Ensemble work is described briefly in the previous “Model Ensemble & Ensemble Post Processing” Section and more extensively in connection with the HMT project in the section below.
Hydrometeorological Testbed (HMT)
Participating CIRA Scientists: Steve Albers, Chris Anderson, Ed Szoke, and Isidora Jankov
1. Long-term Research Objectives and Specific Plans to Achieve Them:
The Hydrometeorological testbed (HMT) is a well-funded, multi-year project (hmt.noaa.gov) designed to improve the use of research quality observations and modeling in operational forecasts of precipitation and stream flow. Three large field campaigns were held in December through March of the past three winter seasons in the American River Basin (ARB) of the Central Sierra Mountains (Fig. 1). CIRA researchers in the Forecast Applications Branch (FAB) are an integral part of ESRL/GSD’s effort to provide high-resolution model analyses and forecasts, as well as forecast interpretation by meteorologists, in support of field operations and NWS operational forecasting.
Fig 1. Basin-scale map of the third full-scale deployment of HMT-West; successfully conducted December, 2007 - March, 2008
The HMT/LAPS analyses (laps.noaa.gov) are used to create web graphics for nowcasting, and they provide gridded initial conditions for experimental numerical weather prediction models being run in support of NWS weather forecast office and river forecast center operations. The analysis software assimilates a wide variety of in-situ and remotely sensed data including GOES satellite and full volume reflectivity and velocity scans from nine WSR-88D radars. Some experimental observation systems are assimilated as well, e.g., the 915 MHz profilers deployed in the HMT domain by ESRL/PSD. We set up and made improvements to our real-time hourly analysis run over the American River Basin running at 3km resolution. The surface relative humidity and wind fields show a wealth of detail related to the topography and larger scale weather systems (Fig. 2). Fig. 3 shows where the precipitation (both liquid equivalent and snowfall) has accumulated over the past day in the HMT domain, as well as the current location and type of precipitation. Cross-sections (Fig. 4) also help show how the three dimensional wind and precipitation fields interact with the terrain. This information is useful for a forecaster who may want to anticipate the evolution of precipitation over the ARB, located in the center of the domain.
Fig 3. LAPS analysis of 24-hr accumulated precipitation (shaded colors), current surface precipitation type (green icons), and 24-hr snow accumulation (orange contours).
Fig 4. LAPS cross-section of clouds, radar reflectivity and precipitation type showing clouds moving over the California coastal range (left), and precipitation over the Sierra Nevada mountain range (right). The section is oriented perpendicular to the Sierras.
Forecasts from NWP models are a primary source of guidance to forecasters at forecast lead teams beyond 6 to 12 hours. NWP efforts for HMT are focused on improving precipitation forecasts in order to improve the timeliness of flash flood warnings and the accuracy of stream and river flow predictions.
During the 2007-2008 winter field campaign, we responded to forecaster input by producing ensemble forecasts out to a 72 hour lead time. We added a few diagnostic variables and participated in system testing to improve the timeliness of model output display on ALPS machines installed during the field campaign.
The project is now moving into an intense developmental phase in which a new forecast domain will be designed, perhaps new ensemble strategies explored, and computing resources determined to install the new ensemble system with the support of California Department of Water funding at the California-Nevada River Forecast Office in 3-4 years.
Daily Forecast Support
For the third year in a row, CIRA researchers provided real-time support to the field experiment in the form of detailed daily discussions of the precipitation forecast over the testbed. This included participation in the daily (7 days a week) conference call and posting text forecasts to the HMT website.
An important part of the HMT program, as outlined at http://www.esrl.noaa.gov/psd/programs/2007/hmt/, is to assess various instrumentation, including new radar technologies, designed to better measure precipitation and determine precipitation type. The ultimate goal is to arrive at more accurate quantitative precipitation estimates (QPE) which, coupled with the advances in high resolution numerical modeling, can lead to improved hydrologic forecasts and warnings. Additional observations during events include special radiosonde launches at frequent intervals to document the characteristics of each storm. All of these special observations require scientists to be on station for each event, but events might be widely spaced in time, so the strategy has been to make a forecast of each event and then staff accordingly, with some of the required staff having to fly in from Boulder as well as Norman, Oklahoma.
