The Indian Ocean has been suggested to play a role in mediating the ENSO precipitation teleconnection to North America in certain years 16. Studies using both model simulations and observations have also shown that tropical diabatic heating anomalies in key regions across the western tropical Pacific, at times independent from ENSO, substantially increase the likelihood of ridging and subsequently drought conditions across California 14, 15. Other studies have shown that traditional indices for describing ENSO variability (i.e., Niño3.4) may not be optimal for capturing the teleconnection to western US precipitation 13. The El Niño Southern Oscillation (ENSO) is known to be the primary driver of seasonal forecast skill across North America 7, 8, 9, 10, yet its signal-to-noise ratio is such that unexpected outcomes will occasionally occur by chance 4, 11, 12. In a seasonal forecasting context, teleconnections are best viewed as probabilistically loading the dice in favor of a certain outcome (i.e., dry versus wet conditions). Given that the economic costs of severe drought can frequently exceed $1B annually across California 5, 6, improving the skill of seasonal precipitation forecasts remains a top priority for water resource managers. As widely documented, the expected positive anomaly of precipitation across California and the Southwest under the major El Niño event of 2015/2016 did not eventuate as anticipated, and instead the devastating drought continued 4. ![]() During the recent severe California drought (years 2012–2016), the challenges for decision-makers under forecast uncertainty were highlighted. Individually, these storms have proven challenging to forecast at lead times beyond the weather time horizon 2, 3. Relatively low precipitation totals combined with high year-to-year variability are often received in the form of a relatively small number of atmospheric rivers across winter months 1. The climatology and variability of precipitation across the western United States present a unique seasonal forecasting challenge. We further show that this approach need not be considered a ‘black box’ by utilizing machine learning interpretability methods to identify the relevant physical processes that lead to prediction skill. For forecasting large-scale spatial patterns of precipitation across the western United States, here we show that these machine learning-based models are capable of competing with or outperforming existing dynamical models from the North American Multi Model Ensemble. After training on thousands of seasons of climate model simulations, the machine learning models are tested for producing seasonal forecasts across the historical observational period (1980-2020). To circumvent this issue, here we explore the feasibility of training various machine learning approaches on a large climate model ensemble, providing a long training set with physically consistent model realizations. The mean annual total precipitation and snowfall maps on this plate are primarily based on thirty-year data during the period 1921 to 1950 inclusive.A barrier to utilizing machine learning in seasonal forecasting applications is the limited sample size of observational data for model training. ![]() A specific gravity of 0.1 for freshly fallen snow is used, which means that ten inches (25.4 cm) of freshly fallen snow is assumed to be equal to one inch (2.54 cm) of rain. Annual precipitation is defined as the sum of rainfall and the assumed water equivalent of snowfall for a given year. Mean Annual Total Precipitation Contained within the 3rd Edition (1957) of the Atlas of Canada is a plate that shows two maps for the annual total precipitation.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |