Wind Ramp Events Validation in NWP Forecast Models during the Second Wind Forecast Improvement Project (WFIP2) Using the Ramp Tool and Metric (RT&M)
Explore the source record for details and available documents.
SEARCH · Search NASA
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Carnegie Mellon Environmental Engineering Seminar, Pittsburgh, PA, April 7, 2023
The National Center for Atmospheric Research (NCAR) is pleased to have led a partnership to advance the state-of-the-science of solar power forecasting by designing, developing, building, deploying, testing, and assessing the SunCast™ Solar Power Forecasting System. The project has included cutting edge research, testing in several geographically- and climatologically-diverse high penetration solar utilities and Independent System Operators (ISOs), and wide dissemination of the research results to raise the bar on solar power forecasting technology. The partners include three other national laboratories, six universities, and industry partners. This public-private-academic team has worked in concert to perform use-inspired research to advance solar power forecasting through cutting-edge research to advance both the necessary forecasting technologies and the metrics for evaluating them. The project has culminated in a year-long, full-scale demonstration of provide irradiance and power forecasts to utilities and ISOs to use in their operations. The project focused on providing elements of a value chain, beginning with the weather that causes a deviation from clear sky irradiance and progresses through monitoring and observations, modeling, forecasting, dissemination and communication of the forecasts, interpretation of the forecast, and through decision-making, which produces outcomes that have an economic value. The system has been evaluated using metrics developed specifically for this project, which has provided rich information on model and system performance. Research was accomplished on the very short range (0-6 hours) Nowcasting system as well as on the longer term (6-72 hour) forecasting system. The shortest range forecasts are based on observations in the field. The shortest range system, built by Brookhaven National Laboratory (BNL) and based on Total Sky Imagers (TSIs) is TSICast, which operates on the shortest time scale with a latency of only a few minutes and forecasts that currently go out to about 15 min. This project has facilitated research in improving the hardware and software so that the new high definition cameras deployed at multiple nearby locations allow discernment of the clouds at varying levels and advection according to the winds observed at those levels. Improvements over “smart persistence” are about 29% for even these very short forecasts. StatCast is based on pyranometer data measured at the site as well as concurrent meteorological observations and forecasts. StatCast is based on regime-dependent artificial intelligence forecasting techniques and has been shown to improve on “smart persistence” forecasts by 15-50%. A second category of short-range forecasting systems employ satellite imagery and use that information to discern clouds and their motion, allowing them to project the clouds, and the resulting blockage of irradiance, in time. CIRACast (the system produced by the Cooperative Institute for Atmospheric Research [CIRA] at Colorado State University) was already one of the more advanced cloud motion systems, which is the reason that team was brought to this project. During the project timeframe, the CIRA team was able to advance cloud shadowing, parallax removal, and implementation of better advecting winds at different altitudes. CIRACast shows generally a 25-40% improvement over Smart Persistence between sunrise and approximately 1600 UTC (Coordinated Universal Time) . A second satellite-based system, MADCast (Multi-sensor Advective Diffusive foreCast system), assimilates data from multiple satellite imagers and profilers to assimilate a fully three-dimensional picture of the cloud into the dynamic core of WRF. During 2015, MADCast (provided at least 70% improvement over Smart Persistence, with most of that skill being derived during partly cloudy conditions. That allows advection of the clouds via the Weather Research and Forecasting (WRF) model dynamics directly. After WRF-Solar™ showed initial success, it was also deployed in nowcasting mode with coarser runs out to 6 hours made hourly. It provided improvements on the order of 50-60% over Smart Persistence for forecasts up to 1600 UTC. The advantages of WRF-Solar-Nowcasting and MADCast were then blended to develop the new MAD-WRF model that incorporates the most important features of each of those models, both assimilating satellite cloud fields and using WRF-So far physics to develop and dissipate clouds. MAE improvements for MAD-WRF for forecasts from 3-6 hours are improved over WRF-Solar-Now by 20%. While all the Nowcasting system components by themselves provide improvement over Smart Persistence, the largest benefit is derived when they are smartly blended together by the Nowcasting Integrator to produce an integrated forecast. The development of WRF-Solar™ under this project has provided the first numerical weather prediction (NWP) model specifically designed to meet the needs of irradiance forecasting. The first augmentation improved the solar tracking algorithm to account for deviations associated with the eccentricity of the Earth’s orbit and the obliquity of the Earth. Second, WRF-Solar™ added the direct normal irradiance (DNI) and diffuse (DIF) components from the radiation parameterization to the model output. Third, efficient parameterizations were implemented to either interpolate the irradiance in between calls to the expensive radiative transfer parameterization, or to use a fast radiative transfer code that avoids computing three-dimensional heating rates but provides the surface irradiance. Fourth, a new parameterization was developed to improve the representation of absorption and scattering of radiation by aerosols (aerosol direct effect). A fifth advance is that the aerosols now interact with the cloud microphysics, altering the cloud evolution and radiative properties, an effect that has been traditionally only implemented in atmospheric computationally costly chemistry models. A sixth development accounts for the feedbacks that sub-grid scale clouds produce in shortwave irradiance as implemented in a shallow cumulus parameterization Finally, WRF-Solar™ also allows assimilation of infrared irradiances from satellites to determine the three dimensional cloud field, allowing for an improved initialization of the cloud field that increases the performance of short-range forecasts. We find that WRF-Solar™ can improve clear sky irradiance prediction by 15-80% over a standard version of WRF, depending on location and cloud conditions. In a formal comparison to the NAM baseline, WRF-Solar™ showed improvements in the Day-Ahead forecast of 22-42%. The SunCast™ system requires substantial software engineering to blend all of the new model components as well as existing publically available NWP model runs. To do this we use an expert system for the Nowcasting blender and the Dynamic Integrated foreCast (DICast®) system for the NWP models. These two systems are then blended, we use an empirical power conversion method to convert the irradiance predictions to power, then apply an analog ensemble (AnEn) approach to further tune the forecast as well as to estimate its uncertainty. The AnEn module decreased RMSE (root mean squared error) by 17% over the blended SunCast™ power forecasts and provided skill in the probabilistic forecast with a Brier Skill Score of 0.55. In addition, we have also developed a Gridded Atmospheric Forecast System (GRAFS) in parallel, leveraging cost share funds. An economic evaluation based on Production Cost Modeling in the Public Service Company of Colorado showed that the observed 50% improvement in forecast accuracy will save their customers $819,200 with the projected MW deployment for 2024. Using econometrics, NCAR has scaled this savings to a national level and shown that an annual expected savings for this 50% forecast error reduction ranges from $11M in 2015 to $43M expected in 2040 with increased solar deployment. This amounts to a $455M discounted savings over the 26 year period of analysis.
Four major research objectives were completed over the course of this study. Three of the objectives were to evaluate three, new, state-of-the-art solar irradiance forecasting models. The fourth objective was to improve the California Independent System Operator’s (ISO) load forecasts by integrating behind-the-meter (BTM) PV forecasts. The three, new, state-of-the-art solar irradiance forecasting models included: the infrared (IR) satellite-based cloud motion vector (CMV) model; the WRF-SolarCA model and variants; and the Optimized Deep Machine Learning (ODML)-training model. The first two forecasting models targeted known weaknesses in current operational solar forecasts. They were benchmarked against existing operational numerical weather prediction (NWP) forecasts, visible satellite CMV forecasts, and measured PV plant power production. IR CMV, WRF-SolarCA, and ODML-training forecasting models all improved the forecast to a significant degree. Improvements varied depending on time of day, cloudiness index, and geographic location. The fourth objective was to demonstrate that the California ISO’s load forecasts could be improved by integrating BTM PV forecasts. This objective represented the project’s most exciting and applicable gains. Operational BTM forecasts consisting of 200,000+ individual rooftop PV forecasts were delivered into the California ISO’s real-time automated load forecasting (ALFS) environment. They were then evaluated side-by-side with operational load forecasts with no BTM-treatment. Overall, ALFS-BTM day-ahead (DA) forecasts performed better than baseline ALFS forecasts when compared to actual load data. Specifically, ALFS-BTM DA forecasts were observed to have the largest reduction of error during the afternoon on cloudy days. Shorter term 30 minute-ahead ALFS-BTM forecasts were shown to have less error under all sky conditions, especially during the morning time periods when traditional load forecasts often experience their largest uncertainties. This work culminated in a GO decision being made by the California ISO to include zonal BTM forecasts into its operational load forecasting system. The California ISO’s Manager of Short Term Forecasting, Jim Blatchford, summarized the research performed in this project with the following quote: “The behind-the-meter (BTM) California ISO region forecasting research performed by Clean Power Research and sponsored by the Department of Energy’s SUNRISE program was an opportunity to verify value and demonstrate improved load forecast capability. In 2016, the California ISO will be incorporating the BTM forecast into the Hour Ahead and Day Ahead load models to look for improvements in the overall load forecast accuracy as BTM PV capacity continues to grow.”
