Tom Warner's research while affiliated with National Center for Atmospheric Research and other places

The transport and dispersion (T&D) models used for air-quality and defense applications require information describing the source parameters and meteorological conditions to forecast concentration and dosage fields. In many cases the source parameters are known and the meteorological conditions are based on observational data or mesoscale-model-generated forecast conditions. This research examines how errors in the input wind fields translate into uncertainty in the contaminant concentration predictions. In particular, this study focuses on street-level errors in the dispersion patterns that occur in “building aware” T&D models that are sensitive to urban designs (e.g. road and building patterns) and release locations relative to the buildings. This problem was evaluated by first creating a “truth” plume for a given release location and wind direction. Then the T&D model uncertainty associated with input wind errors were determined by comparing plumes calculated using wind directions varied at 2° increments to the truth plume. The uncertainty is quantified as fraction of overlap (FOO). The results are evaluated in a control simulation with no buildings, and in two commonly observed city designs (e.g. a regular grid, and hub and spoke configuration). The analysis examines both idealized building configurations along with the urban topography from cities that represent the regular grid and hub and spoke city designs. Results show that the relative impact of the uncertainty in the meteorological conditions and the corresponding sensitivity of the model to variations in the wind direction vary significantly with the release location and city designs. This suggests that some source locations are less (more) sensitive to uncertainty in meteorological conditions and that this information can be factored into the confidence that is placed in emergency response decisions based on this information.

This paper describes an advanced multi-scale weather modeling system, WRF–RTFDDA–LES, designed to simulate synoptic scale (∼2000km) to small- and micro-scale (∼100m) circulations of real weather in wind farms on simultaneous nested grids. This modeling system is built upon the National Center for Atmospheric Research (NCAR) community Weather Research and Forecasting (WRF) model. WRF has been enhanced with the NCAR Real-Time Four-Dimensional Data Assimilation (RTFDDA) capability. FDDA is an effective data assimilation algorithm, which is capable of assimilating diverse weather measurements on model grids and seamlessly providing realistic mesoscale weather forcing to drive a large eddy simulation (LES) model within the WRF framework. The WRF based RTFDDA LES modeling capability is referred to as WRF–RTFDDA–LES. In this study, WRF–RTFDDA–LES is employed to simulate real weather in a major wind farm located in northern Colorado with six nested domains. The grid sizes of the nested domains are 30, 10, 3.3, 1.1, 0.370 and 0.123km, respectively. The model results are compared with wind–farm anemometer measurements and are found to capture many intra-farm wind features and microscale flows. Additional experiments are conducted to investigate the impacts of subgrid scale (SGS) mixing parameters and nesting approaches. This study demonstrates that the WRF–RTFDDA–LES system is a valuable tool for simulating real world microscale weather flows and for development of future real-time forecasting system, although further LES modeling refinements, such as adaptive SGS mixing parameterization and wall-effect modeling, are highly desired.

In collaboration with Xcel Energy and Vasaila Inc., the National Center for Atmospheric Research (NCAR) conducts modeling study to evaluate the existing and the enhanced intensive observation systems for wind power nowcasting and short-range forecasting at a northern Colorado wind farm. The NCAR WRF (Weather Research and Forecasting model) based Real-Time Four-Dimensional Data Assimilation (RTFDDA) and forecasting system, which has been employed to support Xcel Energy operational wind forecast, was used in this study. The observational data include ten met-towers, a 915Hz wind profiler, a sodar and a Windcube Doppler lidar, besides the in-farm met-towers and wind speed and power reports from more than 300 of wind turbines. The WRF-RTFDDA 4-dimensioanl data assimilation algorithm allows to spread and propagate observation information in the WRF model space (x, y, z and time) with weighting functions built according to the observation location and time. The WRF-RTFDDA was set up to run with four nested domains with grid increments of 30, 10, 3.333 and 1.111km respectively. The standard and diverse non-conventional observations are assimilated on coarse grid domains along with the special wind farm observations. In this study, we investigate a) spread of surface observations in PBL according to PBL depth and regimes, b) optimization of horizontal influence radii and steep-terrain adjustment, and c) impact of different observation platforms and data types on 0 - 12 h wind prediction . It is found that PBL mixing and thermodynamic structures are greatly influenced by the PBL parameterization formulation. The range of the data assimilation effect on forecasts relies on weather and PBL regimes. In most cases, assimilation of in-farm and near-farm observations improves up to 12-hour wind power prediction and assimilation of in-farm data can significantly improves 0 - 6 hour forecasts.

