Simultaneous Ensemble Postprocessing for Multiple Lead Times with Standardized Anomalies

Abstract

Separate statistical models are typically fit for each forecasting lead time to postprocess numerical weather prediction (NWP) ensemble forecasts. Using standardized anomalies of both NWP values and observations eliminates most of the lead-time-specific characteristics so that several lead times can be forecast simultaneously. Standardized anomalies are formed by subtracting a climatological mean and dividing by the climatological standard deviation. Simultaneously postprocessing forecasts between +12 and +120 h increases forecast coherence between lead times, yields a temporal resolution as high as the observation interval (e.g., up to 10 min), and speeds up computation times while achieving a forecast skill comparable to the conventional method.

Compound impact of land use and extreme climate on the 2020 fire record of the Brazilian Pantanal

“…The high correlation coefficients in this case reflect the coincident patterns of precipitation and SST seasonality instead of a direct influence of SST on precipitation anomalies. To overcome this problem, our analysis used standardized anomalies (Dabernig et al, 2017) and did not detect the hypothesized relationship between precipitation anomalies and SSTs. Marengo et al (2021) also corroborated that correlations between precipitation and oceanic indices were very low, although these authors made a visual interpretation rather than a statistical analysis.…”

“…Precipitation in the Pantanal is also influenced by the atmospheric moisture carried from the Amazonian rain forest, especially the summer precipitation (Bergier et al, 2018). Lastly, some studies discuss the role of sea surface temperature (SST) in precipitation in the Pantanal (Dabernig et al, 2017; Marengo et al, 2021; Thielen et al, 2020), although more research is needed to improve our knowledge, because there are uncertainties on the topic. Although precipitation trends cannot yet be established for the Pantanal (Marcuzzo et al, 2010; Bergier et al, 2018), the mass of water has undoubtedly been reducing over the course of the last 35 years (Mapbiomas, 2021), which can favour fire occurrence.…”

Probabilistic prediction of the time to hard freeze using seasonal weather forecasts and survival time methods

“…NWPs are often subject to systematic biases. This is especially true for near‐surface temperature forecasts over complex terrain, where differences between the model grid elevation and the true elevation entail systematic differences in the respective near‐surface temperature climatologies (Dabernig et al ., 2017; Keller et al ., 2021). Statistical postprocessing can remove these biases, and thus, in combination with a high‐resolution observation product, downscale the forecasts to that resolution.…”

Lead‐time‐continuous statistical postprocessing of ensemble weather forecasts

“…Most closely related to the present study is the work by Dabernig et al (2017a), who calculate standardized anomalies accounting for seasonality and diurnal cycle via bivariate splines and define EMOS models between the ensemble and observational anomalies. However, we believe that this approach is rather restrictive and that lead-time dependence as well as the diurnal cycle can be accounted for directly in the postprocessing model itself.…”