The Federal Aviation Administration (FAA) is handling nearly 120,000 flights a day through its Air Traffic Management (ATM) system and air traffic congestion is expected to increse substantially over the next 20 years. Weather-induced impacts to throughput and efficiency are the leading cause of flight delays accounting for 70% of all delays with convective weather accounting for 60% of all weather related delays. To support the Next Generation Air Traffic System goal of operating at 3X current capacity in the NAS, ATC decision support tools are being developed to create advisories to assist controllers in all weather constraints. Initial development of these decision support tools did not integrate information regarding weather constraints such as thunderstorms and relied on an additional system to provide that information. Future Decision Support Tools should move towards an integrated system where weather constraints are factored into the advisory of a Decision Support Tool (DST). Several groups such at NASA-Ames, Lincoln Laboratories, and MITRE are integrating convective weather data with DSTs. A survey of current convective weather forecast and observation data show they span a wide range of temporal and spatial resolutions. Short range convective observations can be obtained every 5 mins with longer range forecasts out to several days updated every 6 hrs. Today, the short range forecasts of less than 2 hours have a temporal resolution of 5 mins. Beyond 2 hours, forecasts have much lower temporal. resolution of typically 1 hour. Spatial resolutions vary from 1km for short range to 40km for longer range forecasts. Improving the accuracy of long range convective forecasts is a major challenge. A report published by the National Research Council states improvements for convective forecasts for the 2 to 6 hour time frame will only be achieved for a limited set of convective phenomena in the next 5 to 10 years. Improved longer range forecasts will be probabilistic as opposed to the deterministic shorter range forecasts. Despite the known low level of confidence with respect to long range convective forecasts, these data are still useful to a DST routing algorithm. It is better to develop an aircraft route using the best information available than no information. The temporally coarse long range forecast data needs to be interpolated to be useful to a DST. A DST uses aircraft trajectory predictions that need to be evaluated for impacts by convective storms. Each time-step of a trajectory prediction n&s to be checked against weather data. For the case of coarse temporal data, there needs to be a method fill in weather data where there is none. Simply using the coarse weather data without any interpolation can result in DST routes that are impacted by regions of strong convection. Increasing the temporal resolution of these data can be achieved but result in a large dataset that may prove to be an operational challenge in transmission and loading by a DST. Currently, it takes about 7mins retrieve a 7mb RUC2 forecast file from NOAA at NASA-Ames Research Center. A prototype NCWF6 1 hour forecast is about 3mb in size. A Six hour NCWFG forecast with a 1hr forecast time-step will be about l8mb (6 x 3mb). A 6 hour NCWF6 forecast with a l5min forecast time-step will be about 7mb (24 x 3mb). Based on the time it takes to retrieve a 7mb RUC2 forecast, it will take approximately 70mins to retrieve a 6 hour NCWF forecast with 15min time steps. Until those issues are addressed, there is a need to develop an algorithm that interpolates between these temporally coarse long range forecasts. This paper describes a method of how to use low temporal resolution probabilistic weather forecasts in a DST. The beginning of this paper is a description of some convective weather forecast and observation products followed by an example of how weather data are used by a DST. The subsequent sections will describe probabilistic forecasts followed by a descrtion of a method to use low temporal resolution probabilistic weather forecasts by providing a relevance value to these data outside of their valid times.