Major weather centers, such as National Center for Environment Prediction (NCEP) and ECMWF, produce inter-seasonal weather predictions 6 - 9 months ahead. However, the products from these centers have ~200 km grid sizes, which are too coarse for regional applications. For hydrological applications, such as flood forecasting, watershed control, and water resource planning, detailed spatial and temporal distributions of precipitations are very critical. Existing precipitation downscaling approaches include statistical downscaling algorithms (SDA) and dynamical downscaling algorithms (DDA). SDAs are mostly based on regression using reanalysis and/or hindcasts and may apply for future forecast downscaling. SDAs impose three assumptions: a) the past regression relation is (static) valid for the future, b) there is no feedback of local physical forcing (terrain, coastlines and land-use/soil properties) in response to weather/climate changes and c) downscaling valid at the stations where long historical observations are available. DDAs, by which a regional climate model is embedded (nested) in a global seasonal model, overcome many of the shortcomings of a SDA. However, DDAs are computationally costly and data handling is complicated. In this paper, we present a dynamic-enforced statistical downscaling algorithm (DESDA) for effectively downscaling global-model seasonal forecasts. Four steps are involved with DESDA: 1) using the NCAR Four-Dimensional Data Assimilation (FDDA) modeling system, built upon the Weather Research and Forecasting (WRF) model, to produce 1 - 4 km gridded climatological precipitation-distribution analyses over the eastern Mediterranean region, driven by global analyses; 2) calibrating the gridded model precipitation with available precipitation measurements; 3) Applying an advanced KNN based regression downscaling approach based on the calibrated high-resolution gridded precipitation analysis, NCEP global analysis, and NCEP climate forecasting system (CFS) model 29 years of reforecasts for downscaling the CFS seasonal forecasts of precipitation anomalies; and 4) reconstructing precipitation amounts of the seasonal forecasts on the high-resolution WRF analysis grids. The algorithm and preliminary results will be presented.

The atmosphere can influence the spread of human and agricultural infectious diseases through a number of different mechanisms, including the effect of the atmosphere on the health of the pathogen itself, the health and number of disease vectors, human behavior, wind transport, and flooding. Through knowledge of the statistical or physical relationships between disease incidence, for example outbreaks, and weather or climate conditions, it is possible to translate predictions of the atmosphere into predictions of disease spread or incidence. Medium range forecasts of weeks can allow redistribution of vaccines and medical personnel to locations that will be in greatest need. Inter-seasonal forecasts, e.g. based on the ENSO cycle, can provide long-lead-time information for disease early-warning systems, which can guide the manufacture of vaccines and inform aid agencies about future requirements. And knowledge of longer-term trends in climate conditions, associated, for example, with increases in green-house gases, can be used for development of infectious-disease mitigation and prevention policies. Because of the existence of complex physical, biological, and societal aspects to the links between atmospheric conditions and disease, prediction systems must be constructed based on knowledge of multiple disciplines. To be described in the presentation are activities at the National Center for Atmospheric Research that involve the coupling of atmospheric models with infectious-disease models and decision-support systems. These include 1) the use of operational multi-week weather forecasts to estimate the spatial and temporal variability of the threat of bacterial meningitis in West Africa, 2) climate and spatial risk modeling of human plague in Uganda, 3) a study of how climate variability and human landscape modification interact to influence key aspects of both mosquito vector ecology and human behavior, and how they influence the increased incidence of dengue fever in Mexico, and 4) development of new knowledge about how extreme heat events across the United States and parts of Canada result from changing climate, land use and the interactions between them. In addition, NCAR has an arrangement with the US Centers for Disease Control wherein postdoctoral students are shared between the two organizations in order to provide experiences that will foster research at the interface between climate science and the study of infectious diseases.

