Search

It is often assumed that weather regimes adequately characterize atmospheric circulation variability. However, regime classifications spanning many months and with a low number of regimes may not satisfy this assumption. The first aim of this study is to test such hypothesis for the Euro-Atlantic region. The second one is to extend the assessment of sub-seasonal forecast skill in predicting the frequencies of occurrence of the regimes beyond the winter season. Two regime classifications of four regimes each were obtained from sea level pressure anomalies clustered from October to March and from April to September respectively. Their spatial patterns were compared with those representing the annual cycle. Results highlight that the two regime classifications are able to reproduce most part of the patterns of the annual cycle, except during the transition weeks between the two periods, when patterns of the annual cycle resembling Atlantic Low regime are not also observed in any of the two classifications. Forecast skill of Atlantic Low was found to be similar to that of NAO+, the regime replacing Atlantic Low in the two classifications. Thus, although clustering yearly circulation data in two periods of 6 months each introduces a few deviations from the annual cycle of the regime patterns, it does not negatively affect sub-seasonal forecast skill. Beyond the winter season and the first ten forecast days, sub-seasonal forecasts of ECMWF are still able to achieve weekly frequency correlations of r = 0.5 for some regimes and start dates, including summer ones. ECMWF forecasts beat climatological forecasts in case of long-lasting regime events, and when measured by the fair continuous ranked probability skill score, but not when measured by the Brier skill score. Thus, more efforts have to be done yet in order to achieve minimum skill necessary to develop forecast products based on weather regimes outside winter season. The research leading to these results was funded by the Sub-seasonal to Seasonal climate forecasting for Energy (S2S4E) Project (GA776787), by the New European Wind Atlas (NEWA-II) project (PCIN-2016-029) and by the Ministerio de Economía y Competitividad (MINECO) as part of the CLINSA project (CGL2017-85791-R) and of the Juan de la Cierva-Incorporación Grant (IJCI-2016-29776). We also acknowledge the developers of the s2dverification software package (Manubens et al 2018), used for the data analysis and the visualization of the results presented in this work. The authors would also like to express their gratitude to Pierre-Antoine Bretonniére, Margarida Samsó and Núria Pérez-Zanón for their fundamental contribute in downloading and pre-processing data, and to Dr. Stefano Materia, Dr. Markus Donat, Dr. Simon Wild for their insightful comments and suggestions.

Rainfall variability affects agriculture planning and water resource management. In extreme flood and drought events, lives and property are destroyed. This study aims to improve East Africa’s seasonal rainfall prediction by determining the impact of the standard eight Real-time Multivariate Madden–Julian Oscillation (MJO) (RMM) phases on rainfall and using sea surface temperature (SST) response to test the predictability of the March–May (MAM) and October–December (OND) main rainfall seasons over a period of 33 years (1981–2013). Pearson correlation patterns, composite maps, and regression analyses were applied, and the Brier skill score (BSS) and correlation coefficients (CC) were utilized as validation metrics. Low correspondence of rainfall to MJO 1 and MJO 2 was observed except for the months of November and December. Seasonally, MAM and OND correlation patterns with MJO 2 revealed enhanced rainfall over the highlands and insignificant correspondence over coastal areas. Conversely, enhanced MJO 8 corresponded to suppressed rainfall during the June–August season over the coast and the eastern highlands. MAM rainfall was shown to be predictable using Maritime Continent SST indices, with a BSS of 0.41, while OND rainfall was shown to be predictable using Atlantic and Maritime Continent SSTs with a BSS of 0.62. Positive and negative MJO 2 corresponded, respectively, to enhanced and suppressed rainfall during the OND season and was confirmed to be related to, respectively, a positive and negative Indian Ocean dipole (IOD). An IOD year could possibly be triggered by changes in MJO 2 amplitudes observed as early peaks between February and June.

Accurately predicting drought a few months in advance is important for drought mitigation and agricultural and water resources management, especially for a river basin like that of the Yellow River in North China. However, summer drought predictability over the Yellow River basin is limited because of the low influence from ENSO and the large interannual variations of the East Asian summer monsoon. To explore the drought predictability from an ensemble prediction perspective, 29-year seasonal hindcasts of soil moisture drought, taken directly from several North American multimodel ensemble (NMME) models with different ensemble sizes, were compared with those produced by combining bias-corrected NMME model predictions and variable infiltration capacity (VIC) land surface hydrological model simulations. It was found that the NMME/VIC approach reduced the root-mean-square error from the best NMME raw products by 48% for summer soil moisture drought prediction at the lead-1 season, and increased the correlation significantly. Within the NMME/VIC framework, the multimodel ensemble mean further reduced the error from the best single model by 6%. Compared with the NMME raw forecasts, NMME/VIC had a higher probabilistic drought forecasting skill in terms of a higher Brier skill score and better reliability and resolution of the ensemble. However, the performance of the multimodel grand ensemble was not necessarily better than any single model ensemble, suggesting the need to optimize the ensemble for a more skillful probabilistic drought forecast.

