selected region, forecast variable, and algorithm parameters, and also through the employed algorithms often appearing as a “black box” to the user. In our work, we attempt to make the clustering process more transparent by providing a visual analysis framework to analyse the sensitivity of generated clusters with respect to various factors. The presented framework is coupled to the open-source meteorological ensemble visualization software Met.3D, allowing for interactive specification of clustering parameters and for interactive visual analysis, including 3-D elements. A case study using ensemble prediction data of sudden stratospheric warmings (SSWs) is presented, demonstrating how visualizing similarity between clusterings with different parameters can aid the interpretation of the data.

Quantifying signals and uncertainties in climate models is essential for the detection, attribution, prediction and projection of climate change1,2,3. Although inter-model agreement is high for large-scale temperature signals, dynamical changes in atmospheric circulation are very uncertain4. This leads to low confidence in regional projections, especially for precipitation, over the coming decades5,6. The chaotic nature of the climate system7,8,9 may also mean that signal uncertainties are largely irreducible. However, climate projections are difficult to verify until further observations become available. Here we assess retrospective climate model predictions of the past six decades and show that decadal variations in North Atlantic winter climate are highly predictable, despite a lack of agreement between individual model simulations and the poor predictive ability of raw model outputs. Crucially, current models underestimate the predictable signal (the predictable fraction of the total variability) of the North Atlantic Oscillation (the leading mode of variability in North Atlantic atmospheric circulation) by an order of magnitude. Consequently, compared to perfect models, 100 times as many ensemble members are needed in current models to extract this signal, and its effects on the climate are underestimated relative to other factors. To address these limitations, we implement a two-stage post-processing technique. We first adjust the variance of the ensemble-mean North Atlantic Oscillation forecast to match the observed variance of the predictable signal. We then select and use only the ensemble members with a North Atlantic Oscillation sufficiently close to the variance-adjusted ensemble-mean forecast North Atlantic Oscillation. This approach greatly improves decadal predictions of winter climate for Europe and eastern North America. Predictions of Atlantic multidecadal variability are also improved, suggesting that the North Atlantic Oscillation is not driven solely by Atlantic multidecadal variability. Our results highlight the need to understand why the signal-to-noise ratio is too small in current climate models10, and the extent to which correcting this model error would reduce uncertainties in regional climate change projections on timescales beyond a decade. DMS, AAS, NJD, LH and RE were supported by the Met Office Hadley Centre Climate Programme funded by BEIS and Defra and by the European Commission Horizon 2020 EUCP project (GA 776613). FJDR, LPC, SW and RB also acknowledge the support from the EUCP project (GA 776613) and from the Ministerio de Econom´ıa y Competitividad (MINECO) as part of the CLINSA project (Grant No. CGL2017-85791-R). SW received funding from the innovation programme under the Marie Sk´lodowska-Curie grant agreement H2020-MSCA-COFUND-2016-754433 and PO from the Ramon y Cajal senior tenure programme of MINECO. The EC-Earth simulations were performed on Marenostrum 4 (hosted by the Barcelona Supercomputing Center, Spain) using Auto-Submit through computing hours provided by PRACE.WAM, HP, KMand KP were supported by the German FederalMinistry for Education and Research (BMBF) project MiKlip (grant 01LP1519A). NK, IB, FC and YW were supported by the Norwegian Research Council projects SFE (grant 270733) the Nordic Center of excellent ARCPATH (grant 76654) and the Trond Mohn Foundation, under the project number : BFS2018TMT01 and received grants for computer time from the Norwegian Program for supercomputing (NOTUR2, NN9039K) and storage grants (NORSTORE, NS9039K). JM, LFB and DS are supported by Blue-Action (European Union Horizon 2020 research and innovation program, Grant Number: 727852) and EUCP (European Union Horizon 2020 research and innovation programme under grant agreement no 776613) projects. The National Center for Atmospheric Research (NCAR) is a major facility sponsored by the US National Science Foundation (NSF) under Cooperative Agreement No. 1852977. NCAR contribution was partially supported by the National Oceanic and Atmospheric Administration (NOAA) Climate Program Office under Climate Variability and Predictability Program Grant NA13OAR4310138 and by the US NSF Collaborative Research EaSM2 Grant OCE-1243015.