Ensemble-based mesoscale data assimilation and probabilistic forecasting are traditionally separated in their developments. However, an accurate forecast of probabilistic distribution functions of state variables is in fact equally important for both ensemble-based data assimilation and probabilistic prediction. Poor sampling and forward propagation of initial states and model uncertainties lead to inaccurate probabilistic forecasts and deficient estimate of background error covariance required for Ensemble Kalman Filter data assimilation (EnKF). Thus a well-formulated ensemble prediction system should provide more accurate estimate of the forecast error covariance for EnKF. On the other hand, EnKF is an effective tool for sampling the model initial condition uncertainties that are highly desirable for mesoscale ensemble prediction. It should be noted that mesoscale processes are more complicated than global models and may be dominated by physical processes at times. Thus mesoscale model forecast errors (uncertainties) depend heavily on parameterized physical processes that contain many assumptions and uncertain parameters. A seamless ensemble data assimilation and probabilistic prediction scheme can address the issues on both aspects. An innovative seamless mesoscale ensemble data assimilation and prediction system has been developed at NCAR. The system contains two major sub-modeling systems. One is the NCAR mesoscale Ensemble Real-Time Four Dimensional Data Assimilation (E-RTFDDA) and forecasting system has been developed at NCAR and the other is the NCAR DART (Data Assimilation research Testbed) EnKF modules. E-RTFDDA, built based on WRF and MM5 models, contains diverse ensemble perturbation approaches that take into account of uncertainties in all major modeling system components to produce multi-scale, continuously-cycling probabilistic data assimilation and forecasting. A 30-member E-RTFDDA system with three nested domains with grid sizes of 30, 10 and 3.33 km has been operating for US Army test ranges since September 2007. In the seamless system, the NCAR DART EnKF tools are integrated to E-RTFDDA to formulate an integrated ensemble data assimilation and prediction capability. With this system, EnKF takes advantages of E-RTFDDA by deriving error covariance with an addition of the multiple-perturbation-approach E-RTFDDA forecasts and then it feeds E-RTFDDA with a subset of initial condition perturbations derived from the EnKF perturbation members. Numerical experiments have been conducted for a Cold Air Damming weather event over the Northeastern US to validate and assess the theory and the advantages of the seamless ensemble data assimilation and prediction system. The modeling results and the directions for further improvements will be presented.

Global seasonal forecasts of precipitation are currently produced by the major weather centers. These predictions are available several months in advance at horizontal resolutions of ∼200 km grid-size. They have proved useful to providing an estimate of the expected precipitation over large areas. However, their value is limited for regional applications, for example, hydrological applications such as water resources planning and flood forecast in areas characterized by complex terrain, where information at finer temporal and spatial resolutions is required. Downscaling of global precipitation forecasts to the regional scale is possible through statistical and dynamical approaches. Each of these strategies possesses advantages and limitations in physical, computational and real-time implementation aspects; these have been widely reviewed and discussed in literature. For instance, statistical downscaling is computationally cheap but it relies on reliable long-term records of observed precipitation. These may be sparsely distributed. In contrast, dynamical downscaling techniques which produce regional scale gridded precipitation forecasts using regional climate model nested down from global models, may fill the gaps in sparsely observed areas, but the technique is computationally demanding, in particular if real-time forecasts are desired. The present work combines dynamical and statistical downscaling methods to provide real-time seasonal forecasts of precipitation at high horizontal resolution. These forecasts will serve hydrological applications in the Levant area, where the water budget strongly depends on the fine spatial distribution of seasonal precipitation. Statistical downscaling based on a KNN regression algorithm which makes use of predictors from CFS model outputs and local precipitation data is applied in a twofold manner. In the first instance, it is implemented using the available long-term gauges observations and NCEP CFS global seasonal forecasts to provide seasonal precipitation forecasts at the observation locations. To fill the unobserved gaps, a high-resolution re-analysis of precipitation is produced by downscaling global re-analysis using WRF model and assimilation of meteorological observations. The gridded downscaled precipitation is used as gridded synthetic observations in the KNN-based statistical downscaling procedure of NCEP CFS forecasts: a dynamically enhanced statistical downscaling procedure. A comparative discussion of both methodologies and preliminary results will be presented.

Seasonal precipitation prediction has significant societal and economic impact, particularly for arid and semiarid regions. Current seasonal predictions generally rely on general circulation models (GCMs), which have coarse resolution (~300km). The GCM forecasts provide overall guidance in terms of large and synoptic scale perspectives, but are lack of regional and local details and accuracy that are needed by hydrological applications and water resources planning and management. On the other hand, high-resolution (~10s km) limited-area models have their own issues for operational seasonal forecasting due to unavailability of reliable large-scale drivers and unaffordable computational costs. Thus statistical and dynamical downscaling techniques have emerged to overcome scale mismatch between GCM products and regional (and local) application needs. In this study, a K-Nearest Neighbor (KNN) simulator is used to derive local precipitations based on NCEP Climate Forecast System (CFS) seasonal forecasts and historic rainfall observations. The KNN algorithm is an analog-type approach that queries days within a specified temporal window similar to a given weather feature vector in a GCM forecast. K nearest neighbors is then rank-weighted to derive daily precipitation with the historic observed precipitations. This study focuses on the semiarid area along the southeastern Mediterranean coast. This region is strongly influenced by the Mediterranean climate and complex terrain. Annual precipitation displays strong seasonality and spatial variability. Enhanced seasonal precipitation prediction with local details would benefit the regional hydrological service. Archived CFS seasonal forecasts (1981-2009, and up to 9 months ahead of the initials) are built as our database for weather pattern matching, and observed daily precipitations at stations within the region are compiled from different sources to minimize errors and missing in the observations. Four variables (500 hPa geopotential, sea level pressure, precipitation rate and precipitable water) in CFS output are used as predictors for the KNN daily precipitation forecasts. Monthly precipitation is then simply calculated from the KNN daily products. The analog patterns obtained from the KNN algorithm are cross-examined by another pattern classification tool: Self-organizing Maps (SOM). Seasonal precipitation forecasts (monthly rainfalls) are evaluated using statistics like bias, uncertainty range, and spatial covariability between stations as a measure of forecast skill. Observation incompleteness in the region presents challenge to the KNN application. Thus, retrospective dynamical downscaling with a WRF-based four-dimensional data assimilation system is conducted to reconstruct the regional precipitation climate. The reconstructed precipitations are then used for KNN seasonal rainfall prediction application. Discussion on the downscaling algorithms and results will be presented.

