S2S International Weekly Machine Learning Forecasts: Data & Documentation

The Famine Early Warning Systems Network, FEWS NET, provides objective, evidence-based analyses to help government decision-makers and relief agencies plan for and respond to humanitarian crises as a result of acute food insecurity. NOAA environmental monitoring and forecasting tools are being used to communicate likely weather and climate conditions to shape FEWS NET food security outlooks.

This Product

The Physical Sciences Lab provides experimental precipitation and 2-meter temperature forecasts to support agricultural climatology assumptions used in food security scenario development. These forecasts are based on a machine learning approach using a linear inverse model (LIM). LIM relies on observed relationships between model variables to infer predictable Earth system processes and generate probabilistic forecasts.

Method

LIM is trained using daily differences from average with a 7-day running mean applied, and the current and following month's covariance statistics to train the model using the years 1981-2020. Similar to output from numerical forecast models, for each initialization and lead time, the LIM generates forecasts of temperature and precipitation by propagating the initial conditions forward in time. The forecast is "blended", meaning that LIMs for the month prior and month after are also calculated, and then weighted based on the date of the current month. For example, a forecast initialized on 31 January would have a stronger weighting for the February LIM forecast than the December LIM forecast. Blending is done to improve the reliability of the forecasts and minimize large jumps in the forecasts that can occur transitioning from one month's LIM to the other. Near-real-time forecasts are created using preliminary CHIRPS (Funk et al. 2015) and JRA-55 (Kobayashi et al. 2015) data, with forecasts available at a 3-day lag.

Data Source

All variables except precipitation are from JRA-55 data. Precipitation from the Climate Hazards InfraRed Precipitation with Stations (CHIRPS) dataset is used.


Southwest Asia LIM

One important way to check how well a model performs is to have it “predict” things that we already know — this is referred to as hindcasting. The southwest Asia Jan-Feb-Mar LIM hindcast skill (shown in the figure below) uses a method for assessing skill called anomaly correlation coefficient (ACC), which is comparable to or higher than other subseasonal models. The LIM can also identify periods of elevated forecast skill at the time of forecast, which is achieved using a metric called expected skill. This approach identifies more skillful forecasts (panel b) than those identified using other methods such as the Niño3.4 index and RMM index.

Domains for Each Southwest Asia LIM Variable
Precipitation (22-48 N, 50-80 W)
2-m Temperature (0-50N, 0-120W)
200-hPa Streamfunction (0-90 N, 0-360 W)
Tropical Heating (20 S – 20 N, 0-360 W)
Tropical SST (20 S – 20 N, 0-360 W)
Four maps
ACC for weeks 3-4 forecasts, evaluated from January – March 1982-2020. Panel a) shows ACC for all dates in the record, b) ACC for the 20% of forecasts initialized with the highest expected skill, c) ACC for the 20% with the highest Niño3.4 amplitude, and d) ACC for the 20% with the highest RMM amplitude. The black stippling indicates where the skill of the top 20% of forecasts in each group is statistically significantly different from the skill of the remaining 80% of forecasts at the 95% confidence level.

Breeden, M. L., Albers, J. R., A. Hoell: Subseasonal precipitation forecasts of opportunity over southwest Asia, Weather and Climate Dynamics Discussions, https://doi.org/10.5194/egusphere-2022-555, 2022.

de Andrade, F.M., Coelho, C.A.S. and Cavalcanti, I.F.A.: Global precipitation hindcast quality assessment of the Subseasonal to Seasonal (S2S) prediction project models. Clim. Dyn., 52, 5451–5475. https://doi.org/10.1007/s00382-018-4457-z, 2018.

Funk, C., Peterson, P., Landsfeld, M. Pedreros, D., Verdin, J., Shukla, S., Husak, G., Rowland, J., Harrison, L., Hoell, A., and J. Michaelsen: The climate hazards infrared precipitation with stations—a new environmental record for monitoring extremes. Sci. Data, 2, 150066, https://doi.org/10.1038/sdata.2015.66, 2015.

Kobayashi, S., Ota, Y., Harada, Y., Ebita, A., Moriya, M., Onoda, H., Onogi, Kamahori, H., Kobayashi, C., Hirokazu, E., 605 Miyaoka, K., and Takahashi, K.: The JRA‐55 reanalysis: General specifications and basic characteristics. Journal of the Meteorological Society of Japan, Ser. II, 93(1), 5–48. https://doi.org/10.2151/jmsj.2015‐001, 2015.

Penland, C. and Sardeshmukh, P. D.: The optimal growth of tropical sea surface temperature anomalies J. Clim., 8, 1999–2024, 1995.

Pegion, K., Kirtman, B. P., Becker, E., Collins, D. C., LaJoie, E., Burgman, R., Bell, R., DelSole, T., Min, D., Zhu, Y., Li, W., Sinsky, E., Guan, H., Gottschalck, J., Metzger, E. J., Barton, N. P., Achuthavarier, D., Marshak, J., Koster, R. D., Lin, H., Gagnon, N., Bell, M., Tippett, M. K., Robertson, A. W., Sun, S., Benjamin, S. G., Green, B. W., Bleck, R., and Kim, H.: The Subseasonal Experiment (SubX): A Multimodel Subseasonal Prediction Experiment, Bulletin of the American Meteorological Society, 100(10), 2043-2060, 2019.