The general forecast goals are to give as much advance warning of a potential event (an "Intensive Operational Period" or IOP) as possible, with a go/no go decision needed usually no less than 24 to 48 hours in advance. The ultimate decision to call an IOP rests with the Project Director (this position rotating among several scientists within NOAA), but is of course highly influenced by the forecasters and their confidence in a potential IOP. A conference call among HMT participants occurs every day during the program at 12:30 local mountain time, with the initial business a forecast discussion, followed by further discussion and interpretation leading to a decision on a potential IOP. A written forecast and forecast discussion is also done and posted to the project webpage (http://www.esrl.noaa.gov/psd/programs/2007/hmt/) near or shortly after the conference call. A preliminary version of the forecast discussion is sent to the project directors at least an hour or so ahead of the conference call for planning purposes.
CIRA staff have been an integral part of a larger forecasting cadre that represents a cooperative effort between the National Weather Service (NWS) Weather Forecast Offices (WFO) in the HMT area, which are the Sacramento and Monterey WFOs in California and the Reno WFO in Nevada, the NOAA California-Nevada River Forecast Center in Sacramento, and the NOAA National Centers for Environmental Prediction (NCEP) Heavy Precipitation Branch. During the HMT exercise, the Boulder forecasters were responsible for the written discussion posted to the website and leading the forecast part of the daily conference call. Typically, the forecast discussion involves input from the other participants, particularly when the weather prediction becomes less certain. A prototype AWIPS workstation was installed at the participating WFOs to allow the forecasters there to examine the output from the special model forecasts being run at GSD for the project. There was also occasional participation by two NOAA forecasters from ESRL's Physical Sciences Division, who provide occasional longer range (2 to 3 week) guidance based on their analyses and interpretation of model forecasts.
CIRA and other forecasters in Boulder use a variety of information and model forecasts to make the HMT daily forecast. The standard operational models are found on AWIPS, but the web offers a look at many other models as well as ensemble model forecasts from the NCEP Global Forecast System (GFS), as well as a set of ensemble forecasts from Environment Canada. Analyses of water vapor and other parameters are available from a number of other sites, with one of the favorites out of the University of Hawaii. A set of the most often used sites has been compiled onto a webpage for the project at http://laps.fsl.noaa.gov/szoke/DWB/Hydromet_Test_Bed_fcsthomepage.html. In the shorter range, of course, the forecasters utilize the various 3 km and ensemble runs initialized with LAPS and run locally at GSD that are described in this article.
Post-Season Analysis and Research
In terms of research activities, there are several ongoing HMT projects in collaboration with PSD and CIRES scientists. Initially, a study by Jankov et al (J. Hydrometeor., 2007) evaluated the impact that various microphysical schemes, planetary boundary layer (PBL) schemes, and initial conditions had on Quantitative Precipitation Forecast (QPF) over the HMT area and for events characterized by atmospheric river settings. It was found that for this type of event and for this location, variations in microphysics resulted in a statistically significant impact on simulated precipitation amounts. Several ongoing studies build on this finding and focus on various ways to evaluate performance of the high-resolution numerical model with various microphysics and possibly finding a way to improve QPF in the case of significant precipitation events during the winter in California.