This paper presents a detailed convective forecast accuracy analysis at center and sector levels. The study is aimed to provide more meaningful forecast verification measures to aviation community, as well as to obtain useful information leading to the improvements in the weather translation capacity models. In general, the vast majority of forecast verification efforts over past decades have been on the calculation of traditional standard verification measure scores over forecast and observation data analyses onto grids. These verification measures based on the binary classification have been applied in quality assurance of weather forecast products at the national level for many years. Our research focuses on the forecast at the center and sector levels. We calculate the standard forecast verification measure scores for en-route air traffic centers and sectors first, followed by conducting the forecast validation analysis and related verification measures for weather intensities and locations at centers and sectors levels. An approach to improve the prediction of sector weather coverage by multiple sector forecasts is then developed. The weather severe intensity assessment was carried out by using the correlations between forecast and actual weather observation airspace coverage. The weather forecast accuracy on horizontal location was assessed by examining the forecast errors. The improvement in prediction of weather coverage was determined by the correlation between actual sector weather coverage and prediction. observed and forecasted Convective Weather Avoidance Model (CWAM) data collected from June to September in 2007. CWAM zero-minute forecast data with aircraft avoidance probability of 60% and 80% are used as the actual weather observation. All forecast measurements are based on 30-minute, 60- minute, 90-minute, and 120-minute forecasts with the same avoidance probabilities. The forecast accuracy analysis for times under one-hour showed that the errors in intensity and location for center forecast are relatively low. For example, 1-hour forecast intensity and horizontal location errors for ZDC center were about 0.12 and 0.13. However, the correlation between sector 1-hour forecast and actual weather coverage was weak, for sector ZDC32, about 32% of the total variation of observation weather intensity was unexplained by forecast; the sector horizontal location error was about 0.10. The paper also introduces an approach to estimate the sector three-dimensional actual weather coverage by using multiple sector forecasts, which turned out to produce better predictions. Using Multiple Linear Regression (MLR) model for this approach, the correlations between actual observation and the multiple sector forecast model prediction improved by several percents at 95% confidence level in comparison with single sector forecast.
Background: Short-term forecasts of infectious disease burden can contribute to situational awareness and aid capacity planning. Based on best practice in other fields and recent insights in infectious disease epidemiology, one can maximise the predictive performance of such forecasts if multiple models are combined into an ensemble. Here, we report on the performance of ensembles in predicting COVID-19 cases and deaths across Europe between 08 March 2021 and 07 March 2022. Methods: We used open-source tools to develop a public European COVID-19 Forecast Hub. We invited groups globally to contribute weekly forecasts for COVID-19 cases and deaths reported by a standardised source for 32 countries over the next 1–4 weeks. Teams submitted forecasts from March 2021 using standardised quantiles of the predictive distribution. Each week we created an ensemble forecast, where each predictive quantile was calculated as the equally-weighted average (initially the mean and then from 26th July the median) of all individual models’ predictive quantiles. We measured the performance of each model using the relative Weighted Interval Score (WIS), comparing models’ forecast accuracy relative to all other models. We retrospectively explored alternative methods for ensemble forecasts, including weighted averages based on models’ past predictive performance. Results: Over 52 weeks, we collected forecasts from 48 unique models. We evaluated 29 models’ forecast scores in comparison to the ensemble model. We found a weekly ensemble had a consistently strong performance across countries over time. Across all horizons and