Ensemble forecasting has become a standard practice in numerical weather prediction in forecasting centres across the world. The large data sets generated by ensemble forecasting systems carry much information, that is difficult to analyse in short time periods, requiring well-designed workflows in order to be useful. Clustering is one of the ensemble analysis methods that are applied to discover similarities between ensemble members. Cluster analysis involves different steps like dimensionality reduction, core clustering algorithm and evaluation. A large of number of methods have been proposed in the literature for each of these steps, however, only few have been applied to clustering of ensemble forecasts. A major challenge is that for a given ensemble forecast, different choices of methods and data domains can lead to very different clustering results. For example, Kumpf et al. (2018, IEEE Transact. Vis. Comp. Graph.) have demonstrated the sensitivity of clustering results to even small changes in the considered domain. The challenge equally exists for choices in clustering methods and method parameters.In our work, we are attempting to open up the clustering black box by introducing a visualization workflow that makes transparent to the user how different choices in methods and method parameters lead to different clustering results. To achieve this, a clustering analysis library that works in tandem with the ensemble visualization software “Met.3D” () is being developed. We present the current state of the system and demonstrate its use by analysing an ensemble forecast case study.

The Greenland Ice Sheet (GrIS) mass loss has been accelerating at a rate of about 20 ± 10 Gt/yr2 since the end of the 1990's, with around 60 % of this mass loss directly attributed to enhanced surface meltwater runoff. However, in the climate and glaciology communities, different approaches exist on how to model the different surface mass balance (SMB) components using: (1) complex physically-based climate models which are computationally expensive; (2) intermediate complexity energy balance models; (3) simple and fast positive degree day models which base their inferences on statistical principles and are computationally highly efficient. Additionally, many of these models compute the SMB components based on different spatial and temporal resolutions, with different forcing fields as well as different ice sheet topographies and extents, making inter-comparison difficult. In the GrIS SMB model intercomparison project (GrSMBMIP) we address these issues by forcing each model with the same data (i.e., the ERA-Interim reanalysis) except for two global models for which this forcing is limited to the oceanic conditions, and at the same time by interpolating all modelled results onto a common ice sheet mask at 1 km horizontal resolution for the common period 1980–2012. The SMB outputs from 13 models are then compared over the GrIS to (1) SMB estimates using a combination of gravimetric remote sensing data from GRACE and measured ice discharge, (2) ice cores, snow pits, in-situ SMB observations, and (3) remotely sensed bare ice extent from MODerate-resolution Imaging Spectroradiometer (MODIS). Our results reveal that the mean GrIS SMB of all 13 models has been positive between 1980 and 2012 with an average of 340 ± Gt/yr, but has decreased at an average rate of −7.3 Gt/yr2 (with a significance of 96 %), mainly driven by an increase of 8.0 Gt/yr2 (with a significance of 98 %) in meltwater runoff. Spatially, the largest spread among models can be found around the margins of the ice sheet, highlighting the need for accurate representation of the GrIS ablation zone extent and processes driving the surface melt. In addition, a higher density of in-situ SMB observations is required, especially in the south-east accumulation zone, where the model spread can reach 2 mWE/yr due to large discrepancies in modelled snowfall accumulation. Overall, polar regional climate models (RCMs) perform the best compared to observations, in particular for simulating precipitation patterns. However, other simpler and faster models have biases of same order than RCMs with observations and remain then useful tools for long-term simulations. Finally, it is interesting to note that the ensemble mean of the 13 models produces the best estimate of the present day SMB relative to observations, suggesting that biases are not systematic among models.