Accurate wind and severe-weather forecasts are crucial for wind-energy production and grid-load management. Most of the prevailing wind power forecast methods rely heavily on statistical approaches that typically do not deal directly with weather processes. Currently employed numerical weather prediction (NWP) models are deemed insufficiently accurate by many industry stakeholders for wind power prediction, even though they have been used for such applications. The reason is partly because the NWP products used for power forecasting are typically produced by the coarse-resolution models at the major operational weather centers. Although a few of high-resolution models are run by some wind energy industries, most of these model do not contain advanced data assimilation capabilities that are required to initialize the model prediction with the important high-resolution weather information. In this presentation, we introduce the NCAR Real-Time Four Dimensional Data Assimilation (RTFDDA) and forecasting system that has been developed to specifically analyze and predict meteorological conditions over small regions. Operational RTFDDA systems have been implemented across the United States and other global regions to support tens of other weather-critical applications in the last nine years. The system provides rapidly updated, multi-scale weather analyses and forecasts with the fine-mesh domain having a 0.5 - 3 km grid increment. The presentation will focus on the modification and improvements to the NWP technologies in RTFDDA for wind energy applications. The technologies include the use of a) an advanced mesoscale weather model (WRF, Weather Research and Forecasting model) with a continuous 4-D data assimilation scheme, b) an effective data quality-control procedures that handles the ingestion of diverse weather data sources, c) special algorithms for the assimilation of wind-farm measurements including met-tower and turbine nacelle wind speeds, d) ensemble-based probabilistic data analysis and forecasting, e) a sophisticated land-surface model and land-surface data assimilation system, f) model nests from the synoptic scale to the intra-farm microscales, and g) advanced model post-processing for prediction-error correction and power-generation calibration. These technologies are being utilized as part of a research and development partnership with Xcel Energy Services, Inc., which purchases power from wind farms across the US. Results for a large wind farm (with 274 wind turbines) in northern Colorado are analyzed in detail to illustrate the capabilities of each of the technologies described above. Inherent challenges in wind modeling and the predictability of intra-farm-scale weather processes and the high priority research directions for wind forecasting will also be discussed.

Forecasting wind power is highly desired for power grid operation, and it is extremely challenging as well because of the sporadic temporal nature of atmospheric winds. Power companies must anticipate the magnitude and timing of wind power in order to balance their power load. Numerical weather prediction (NWP) models such as WRF are very useful tools in forecasting mesoscale weather and winds at wind farms. Nonetheless, it is well known that model physics can have a profound impact on the model solution. Furthermore, many of the WRF model physics packages were designed for idealized scenarios and over flat terrain, which may not be performed in the same way over complex terrain regions where many wind farms locate. Research Applications Laboratory (RAL) at National Center for Atmospheric Research (NCAR), in collaboration with Xcel Energy to develop a real time wind power forecast systems for various wind farms including those in Colorado. One of the core components of the system is a high-resolution NCAR/ATEC WRF-ARW-based Real-Time Four-Dimensional Data Assimilation (RTFDDA) and forecasting system (Deltax = 3.3 km). This study evaluated different physics parameterizations and combinations in cloud microphysics, cumulus parameterization, land surface model, surface layer scheme, and planetary boundary layer parameterization in forecasting wind speed in two summer and winter ramp-up events in a Northern Colorado wind farm, respectively. We will compare and contrast the performance of different physics parameterizations and combinations between the summer (convective storms) and winter (cold frontal passage) ramp-up cases.