Benjamin W. Green

West Nile virus (WNV) infection has caused over 30,000 human cases of the severe, neuroinvasive form of the disease (West Nile virus Neuroinvasive Disease; WNND) and nearly 3,000 deaths in the U.S. Despite known links to observable climate factors, no effective nationwide WNV or WNND forecast exists. We aimed to produce a skillful, nationwide WNND forecast built upon regionally varying relationships between climate factors and WNND. After examining the relationships between climate conditions and annual WNND caseload for 11 regions in the U.S., we incorporated the most salient climate factors—most commonly drought and temperature—into a regionally determined nationwide WNND statistical forecast model using a Bayesian regression framework. We retrospectively generated forecasts from 2005 to 2022 and compared forecast skill against various benchmarks, including a simple, historical case‐driven model. Our regional, climate‐informed WNND retrospective forecasts outperformed a benchmark model only informed by historical WNND case data across all regions, as well as in a nationally aggregated score (univariable: 18.8% [4.7%–27.7%], bivariable: 21.8% [7.0%–30.7%] improvement). The regional forecasts also outperformed an ensemble model generated from a recent WNV forecasting competition and a parallel, county‐level, regional climate‐informed forecast outperformed forecasts from the same competition. Importantly, our approach to WNND forecast development aggregated county‐level data to broader regions to boost statistical signal and capture the regionally varying influences of climate conditions on annual WNND caseload. The advances here represent a potential path toward actionable broad‐scale WNV forecasts.

Background: West Nile virus (WNV) infection has caused over 30,000 human cases of the severe, neuroinvasive form of the disease (West Nile virus Neuroinvasive Disease; WNND) and nearly 3,000 deaths in the U.S. since its introduction in 1999. Despite spatiotemporal variation in the impact of WNV and its known links to various climate factors, no effective nationwide WNV or WNND forecast exists. Objectives: We aim to produce a skillful, nationwide WNND forecast built upon regionally varying relationships between climate factors and WNND. Methods: We examined the impact of climate conditions on annual WNND caseload for 11 ecologically meaningful regions in the U.S. The most salient climate factors were incorporated into a regionally determined nationwide WNND forecast model. We retrospectively generated forecasts from 2005-2022 using observed climate conditions and compared forecast skill against various benchmarks, including a simple, historical case-driven model. Forecast skill was assessed by weighted interval scoring. Results: Regional, climate-informed WNND retrospective forecasts outperformed a benchmark model only informed by historical WNND case data across all regions, as well as in a nationally aggregated score (univariate: 18.1% [3.7–27.0%], bivariate: 23.9% [9.0–32.8%]). Additionally, the regional forecasts outperformed an ensemble model generated from the 2022 CDC WNV Forecasting Challenge and a parallel, county-level, regional climate-informed forecast outperformed forecasts from the same Challenge. Drought and temperature were the climate factors most consistently linked to WNND and incorporated into our forecast model. Discussion: We show a retrospectively generated WNND forecast for the continental U.S. that considerably improved upon simple forecasts based on historical case distributions. This forecast aggregated county-level data to broader regions to boost statistical signal and capture the regionally varying influences of climate conditions on annual WNND caseload. The advances here represent a potential path toward actionable broad-scale WNV forecasts.

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.