An ongoing study in collaboration with several scientists from NOAA/PSD (Jian-Wen Bao, Paul Neiman and Allen White) and CIRES (Huiling Yuan) focuses on a detailed analysis of a high-resolution numerical model with various microphysics and its performance in cases of atmospheric river events. For this purpose, simulations of the representative events have been performed by utilizing the WRF-ARW numerical model with four different microphysics (Lin, Thompson, WSM6 and Schultz). The evaluation consisted of comparisons of the flow and cloud structure against observations from experimental radars deployed for the HMT project. This study has revealed more details about performance of various microphysical schemes for this type of event with mountainous terrain and different precipitation regimes (Bright Band vs. Non-Bright Band). Even though various microphysics have demonstrated a large diversity in their solutions, it has been found that all model configurations had a tendency to overestimate simulated precipitation amounts possibly due to the model’s tendency to overestimate the moisture content and the upslope wind component’s duration and intensity.
Fig 5. Time series of observations from BBY and CZD between 22 UTC 29 December and 22 UTC 31 December 2005. (a) Time-height section of hourly-averaged wind profiles (wind flags 25 m s-1; barbs = 5 m s-1; half-barbs = 2.5 m s-1), along-front isotachs (directed from 230 degrees; red shading >20 m s-1), bright band melting-level height (bold black dots), and axes of maximum thermal wind-derived warm and cold advection (red and blue dashed lines, respectively), from the wind profiler at BBY. (b) Time-height section of ~1.5-min radar reflectivity (dBZ) from the S-band radar at CZD. (c) Time-series traces of 30-min IWV (cm; green) from the GPS sensor at BBY and 2-min rain accumulation (mm) and rain rate (RR = mm h-1; 10-min averaging period) data recorded at the rain gauge at CZD. The red- and blue- shaded bars in the bottom panel denote warm- and cold-frontal transitions, respectively.
The finding about the WRF-ARW model tendency to overestimate the intensity and duration of the upslope wind component has served as a motivation for a study to quantify the impact that various microphysical schemes and intensity of upslope wind component as well as the interaction between the two have on simulated orographically induced rainfall. For this purpose, 12-hour high resolution WRF-ARW model simulations of a typical atmospheric river event will be used. The simulations will include variations in microphysical schemes and wind initial conditions. The initial wind perturbations will be generated by Monte Carlo sampling from Gaussian distribution using analysis based covariance estimate. The impact of microphysics and wind variations on the simulated rainfall will be quantified by using the factor separation methodology. This method measures absolute, relative and synergistic contribution of each variation to the simulated rainfall.
Further evaluation of the WRF-ARW model with various microphysics performances will be assessed by producing synthetic satellite imagery and using an objective measure of difference in various microphysics compared to observations. This research will be performed in collaboration with CIRA scientists Louie Grasso and Manajit Sengupta. To make this research possible, a production of observational operators for WRF-ARW model has been developed as a part of collaborative work with a group of CIRA scientists (Renate Brummer, Louie Grasso, Manajit Sengupta, Dusanka and Milija Zupanski) on a WRF/RAMS Synthetic Imagery Project. An example of a synthetic satellite image (brightness temperature) from a WRF-ARW model simulation using Schultz microphysics over the HMT domain is presented in Fig. 6
Fig. 6. Synthetic brightness temperature calculated from WRF-ARW model output using WSM6 microphysics valid at 06 UTC 30 december 2005.
An NSF proposal in collaboration with CIRES scientist Tomislava Vukicevic for work which will focus on improving nonconvective precipitation accuracy by objectively estimating the individual contributions of parameterized processes representing generation and depletion of the various hydrometeors has been recently granted. The new method will be developed and tested using the Weather Research and Forecasting (WRF) model and this model’s 4-dimensional variational (4DVAR) data assimilation system. In this study, once again, the focus will be on modeling and prediction of precipitation associated with strong synoptic scale forcing mechanisms and topography.