We analyze the quasi‐biennial oscillation (QBO) variability of historical and decadal hindcast simulations of the MiKlip (Mittelfristige Klimavorhersagen) decadal prediction system using the higher resolved version of the Max Planck Institute Earth System Model. We find a realistic variability of the QBO in historical simulations when changing from the Coupled Model Intercomparison Project Phase 5 (CMIP5) to the Coupled Model Intercomparison Project Phase 6 (CMIP6) external forcing. This agreement between the simulated and the observed QBO is improved by the initialization of decadal hindcast simulations with CMIP6 forcing in the first three lead years. In the decadal hindcast simulations, the agreement is similar to a persistence forecast in the first five lead years and higher than the persistence forecast in the later lead years. We find a strong relation between the QBO and the ozone variability in the stratosphere and conclude that the change of the ozone data from CMIP5 to CMIP6 leads to the improved QBO variability and prediction skill in our simulations.

Multi-year droughts strongly impact food production and water management. Thus, predictions for the next decade are required for decision makers. This study analyzes the decadal prediction skill of the Global Precipitation Climatology Centre Drought Index (GPCC‑DI) and its components, namely the Standardized Precipitation Index (SPI‑DWD) adapted by the German Meteorological Service (Deutscher Wetterdienst, DWD) and the Standardized Precipitation Evapotranspiration Index (SPEI) within the German global decadal prediction system. The decadal predictions are recalibrated. The prediction skills of the two prediction types ensemble mean predictions and probabilistic predictions are evaluated against those of the commonly applied reference predictions observed climatology and uninitialized simulations. The evaluation of 4‑year mean droughts for the lead-year period 1–4 at 5° spatial resolution shows high prediction skills for the SPEI in the tropics, especially northern Africa, and several heterogeneously distributed hot spots for the SPI‑DWD. The advantage of GPCC‑DI is its global coverage, but it hardly enhances the SPI‑DWD and SPEI skills. The recalibration clearly enhances ensemble mean prediction skills in slightly improving correlations and in strongly reducing standard deviations as well as large conditional biases in decadal predictions. For probabilistic predictions, impacts of conditional biases and recalibration are less prominent. To meet user requirements decadal drought predictions with higher temporal and spatial resolutions are analyzed. 1‑year mean droughts for lead year 1 mostly show smaller prediction skills than 4‑year means because of larger small-scale noise, but some regions reveal improved skills due to regional processes predictable at the 1‑year time scale, e.g. over the western United States. Drought predictions at 2° resolution show similar spatial skill patterns with enhanced fine-scale structures mostly without losing prediction skill. A user-oriented evaluation of the decadal GPCC‑DI prediction for the severe North African drought of 2008–2011 reproduces most observed drought index tendencies in both prediction types, but intensities are often underestimated. Finally, the decadal GPCC‑DI prediction for 2018–2021 presents a drought over North Africa and Arabia and wetting over the Northern Hemisphere in both prediction types. For 2018, predicted patterns are similar but with smoothed intensities. In summary, decadal drought prediction skill depends on the indices, time periods, and areas considered. However, the analyzed drought indices can provide skillful high-resolution information for several future time periods and regions meeting user needs for decadal drought predictions.