locations, the ensemble performed better on relative WIS than 83% of participating models’ forecasts of incident cases (with a total N=886 predictions from 23 unique models), and 91% of participating models’ forecasts of deaths (N=763 predictions from 20 models). Across a 1–4 week time horizon, ensemble performance declined with longer forecast periods when forecasting cases, but remained stable over 4 weeks for incident death forecasts. In every forecast across 32 countries, the ensemble outperformed most contributing models when forecasting either cases or deaths, frequently outperforming all of its individual component models. Among several choices of ensemble methods we found that the most influential and best choice was to use a median average of models instead of using the mean, regardless of methods of weighting component forecast models. Conclusions: Our results support the use of combining forecasts from individual models into an ensemble in order to improve predictive performance across epidemiological targets and populations during infectious disease epidemics. Our findings further suggest that median ensemble methods yield better predictive performance more than ones based on means. Our findings also highlight that forecast consumers should place more weight on incident death forecasts than incident case forecasts at forecast horizons greater than 2 weeks. Funding: AA, BH, BL, LWa, MMa, PP, SV funded by National Institutes of Health (NIH) Grant 1R01GM109718, NSF BIG DATA Grant IIS-1633028, NSF Grant No.: OAC-1916805, NSF Expeditions in Computing Grant CCF-1918656, CCF-1917819, NSF RAPID CNS-2028004, NSF RAPID OAC-2027541, US Centers for Disease Control and Prevention 75D30119C05935, a grant from Google, University of Virginia Strategic Investment Fund award number SIF160, Defense Threat Reduction Agency (DTRA) under Contract No. HDTRA1-19-D-0007, and respectively Virginia Dept of Health Grant VDH-21-501-0141, VDH-21-501-0143, VDH-21-501-0147, VDH-21-501-0145, VDH-21-501-0146, VDH-21-501-0142, VDH-21-501-0148. AF, AMa, GL funded by SMIGE - Modelli statistici inferenziali per governare l'epidemia, FISR 2020-Covid-19 I Fase, FISR2020IP-00156, Codice Progetto: PRJ-0695. AM, BK, FD, FR, JK, JN, JZ, KN, MG, MR, MS, RB funded by Ministry of Science and Higher Education of Poland with grant 28/WFSN/2021 to the University of Warsaw. BRe, CPe, JLAz funded by Ministerio de Sanidad/ISCIII. BT, PG funded by PERISCOPE European H2020 project, contract number 101016233. CP, DL, EA, MC, SA funded by European Commission - Directorate-General for Communications Networks, Content and Technology through the contract LC-01485746, and Ministerio de Ciencia, Innovacion y Universidades and FEDER, with the project PGC2018-095456-B-I00. DE., MGu funded by Spanish Ministry of Health / REACT-UE (FEDER). DO, GF, IMi, LC funded by Laboratory Directed Research and Development program of Los Alamos National Laboratory (LANL) under project number 20200700ER. DS, ELR, GG, NGR, NW, YW funded by National Institutes of General Medical Sciences (R35GM119582; the content is solely the responsibility of the authors and does not necessarily represent the official views of NIGMS or the National Institutes of Health). FB, FP funded by InPresa, Lombardy Region, Italy. HG, KS funded by European Centre for Disease Prevention and Control. IV funded by Agencia de Qualitat i Avaluacio Sanitaries de Catalunya (AQuAS) through contract 2021-021OE. JDe, SMo, VP funded by Netzwerk Universitatsmedizin (NUM) project egePan (01KX2021). JPB, SH, TH funded by Federal Ministry of Education and Research (BMBF; grant 05M18SIA). KH, MSc, YKh funded by Project SaxoCOV, funded by the German Free State of Saxony. Presentation of data, model results and simulations also funded by the NFDI4Health Task Force COVID-19 ( https://www.nfdi4health.de/task-force-covid-19-2 ) within the framework of a DFG-project (LO-342/17-1). LP, VE funded by Mathematical and Statistical modelling project (MUNI/A/1615/2020), Online platform for real-time monitoring, analysis and management of epidemic situations (MUNI/11/02202001/2020); VE also supported by RECETOX research infrastructure (Ministry of Education, Youth and Sports of the Czech Republic: LM2018121), the CETOCOEN EXCELLENCE (CZ.02.1.01/0.0/0.0/17-043/0009632), RECETOX RI project (CZ.02.1.01/0.0/0.0/16-013/0001761). NIB funded by Health Protection Research Unit (grant code NIHR200908). SAb, SF funded by Wellcome Trust (210758/Z/18/Z).