One of the current ongoing activities is built on a theoretical study performed in collaboration with CIRES scientist Tomislava Vukicevic and FAB scientist John McGinley (Vukicevic et al., 2008). The main focus of this study was a development of a technique that unifies evaluation of the forecast uncertainties produced either by initial conditions or different model versions, or both. The technique consists of first diagnosing the performance of the forecast ensemble which is based on explicit use of the analysis uncertainties, and then optimizing the ensemble forecast using results of the diagnosis. The technique includes explicit evaluation of probabilities which are associated with the Gaussian stochastic representation of both the analysis and forecast. It combines the technique for evaluating the analysis error covariance that was first presented in the Ensemble Transform data assimilation method developed by Bishop et al in 2001 and the standard Monte Carlo approach for computing samples from the known Gaussian distribution. The current activity consists of applying the theoretical approach on “real” data. For this purpose, simulations of atmospheric river events performed by using WRF-ARW model with mixed physics and mixed initial conditions are used.
CIRA researchers actively participated in a Quantitative Precipitation Evaluation (QPE) project funded by the GSD Office of the Director and led by Steven Mullen from Arizona State University. The main goal of this study was to evaluate a possibility of producing a better precipitation analysis by using an ensemble of WRF model forecasts to recover a background error covariance. Preliminary results were presented in a recent seminar and first draft of an official journal publication has been made. CIRA researchers will take part in an extension of this study which will involve an implementation of the experimental radar data available for the HMT 30-31 December 2005 IOP. The experimental x-band radar data have been obtained from PSD scientist David Kingsmill.
One additional HMT activity consisted of taking part in a team effort to test a newly developed observations-based forecast model verification tool by a group of PSD scientists (Dan Gottas, Sara Michelson, Jian-Wen Bao, Paul Neiman, Allen White, Seth Gutman, Marty Ralph, Dave Kingsmill and Tim Schneider) for atmospheric rivers and their impacts on coastal orographic precipitation enhancement. The tool focuses on water vapor flux as a major determinant of orographic precipitation. The water vapor transport is estimated by using wind profilers and GPS-met (Integrated Water Vapor) IWV data.
Working with the California Dept. of Water Resources, we coordinated the HMT datasets that were presented using Science On a Sphere at the California State Fair in late summer 2007. We also gave a number of on-site SOS demos with these and other datasets at the fair. We also helped with the preparation of graphics being shown with Science On a Sphere for an interview of HMT participants by a National Geographic film crew.
3. Comparison of Objectives vs Actual Accomplishments for Report Period:
Our accomplishments for this project compare favorably with the goals projected in the statement of work.
5. Research Linkages/Partnerships/Collaborators:
Jankov, I., P. J. Schultz, C. J. Anderson, and S. E. Koch, 2007: The impact of different physical parameterizations and their interactions on cold season QPF. J. Hydrometeor.8, 1141-1151.
Jankov, I., W. A. Gallus, Jr., M. Segal, and S. E. Koch, 2007: Influence of initial conditions on the WRF-ARW model QPF response to physical parameterizations changes. Wea. Forecasting22, 501-519.
Jankov, I., J. W. Bao, P. J. Neimen, P. J. Schultz, and A. B. White, 2008: Evaluation of microphysical algorithms in WRF-ARW model simulations of atmospheric river events affecting the California coast. European Geosciences Union General Assembly 2008, Vienna, Austria, EGU
Schultz, P., S. C. Albers, C. J. Anderson, D. Birkenheuer, I. Jankov, and J. McGinley, 2008: A computationally efficient method for initializing numerical weather models with explicit representation of moist convection. Submitted to Wea. Forecasting.
Vukicevic, T., I. Jankov, and J. McGinley, 2008: Diagnosis and optimization of ensemble forecasts. Mon. Wea. Rev., 136, pp. 1054-1074.
Yuan, H., J. A. McGinley, P. J. Schultz, C. J. Anderson, and C. Lu, 2008: Short-range precipitation forecasts from time-lagged multimodel ensembles during the HMT-West-2006 campaign. Accepted by J. Hydrometeor.
Yuan, H., C. Lu, J. McGinley, P. Schultz, B. Jamison, L. Wharton, and C. Anderson, 2008: Short-range precipitation forecasts using time-lagged multimodel ensembles. Submitted to Wea. Forecasting