NOAA has been developing a fully coupled Earth system model under the Unified Forecast System framework that will be responsible for global (ensemble) predictions at lead times of 0–35 days. The development has involved several prototype runs consisting of bimonthly initializations over a 7-yr period for a total of 168 cases. This study leverages these existing (baseline) prototypes to isolate the impact of substituting (one-at-a-time) parameterizations for convection, microphysics, and planetary boundary layer on 35-day forecasts. Through these physics sensitivity experiments, it is found that no particular configuration of the subseasonal-length coupled model is uniformly better or worse, based on several metrics including mean-state biases and skill scores for the Madden–Julian oscillation, precipitation, and 2-m temperature. Importantly, the spatial patterns of many “first-order” biases (e.g., impact of convection on precipitation) are remarkably similar between the end of the first week and weeks 3–4, indicating that some subseasonal biases may be mitigated through tuning at shorter time scales. This result, while shown for the first time in the context of subseasonal prediction with different physics schemes, is consistent with findings in climate models that some mean-state biases evident in multiyear averages can manifest in only a few days. An additional convective parameterization test using a different baseline shows that attempting to generalize results between or within modeling systems may be misguided. The limitations of generalizing results when testing physics schemes are most acute in modeling systems that undergo rapid, intense development from myriad contributors—as is the case in (quasi) operational environments.

Tropical cyclone (TC) intensity is strongly influenced by surface fluxes of momentum and moist enthalpy (typically parameterized in terms of “exchange coefficients” Cd and Ck, respectively). The behavior of Cd and Ck remains quite uncertain especially in high wind conditions over the ocean; moreover, moist enthalpy flux is extremely sensitive to sea surface temperature (SST). This study focuses on numerical simulations of Hurricane Katrina (2005) from an atmosphere–ocean coupled modeling system to examine the combined impacts of air–sea flux parameterizations and ocean cooling on TC evolution. Three momentum flux options—which make Cd increase, level off, or decrease at hurricane-force wind speeds—with five different Ck curves are tested. Maximum 10-m wind speed Vmax is highly sensitive to Cd, with weaker sensitivities for minimum sea level pressure Pmin and track. Atmosphere-only runs that held SST fixed yielded TCs with Pmin substantially deeper than observations. Introducing ocean coupling weakens TC intensity with much more realistic Pmin. The coupled run with the flux parameterization that decreases Cd at high wind speeds yields a simulated TC intensity most consistent with observations. This Cd parameterization produces TCs with the highest Vmax. Increasing Ck generally increases surface heat fluxes and thus TC intensity. For coupled runs using the default Ck parameterization, the simulated SST fields are similar (regardless of Cd parameterization) and agree well with satellite observations. The mesoscale oceanic eddies, which are well resolved in the ocean model, contribute to the magnitude of TC-induced SST cooling and greatly influence TC intensity.

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.

The tropical cyclone (TC) boundary layer (TCBL)—featuring extreme winds over a rough ocean—is difficult to study observationally. With increasing computational power, high-resolution large-eddy simulation (LES) has become an attractive tool to advance understanding of the TCBL. Here, an idealized Cartesian-based LES is employed to investigate boundary layers driven by extreme TC-like winds. The LES includes the effects of centripetal acceleration through an “effective” Coriolis parameter f* = f + 2Vg/R, with the Earth Coriolis parameter f, gradient wind Vg, and (fixed) radius R. Multiple LES experiments are conducted to elucidate how the boundary layer develops and persists in the strongly rotating TC environment. In all simulations, an overshooting jet develops, the height of which increases with Vg, R, and surface drag. Normalized jet strength also increases with R and drag but decreases with Vg. Turbulent diffusivity Km—which must be parameterized in mesoscale and global models but can be diagnosed by LES—varies considerably both within and among simulations. Also evident is a pseudo-inertial oscillation with a period close to the theoretical 2π/f* and an amplitude that decreases exponentially with time. The LES simulations agree with the linear theory for partial-slip Ekman spirals, except when the effects of Km overwhelmingly counter the effects of Vg.

