Analog-Based Methods

FY2019 Accomplishments

The Analog Ensemble (AnEn)

The analog ensemble has been applied with success to the following applications:

  • Generating 0-240 hour wind and solar power probabilistic forecasts for a site in Kuwait. The AnEn has been coupled with the Shaacke Shuffle technique to generate reliable probabilistic forecasts of aggregated wind+solar power.
Figure 1. Example of AnEn (a, c, e) and AnEn+SS (b, d, f) ensemble forecast for total wind power (a, b), total solar power (c, d) and total wind+solar power (e, f). The shaded area indicates the 10-90th and the 25-75th quantile range and the solid black line the verifying observations
Figure 1. Example of AnEn (a, c, e) and AnEn+SS (b, d, f) ensemble forecast for total wind power (a, b), total solar power (c, d) and total wind+solar power (e, f). The shaded area indicates the 10-90th and the 25-75th quantile range and the solid black line the verifying observations
  • The bias correction technique for rare events has been described in a paper (Alessandrini et al. 2019) and has been included in the official version of the code distributed through the Github platform. In Figure 2, a comparison between AnEn and AnEn with the bias correction is presented in terms of bias as a function of the predicted wind speed.
Figure 2. AnEn (analog ensemble in its original formulation) and AnEnBc (Analog Ensemble with the bias correction algorithm) ensemble mean bias as a function of the wind speed from Global Environmental Multiscale (GEM) model averaged over equally populated bins. 1-year training period and the shorter 9-month training period are used for both AnEn (analog ensemble in its original formulation) and AnEnBc. The error bars indicate the 95% bootstrap confidence intervals.
Figure 2. AnEn (analog ensemble in its original formulation) and AnEnBc (Analog Ensemble with the bias correction algorithm) ensemble-mean bias as a function of the wind speed from Global Environmental Multiscale (GEM) model averaged over equally populated bins. The 1-year training period and the shorter 9-month training period are used for both AnEn (analog ensemble in its original formulation) and AnEnBc. The error bars indicate the 95% bootstrap confidence intervals.

FY2020 PLANS

In FY20 the potential of the analog ensemble technique will be further explored for several applications: 1) improving WRF-CHEM operational air-quality predictions over New Delhi (India); 2) consolidating real-time operational forecasting of tropical cyclones rapid intensification; 3) exploring the use of AnEn to post-process predictions from neural networks or convolutional neural networks systems for tropical cyclone intensity predictions; and 4) improving air-quality predictions during wild fires over the US.  

REFERENCES

Alessandrini, S.; Sperati, S.; Delle Monache, L. Improving the Analog Ensemble Wind Speed Forecasts for Rare Events. Mon. Weather Rev. 2019, 147, 2677–2692.

Alessandrini, S., Delle Monache, L., Rozoff, C.M. and Lewis, W.E., 2018. Probabilistic Prediction of Tropical Cyclone Intensity with an Analog Ensemble. Monthly Weather Review, 146(6), pp.1723-1744

Clark, M.; Gangopadhyay, S.; Hay, L.; Rajagopalan, B.; Wilby, R. The Schaake shuffle: A method for reconstructing space-time variability in forecasted precipitation and temperature fields. J. Hydrometeorol. 2004.

Sperati, S., Alessandrini, S. and Delle Monache, L., 2017. Gridded probabilistic weather forecasts with an analog ensemble. Quarterly Journal of the Royal Meteorological Society, 143(708), pp.2874-2885.