International audience; the accumulation of magma within the Monti Sabatini Volcanic District (MSVD), italy, coupled with the extensional tectonics of the region, pose both volcanic and tectonic hazards to the city of Rome, located 20 km to the southeast. We combine 40 Ar/ 39 Ar geochronology of volcanic deposits and a geomorphologic/stratigraphic/paleomagnetic study of fluvial terraces to determine the recurrence interval and the time elapsed since the last eruption of the MSVD. Moreover, we provide a date for the youngest known eruption of the MSVD and assess the timing of the most recent volcanic phase. Results of this study show: (i) The most recent eruptive phase occurred between 100 ka and 70 ka; (ii) the anomalously high elevation of the MIS 5 terrace indicates that it was concurrent with 50 m of uplift in the volcanic area; (iii) the time since the last eruption (70 ka) exceeds the average recurrence interval (39 ky) in the last 300 ky, as well as the longest previous dormancy (50 ky) in that time span. (iv) the current duration of dormancy is similar to the timespan separating the major explosive phase that occurred 590-450 ka. The magnitude and patterns of deformation of a volcanic field provide constraints on the magmatic processes operating beneath a volcano. A recent geomorphological study 1 reconstructed a series of paleo-surfaces along a 40-km-long stretch of the Tiber River Valley north of Rome, between Magliano Sabina and Monterotondo, located at the eastern margin of the Monti Sabatini Volcanic District (MSVD) (Fig. 1a). These paleo-surfaces are interpreted as fluvial terraces formed through the interplay between regional uplift and glacio-eustasy (e.g. 2) during a regressive phase that formed the hydrographic network of the Paleo-Tiber River since the end of the Santernian (1.78-1.5 Ma; lower Calabrian) 3. The reconstructed rates of uplift during the last 1.8 Ma recognized two major pulses: 0.86 through 0.5 Ma, and 0.25 Ma through the present time 1. The coincidence of the uplift pulses with the ages of the main volcanic phases 1 are interpreted as mainly driven by uprising magma bodies from a metasomatized mantle source of the Roman Magmatic Province (e.g. 4-6). This "magmatic" uplift overlaps a smaller isostatic component on the Tyrrhenian Sea Margin of central Italy. In the present study, we further refine the chronostratigraphy of the paleo-surfaces in the area previously investigated, providing five new 40 Ar/ 39 Ar age determinations on volcanic products intercalated within the

Our decadal climate prediction system, which is based on the Max-Planck-Institute Earth System Model, is initialized from a coupled assimilation run that utilizes nudging to selected state parameters from reanalyses. We apply full-field nudging in the atmosphere and either full-field or anomaly nudging in the ocean. Full fields from two different ocean reanalyses are considered. This comparison of initialization strategies focuses on the North Atlantic Subpolar Gyre (SPG) region, where the transition from anomaly to full-field nudging reveals large differences in prediction skill for sea surface temperature and ocean heat content (OHC). We show that nudging of temperature and salinity in the ocean modifies OHC and also induces changes in mass and heat transports associated with the ocean flow. In the SPG region, the assimilated OHC signal resembles well OHC from observations, regardless of using full fields or anomalies. The resulting ocean transport, on the other hand, reveals considerable differences between full-field and anomaly nudging. In all assimilation runs, ocean heat transport together with net heat exchange at the surface does not correspond to OHC tendencies, the SPG heat budget is not closed. Discrepancies in the budget in the cases of full-field nudging exceed those in the case of anomaly nudging by a factor of 2–3. The nudging-induced changes in ocean transport continue to be present in the free running hindcasts for up to 5 years, a clear expression of memory in our coupled system. In hindcast mode, on annual to inter-annual scales, ocean heat transport is the dominant driver of SPG OHC. Thus, we ascribe a significant reduction in OHC prediction skill when using full-field instead of anomaly initialization to an initialization shock resulting from the poor initialization of the ocean flow.