Wind resource assessment and wind power forecasting are used in research and industry to anticipate future power output at scales ranging from individual wind turbines to entire wind farms. Probabilistic day-ahead wind forecasting is useful for anticipating how a wind farm could potentially participate in the day-ahead market by providing upper and lower bounds for expected power generation, thus informing grid operators of its uncertainty. Understanding this uncertainty is part of a larger project focused on building a platform that combines efforts in weather forecasting, aerodynamic and economic modeling to create maximum value of a wind plant to better provide services to the grid. This effort is also known as the Atmosphere to Electrons to Grid (A2E2G) project. One method for producing a probabilistic forecast is through the analog ensemble approach (Delle Monache et al., 2011). This method leverages historical forecasts and their corresponding observations as a training data set from which future forecasts can be made. For some future forecast, the most similar historical forecasts (analogs) are identified on a regular time basis such as once per a 3-hour window. The most similar analogs, based on a metric such as root mean square error (RMSE), are recorded and their corresponding verifying observations are used as an ensemble member for this future forecast. Prior work in this area demonstrates improvements over raw Numerical Weather Prediction (NWP) forecasts and shows skill similar to techniques such as logistic regression and machine learning (Delle Monache et al., 2013; Alessandrini et al., 2015). Here, we take the High-Resolution Rapid Refresh model (HRRR) day-ahead forecast (0-36 hours) to create a probabilistic day-ahead forecast using an analog ensemble approach. The HRRR has an hourly temporal resolution, with a spatial resolution of 3 km. The 12 UTC HRRR model run is downloaded every day for one year from August 2019 - July 2020, with the first 11 months serving as a bank of analogs from which the forecasting algorithm can create a probabilistic forecast. Once downloaded, the original HRRR forecast is temporally interpolated to 5-minutes, aligning with both the temporal resolution of the observations as well as the timescale relevant for day-ahead power forecasts. The forecast is validated at the M2 tower at the Flatirons Campus of the National Renewable Energy Laboratory (NREL) at a typical wind turbine height of 80 m. Variables such as wind speed, wind direction, and turbulence intensity are incorporated into the probabilistic forecast model and weighted according to their relative importance to the forecast. Based on metrics such as mean bias error (MBE), mean absolute error (MAE), and root mean square error, the analog ensemble forecast outperforms the raw HRRR forecast during the testing period of July 2020. Figure 1 illustrates an example day-ahead forecast compared against the verifying observations. The general variability and ramps are captured throughout the day, with potential to further improve the analog ensemble model through machine learning techniques.
The expected peak wind speed for the day is an important element in the daily morning forecast for ground and space launch operations at Kennedy Space Center (KSC) and Cape Canaveral Air Force Station (CCAFS). The 45th Weather Squadron (45 WS) must issue forecast advisories for KSC/CCAFS when they expect peak gusts for >= 25, >= 35, and >= 50 kt thresholds at any level from the surface to 300 ft. In Phase I of this task, the 45 WS tasked the Applied Meteorology Unit (AMU) to develop a cool-season (October - April) tool to help forecast the non-convective peak wind from the surface to 300 ft at KSC/CCAFS. During the warm season, these wind speeds are rarely exceeded except during convective winds or under the influence of tropical cyclones, for which other techniques are already in use. The tool used single and multiple linear regression equations to predict the peak wind from the morning sounding. The forecaster manually entered several observed sounding parameters into a Microsoft Excel graphical user interface (GUI), and then the tool displayed the forecast peak wind speed, average wind speed at the time of the peak wind, the timing of the peak wind and the probability the peak wind will meet or exceed 35, 50 and 60 kt. The 45 WS customers later dropped the requirement for >= 60 kt wind warnings. During Phase II of this task, the AMU expanded the period of record (POR) by six years to increase the number of observations used to create the forecast equations. A large number of possible predictors were evaluated from archived soundings, including inversion depth and strength, low-level wind shear, mixing height, temperature lapse rate and winds from the surface to 3000 ft. Each day in the POR was stratified in a number of ways, such as by low-level wind direction, synoptic weather pattern, precipitation and Bulk Richardson number. The most accurate Phase II equations were then selected for an