The Unusual Nature of West Coast NCDC Climate Divisions’ Seasonal Precipitation Anomaly Patterns Relative to ENSO Phase for the 2015−16 and 2016−17 Seasons, as Compared to Long-Term 1895−96 to 2014−15 Climatological Tendencies

In a recent study (Fisk, 2016), the existence and character of statistical anomaly modes in July-June total precipitation were investigated for 12 California, Oregon, and Washington NCDC Climate Divisions located near the Pacific Ocean west of the Cascade Mountains. Period of record considered was the 1895-96 thru 2014-15 seasons (n=120). The K-Means Clustering methodology, integrated with the V-Fold Cross Validation Algorithm, an iterative training sample type procedure which optimizes the number of clusters resolved was utilized to identify the mean (standardized) anomaly patterns by cluster, by division. Results produced seven clusters. Then, through referencing and manipulation of the bi-monthly statistics from the Multivariate ENSO (MEI) Index (Wolter, 2015), a hierarchy of ENSO designations based on mean July-June bi-monthly rankings was created: “Strong La Nina”, “La Nina”, “Neutral”, “Weak El Nino”, “Moderate El Nino”, and “Strong El Nino”. These designations were then matched up in cross-tabulation fashion with those of the clusters, and from further manipulation of the 42 cross-tab frequencies, a table of Bayesian probabilities was created which estimated the conditional probabilities of given cluster anomaly patterns being realized, given the particular ENSO designation in place. In addition to providing conditional probabilities of expected precipitation regime types up and down the near Pacific Coast, the probabilities could also be useful in further assessing, after the fact, the anomaly character of the seasonal precipitation pattern relative to ENSO expectations.

The 2015-16 El Nino was generally regarded as one of the three strongest in history (it was still in process at the time of the 2016 study so the seasonal precipitation figures were not yet available) and the Bayesian probabilities, in particular, gave a greater than 80 percent chance that Southern California would be wet or very wet for the rain season, but instead, drought prevailed, extending the multi-year pattern to its fifth consecutive season. In contrast, Oregon and Washington were quite wet. Inspecting the cluster designation framework, the 2015-16 pattern resembled most closely the “Oregon & Washington Exceptionally Wet, California South Coast Dry” and “Dry South and Central California and Wet Oregon and Washington” modes, but the “Strong El Nino” Bayesian probabilities in the table for these two were each zero! Indeed, the 2015-16 El Nino behaved more like a Strong La Nina than a strong El Nino, a result not experienced over the entire NCDC Climate division period of record, and probably at least further back thru 1871-72 which is the oldest season with MEI indices.

The 2016-17 season, a Neutral or La Nina for much of its duration, brought heavy rains up and down the coast, including (unexpectedly) Southern California – the long drought pattern essentially ending. Of the seven idealized modes, the closest affinity was to the “Oregon & Washington Exceptionally Wet, California South Coast Dry” pattern, one with conditional probabilities of 7.1% for La Nina’s 2.5% for Neutrals); yet in the case of 2016-17, all of the twelve divisions had above average July-June rainfall, a rare occurrence for the near Pacific Coast. In fact, the lowest standardized anomaly of the twelve, +0.59z, is the highest such figure of the entire 122-year history.

The 2016 clustering and Bayesian table results are revisted, with the 2015-16 and 2016-17 results described in their context, along with supplementary graphs and tables