The purpose of this study is the selection of an appropriate coupled general circulation model (CGCM) for seasonal weather prediction of winter temperatures in two cities of Pakistan: Islamabad and Lahore. The data used in the study comprises 32 years (1969–2000) of real-time observations from the Pakistan Meteorological Department (PMD) and simulation data from CGCMs, including the European Centre for Medium Range Weather Forecasts (ECMWF) and the UK Met Office (UKMO). This study attempts to establish accurate average seasonal weather predictions using a single-model Ensemble Prediction System (EPS). Accurate seasonal prediction is important for developing an early warning system. The probabilistic forecast is calculated for the lower tercile (i.e. [0, 0.33]) keeping in view the trend of winter temperature for both cities. The initial conditions (ICs) obtained from the CGCMs have been alternatively processed for better results. The probabilistic forecast obtained from the EPS has been further verified against PMD observations and European Reanalysis Data (ERA-40) using two different measures of probabilistic skill: the Brier Skill Score (BSS) and the Relative Operating Characteristic Skill Score (ROCSS). The ECMWF and UKMO models were used for both of the cities under study. The results demonstrate that the ECMWF model is more suitable for Islamabad, whereas the UKMO model is more suitable for Lahore. The results from these models might be generalised for other areas at the same latitudes with similar topographic and climatic conditions.

Probabilistic forecasts of U.S. Drought Monitor (USDM) intensification over 2-, 4-, and 8-week time periods are developed based on recent anomalies in precipitation, evapotranspiration, and soil moisture. These statistical forecasts are computed using logistic regression with cross validation. While recent precipitation, evapotranspiration, and soil moisture do provide skillful forecasts, it is found that additional information on the current state of the USDM adds significant skill to the forecasts. The USDM state information takes the form of a metric that quantifies the “distance” from the next-higher drought category using a nondiscrete estimate of the current USDM state. This adds skill because USDM states that are close to the next-higher drought category are more likely to intensify than states that are farther from this threshold. The method shows skill over most of the United States but is most skillful over the north-central United States, where the cross-validated Brier skill score averages 0.20 for both 2- and 4-week forecasts. The 8-week forecasts are less skillful in most locations. The 2- and 4-week probabilities have very good reliability. The 8-week probabilities, on the other hand, are noticeably overconfident. For individual drought events, the method shows the most skill when forecasting high-amplitude flash droughts and when large regions of the United States are experiencing intensifying drought.

There have been numerous statistical and dynamical downscaling model comparisons. However, differences in model skill can be distorted by inconsistencies in experimental set-up, inputs and output format. This paper harmonizes such factors when evaluating daily precipitation downscaled over the Iberian Peninsula by the Statistical DownScaling Model (SDSM) and two configurations of the dynamical Weather Research and Forecasting Model (WRF) (one with data assimilation (D) and one without (N)). The ERA-Interim reanalysis at 0.75 degrees resolution provides common inputs for spinning-up and driving the WRF model and calibrating SDSM. WRF runs and SDSM output were evaluated against ECA&D stations, TRMM, GPCP and EOBS gridded precipitation for 2010-2014 using the same suite of diagnostics. Differences between WRF and SDSM are comparable to observational uncertainty, but the relative skill of the downscaling techniques varies with diagnostic. The SDSM ensemble mean, WRF-D and ERAI have similar correlation scores (r=0.45-0.7), but there were large variations amongst SDSM ensemble members (r=0.3-0.6). The best Linear Error in Probability Space (LEPS=0.001-0.007) and simulations of precipitation amount were achieved by individual members of the SDSM ensemble. However, the Brier Skill Score shows these members do not improve the prediction by ERA-Interim, whereas precipitation occurrence is reproduced best by WRF-D. Similar skill was achieved by SDSM when applied to station or gridded precipitation data. Given the greater computational demands of WRF compared with SDSM, clear statements of expected value-added are needed when applying the former to climate impacts and adaptation research. This work was supported by Grants (CGL2013-45198-C2-1-R and CGL2016-76561-R, MINECO/ERDF, UE). S.J.G.R. was supported by a FPI Postdoctoral Research Grant (MINECO, BES-2014-069977). Additional funding was received from the University of the Basque Country (UPV/EHU, GIU 17/002). Intensive computational resources used in the project were provided by I2BASQUE. The authors thank the creators of the WRF/ARW and WRFDA systems for making them freely available to the community. The authors acknowledge the E-OBS dataset from the EU-FP6 project ENSEMBLES and the data providers in the ECA&D project. NOAA_OI_SST_V2 data provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA. Comments by two anonymous reviewers have significantly improved the manuscript.

Most solar flares originate in sunspot groups, where magnetic field changes lead to energy build-up and release. However, few flare-forecasting methods use information of sunspot-group evolution, instead focusing on static point-in-time observations. Here, a new forecast method is presented based upon the 24-hr evolution in McIntosh classification of sunspot groups. Evolution-dependent $\geqslant$C1.0 and $\geqslant$M1.0 flaring rates are found from NOAA-numbered sunspot groups over December 1988 to June 1996 (Solar Cycle 22; SC22) before converting to probabilities assuming Poisson statistics. These flaring probabilities are used to generate operational forecasts for sunspot groups over July 1996 to December 2008 (SC23), with performance studied by verification metrics. Major findings are: i) considering Brier skill score (BSS) for $\geqslant$C1.0 flares, the evolution-dependent McIntosh-Poisson method ($\text{BSS}_{\text{evolution}}=0.09$) performs better than the static McIntosh-Poisson method ($\text{BSS}_{\text{static}} = -0.09$); ii) low BSS values arise partly from both methods over-forecasting SC23 flares from the SC22 rates, symptomatic of $\geqslant$C1.0 rates in SC23 being on average $\approx$80% of those in SC22 (with $\geqslant$M1.0 being $\approx$50%); iii) applying a bias-correction factor to reduce the SC22 rates used in forecasting SC23 flares yields modest improvement in skill relative to climatology for both methods ($\mathrm{BSS}^{\mathrm{corr}}_{\mathrm{static}} = 0.09$ and $\mathrm{BSS}^{\mathrm{corr}}_{\mathrm{evolution}} = 0.20$) and improved forecast reliability diagrams., 21 pages, 9 figures

Cho, M.; Yoon, H.-D.; Do, K.; Son, S., and Kim, I.–H., 2019. Comparative study on the numerical simulation o bathymetric changes under storm condition. In: Lee, J.L.; Yoon, J.-S.; Cho, W.C.; Muin, M., and Lee, J. (eds.), The 3rd International Water Safety Symposium. Journal of Coastal Research, Special Issue No. 91, pp. 106-110. Coconut Creek (Florida), ISSN 0749-0208.Topographic changes in the nearshore mainly occur in strong storm condition. To understand bathymetric changes, attempts were made using several numerical models. Nowadays, Delft3D and XBeach are one of the most popular numerical models for predictions of morphodynamics in the nearshore. In this study, a comparative study of Delft3D and XBeach was performed to understand bathymetric changes under storm condition. The model parameters were calibrated with field observations. Storm surge and storm-induced wave field were re-created to simulate morphodynamics modelling. Bathymetric evolution, as well as shoreline changes were in a good agreement with field observations. BSS (Brier Skill Score) evaluation was applied to compare the accuracy of predictions of both models.

We derive autocorrelation-robust asymptotic variances of the Brier score and Brier skill score, which are generally applicable in circumstances with weak serial correlation. A simulation experiment and an empirical application from macroeconomics underscore the importance of taking care of serial correlation. We find that the conventional variances are too conservative to account for the sampling variability in estimating the Brier (skill) score.

We discuss a novel three-tier hierarchical approach to the validation of an end-to-end seasonal climate forecast system. We present a malaria transmission simulation model (MTSM) driven with output from the DEMETER multi-model seasonal climate predictions, to produce probabilistic hindcasts of malaria prevalence. These prevalence hindcasts are second-tier validated against estimates from the MTSM driven with ERA-40 gridded analyses. The DEMETER–MTSM prevalence hindcasts are shown to be (tier-2) skilful for the one-month lead seasonal predictions as well as for the period covering the seasonal malaria peak with a 4–6 month forecast window for the event prevalence above the median. Interestingly, the tier-2 Brier skill score for the forecast window of the hindcasts starting in February, for the event prevalence above the median, is higher than for either the tier-1 precipitation or temperature forecasts, which were the MTSM driving variables.

The targeting procedure developed at ECMWF is used to make ensembles specially designedfor northern Europe and parts of the north Atlantic Ocean. A total of 35 ensembles are integrated,consisting of 20 winter cases and 15 summer cases in 1997, each consisting of 20 membersplus one control forecast. The ensembles are run up to day 10, and the ensemble spread insidethe target area continues to increase all through the 10 days. Two distinct regimes of increasecan be found, the first increase is consistent with the perturbations moving in and through thetarget area, it is hypothesised that the latter increase in ensemble spread around forecast day 5–7is connected with increasing non-linearity. The performance of the experimental ensembles iscompared to the operational ensemble prediction system (EPS) at ECMWF, both with all 50members and with only 20 members. The spread increases when the number of members in theensemble prediction system is increased, and the spread increases inside the target area whentargeting is applied.We find that the increase in spread when going from EPS with 50 membersto the targeted ensembles is larger than when going from 20 to 50 ensemble members of theoperational sets. Clearly targeting must be an option when predicting for a sub-domain of thehemisphere. Looking at other measures, such as the Brier skill score (BSS), relative operatingcharacteristic (ROC) curves and cost/loss analyses, the impact of the targeting is modest forthe winter cases, but the impact for the summer cases is evident. For the winter cases a largepart of the operational perturbations were located in the same area as the targeted perturbations,and the differences for the two sets are small. For the summer cases the operational perturbationswere mostly split between two locations and hence the targeting will give results differing morefrom the operational. DOI: 10.1034/j.1600-0870.2001.01132.x