The MPI‐ESM1.2 is the latest version of the Max Planck Institute Earth System Model and is the baseline for the Coupled Model Intercomparison Project Phase 6 and current seasonal and decadal climate predictions. This paper evaluates a coupled higher‐resolution version (MPI‐ESM1.2‐HR) in comparison with its lower‐resolved version (MPI‐ESM1.2‐LR). We focus on basic oceanic and atmospheric mean states and selected modes of variability, the El Niño/Southern Oscillation and the North Atlantic Oscillation. The increase in atmospheric resolution in MPI‐ESM1.2‐HR reduces the biases of upper‐level zonal wind and atmospheric jet stream position in the northern extratropics. This results in a decrease of the storm track bias over the northern North Atlantic, for both winter and summer season. The blocking frequency over the European region is improved in summer, and North Atlantic Oscillation and related storm track variations improve in winter. Stable Atlantic meridional overturning circulations are found with magnitudes of ~16 Sv for MPI‐ESM1.2‐HR and ~20 Sv for MPI‐ESM1.2‐LR at 26°N. A strong sea surface temperature bias of ~5°C along with a too zonal North Atlantic current is present in both versions. The sea surface temperature bias in the eastern tropical Atlantic is reduced by ~1°C due to higher‐resolved orography in MPI‐ESM‐HR, and the region of the cold‐tongue bias is reduced in the tropical Pacific. MPI‐ESM1.2‐HR has a well‐balanced radiation budget and its climate sensitivity is explicitly tuned to 3 K. Although the obtained reductions in long‐standing biases are modest, the improvements in atmospheric dynamics make this model well suited for prediction and impact studies.

A new release of the Max Planck Institute for Meteorology Earth System Model version 1.2 (MPI‐ESM1.2) is presented. The development focused on correcting errors in and improving the physical processes representation, as well as improving the computational performance, versatility, and overall user friendliness. In addition to new radiation and aerosol parameterizations of the atmosphere, several relatively large, but partly compensating, coding errors in the model's cloud, convection, and turbulence parameterizations were corrected. The representation of land processes was refined by introducing a multilayer soil hydrology scheme, extending the land biogeochemistry to include the nitrogen cycle, replacing the soil and litter decomposition model and improving the representation of wildfires. The ocean biogeochemistry now represents cyanobacteria prognostically in order to capture the response of nitrogen fixation to changing climate conditions and further includes improved detritus settling and numerous other refinements. As something new, in addition to limiting drift and minimizing certain biases, the instrumental record warming was explicitly taken into account during the tuning process. To this end, a very high climate sensitivity of around 7 K caused by low‐level clouds in the tropics as found in an intermediate model version was addressed, as it was not deemed possible to match observed warming otherwise. As a result, the model has a climate sensitivity to a doubling of CO2 over preindustrial conditions of 2.77 K, maintaining the previously identified highly nonlinear global mean response to increasing CO2 forcing, which nonetheless can be represented by a simple two‐layer model., Key Points An updated version of the Max Planck Institute for Meteorology Earth System Model (MPI‐ESM1.2) is presentedThe model includes both code corrections and parameterization improvementsDespite this, the model maintains an equilibrium climate sensitivity, which rises with warming

Decadal climate predictions use initialized coupled model simulations that are typically affected by a drift toward a biased climatology determined by systematic model errors. Model drifts thus reflect a fundamental source of uncertainty in decadal climate predictions. However, their analysis has so far relied on ad-hoc assessments of empirical and subjective character. Here, we define the climate model drift as a dynamical process rather than a descriptive diagnostic. A unified statistical Bayesian framework is proposed where a state-space model is used to decompose systematic decadal climate prediction errors into an initial drift, seasonally varying climatological biases and additional effects of co-varying climate processes. An application to tropical and south Atlantic sea-surface temperatures illustrates how the method allows to evaluate and elucidate dynamic interdependencies between drift, biases, hindcast residuals and background climate. Our approach thus offers a methodology for objective, quantitative and explanatory error estimation in climate predictions.

Predictive skill of North Atlantic upper-ocean salinity has, in contrast to upper-ocean temperature, so far not received much attention, despite recent evidence that predictability of abundance and distribution of marine ecosystem species can be inferred from salinity predictions. Here, we use initialized decadal prediction experiments with six climate models to assess the predictive skill of North Atlantic upper-ocean salt content. Based on the multi-model ensemble mean, we demonstrate decadal-scale predictive in the entire subpolar North Atlantic and the eastern part of the Nordic Seas. Based on our own decadal prediction experiments, we attribute the skill at longer lead times to a delayed response to the initialization of the North Atlantic ocean circulation. &nbsp