RESUMEN
West Nile virus (WNV) is the leading cause of mosquito-borne illness in the continental United States (CONUS). Spatial heterogeneity in historical incidence, environmental factors, and complex ecology make prediction of spatiotemporal variation in WNV transmission challenging. Machine learning provides promising tools for identification of important variables in such situations. To predict annual WNV neuroinvasive disease (WNND) cases in CONUS (2015-2021), we fitted 10 probabilistic models with variation in complexity from naïve to machine learning algorithm and an ensemble. We made predictions in each of nine climate regions on a hexagonal grid and evaluated each model's predictive accuracy. Using the machine learning models (random forest and neural network), we identified the relative importance and variation in ranking of predictors (historical WNND cases, climate anomalies, human demographics, and land use) across regions. We found that historical WNND cases and population density were among the most important factors while anomalies in temperature and precipitation often had relatively low importance. While the relative performance of each model varied across climatic regions, the magnitude of difference between models was small. All models except the naïve model had non-significant differences in performance relative to the baseline model (negative binomial model fit per hexagon). No model, including the ensemble or more complex machine learning models, outperformed models based on historical case counts on the hexagon or region level; these models are good forecasting benchmarks. Further work is needed to assess if predictive capacity can be improved beyond that of these historical baselines.
RESUMEN
Monthlong hindcasts of the Madden-Julian oscillation (MJO) from the atmospheric Flow-following Icosahedral Model coupled with an icosahedral-grid version of the Hybrid Coordinate Ocean Model (FIM-iHYCOM), and from the coupled Climate Forecast System, version 2 (CFSv2), are evaluated over the 12-yr period 1999-2010. Two sets of FIM-iHYCOM hindcasts are run to test the impact of using Grell-Freitas (FIM-CGF) versus simplified Arakawa-Schubert (FIM-SAS) deep convection parameterizations. Each hindcast set consists of four time-lagged ensemble members initialized weekly every 6 h from 1200 UTC Tuesday to 0600 UTC Wednesday. The ensemble means of FIM-CGF, FIM-SAS, and CFSv2 produce skillful forecasts of a variant of the Real-time Multivariate MJO (RMM) index out to 19, 17, and 17 days, respectively; this is consistent with FIM-CGF having the lowest root-mean-square errors (RMSEs) for zonal winds at both 850 and 200 hPa. FIM-CGF and CFSv2 exhibit similar RMSEs in RMM, and their multimodel ensemble mean extends skillful RMM prediction out to 21 days. Conversely, adding FIM-SAS-with much higher RMSEs-to CFSv2 (as a multimodel ensemble) or FIM-CGF (as a multiphysics ensemble) yields either little benefit, or even a degradation, compared to the better single-model ensemble mean. This suggests that multiphysics/multimodel ensemble mean forecasts may only add value when the individual models possess similar skill and error. An atmosphere-only version of FIM-CGF loses skill after 11 days, highlighting the importance of ocean coupling. Further examination reveals some sensitivity in skill and error metrics to the choice of MJO index.