Repeat-pass spaceborne interferometric synthetic aperture radar (InSAR) is commonly used to measure surface deformation; phase delays due to atmospheric water vapour may have significant impact on the accuracy of these measurements. In recent years, there has been a growing interest in using forecasts and analyses from numerical weather prediction (NWP) models – which can provide good estimates of the atmospheric state – to correct for atmospheric phase delays. In this study, three separate estimates of atmospheric water vapour content from NWP output are used in combination with Environmental Satellite (Envisat) Advanced Synthetic Aperture Radar (ASAR) data over the Pearl River Delta region in South China to mitigate atmospheric distortion. The NWP-based estimates are derived from: (1) interpolation of National Centers for Environmental Prediction (NCEP) Final Operational Global Analysis (FNL) data; (2) Weather Research and Forecasting (WRF) model simulations initialized with FNL analysis without additional data assimilation; and (3) WRF simulations initialized with a three-dimensional variational (3DVar) data assimilation system that ingests additional meteorological observations. The accuracy of the atmospheric corrections from these different NWP model outputs is further verified quantitatively with precipitable water vapour (PWV) data from several ground-based global positioning system (GPS) stations in Hong Kong. Inter-comparison shows a good agreement between the PWV derived from the WRF-3DVar simulations and the GPS measurements, suggesting that atmospheric correction by convection-permitting WRF simulations initialized with mesoscale data assimilation may effectively mitigate atmospheric distortion in InSAR measurements, especially for coastal areas.

Current numerical simulations of tropical cyclones (TCs) use a horizontal grid spacing as small as Δx = 10³ m, with all boundary layer (BL) turbulence parameterized. Eventually, TC simulations can be conducted at Large Eddy Simulation (LES) resolution, which requires Δx to fall in the inertial subrange (often 2 m) to adequately resolve the large, energy‐containing eddies. Between the two lies the so‐called “terra incognita” because some of the assumptions used by mesoscale models and LES to treat BL turbulence are invalid. This study performs several 4–6 h simulations of Hurricane Katrina (2005) without a BL parameterization at extremely fine Δx [333, 200, and 111 m, hereafter “Large Eddy Permitting (LEP) runs”] and compares with mesoscale simulations with BL parameterizations (Δx = 3 km, 1 km, and 333 m, hereafter “PBL runs”). There are profound differences in the hurricane BL structure between the PBL and LEP runs: the former have a deeper inflow layer and secondary eyewall formation, whereas the latter have a shallow inflow layer without a secondary eyewall. Among the LEP runs, decreased Δx yields weaker subgrid‐scale vertical momentum fluxes, but the sum of subgrid‐scale and “grid‐scale” fluxes remain similar. There is also evidence that the size of the prevalent BL eddies depends upon Δx, suggesting that convergence to true LES has not yet been reached. Nevertheless, the similarities in the storm‐scale BL structure among the LEP runs indicate that the net effect of the BL on the rest of the hurricane may be somewhat independent of Δx.

Tropical cyclones (TCs) are strongly influenced by fluxes of momentum and moist enthalpy across the air–sea interface. These fluxes cannot be resolved explicitly by current-generation numerical weather prediction models, and therefore must be accounted for via empirical parameterizations of surface exchange coefficients (CD for momentum and Ck for moist enthalpy). The resultant model uncertainty is examined through hundreds of convection-permitting Weather Research and Forecasting Model (WRF) simulations of Hurricane Katrina (2005) by varying four key parameters found in commonly used parameterizations of the exchange coefficient formulas. Two of these parameters effectively act as multiplicative factors for the exchange coefficients over all wind speeds (one each for CD and Ck); the other two parameters control the behavior of CD at very high wind speeds (i.e., above 33 m s−1). It is found that both the intensity and the structure of TCs are highly dependent upon the two multiplicative parameters. The multiplicative parameter for CD has a considerably larger impact than the one for Ck on the relationship between maximum 10-m wind speed and minimum sea level pressure: CD alters TC structure, with higher values shifting the radius of maximum winds inward and strengthening the low-level inflow; Ck only affects structure by uniformly strengthening/weakening the primary and secondary circulations. The TC exhibits the greatest sensitivities to the two multiplicative parameters after a few hours of model integration, suggesting that these parameters could be estimated by assimilating near-surface observations. The other two parameters are likely more difficult to estimate because the TC is only marginally sensitive to them in small areas of high wind speed.