independent verification. The Phase I and II forecast methods were compared using an independent verification data set. The two methods were compared to climatology, wind warnings and advisories issued by the 45 WS, and North American Mesoscale (NAM) model (MesoNAM) forecast winds. The performance of the Phase I and II methods were similar with respect to mean absolute error. Since the Phase I data were not stratified by precipitation, this method's peak wind forecasts had a large negative bias on days with precipitation and a small positive bias on days with no precipitation. Overall, the climatology methods performed the worst while the MesoNAM performed the best. Since the MesoNAM winds were the most accurate in the comparison, the final version of the tool was based on the MesoNAM winds. The probability the peak wind will meet or exceed the warning thresholds were based on the one standard deviation error bars from the linear regression. For example, the linear regression might forecast the most likely peak speed to be 35 kt and the error bars used to calculate that the probability of >= 25 kt = 76%, the probability of >= 35 kt = 50%, and the probability of >= 50 kt = 19%. The authors have not seen this application of linear regression error bars in any other meteorological applications. Although probability forecast tools should usually be developed with logistic regression, this technique could be easily generalized to any linear regression forecast tool to estimate the probability of exceeding any desired threshold . This could be useful for previously developed linear regression forecast tools or new forecast applications where statistical analysis software to perform logistic regression is not available. The tool was delivered in two formats - a Microsoft Excel GUI and a Tool Command Language/Tool Kit (Tcl/Tk) GUI in the Meteorological Interactive Data Display System (MIDDS). The Microsoft Excel GUI reads a MesoNAM text file containing hourly forecasts from 0 to 84 hours, from one model run (00 or 12 UTC). The GUI then displays e peak wind speed, average wind speed, and the probability the peak wind will meet or exceed the 25-, 35- and 50-kt thresholds. The user can display the Day-1 through Day-3 peak wind forecasts, and separate forecasts are made for precipitation and non-precipitation days. The MIDDS GUI uses data from the NAM and Global Forecast System (GFS), instead of the MesoNAM. It can display Day-1 and Day-2 forecasts using NAM data, and Day-1 through Day-5 forecasts using GFS data. The timing of the peak wind is not displayed, since the independent verification showed that none of the forecast methods performed significantly better than climatology. The forecaster should use the climatological timing of the peak wind (2248 UTC) as a first guess and then adjust it based on the movement of weather features.
Abstract Background West Nile virus (WNV) is the most common cause of mosquito-borne disease in the continental USA, with an average of ~1200 severe, neuroinvasive cases reported annually from 2005 to 2021 (range 386–2873). Despite this burden, efforts to forecast WNV disease to inform public health measures to reduce disease incidence have had limited success. Here, we analyze forecasts submitted to the 2022 WNV Forecasting Challenge, a follow-up to the 2020 WNV Forecasting Challenge. Methods Forecasting teams submitted probabilistic forecasts of annual West Nile virus neuroinvasive disease (WNND) cases for each county in the continental USA for the 2022 WNV season. We assessed the skill of team-specific forecasts, baseline forecasts, and an ensemble created from team-specific forecasts. We then characterized the impact of model characteristics and county-specific contextual factors (e.g., population) on forecast skill. Results Ensemble forecasts for 2022 anticipated a season at or below median long-term WNND incidence for nearly all (> 99%) counties. More counties reported higher case numbers than anticipated by the ensemble forecast median, but national caseload (826) was well below the 10-year median (1386). Forecast skill was highest for the ensemble forecast, though the historical negative binomial baseline model and several team-submitted forecasts had similar forecast skill. Forecasts utilizing regression-based frameworks tended to have more skill than those that did not and models using climate, mosquito surveillance, demographic, or avian data had less skill than those that did not, potentially due to overfitting. County-contextual analysis showed strong relationships with the number of years that WNND had been reported and permutation entropy (historical variability). Evaluations based on weighted interval score and logarithmic scoring metrics produced similar results. Conclusions The relative success of the ensemble forecast, the best forecast for 2022, suggests potential gains in community ability to forecast WNV, an improvement from the 2020 Challenge. Similar to the previous challenge, however, our results indicate that skill was still limited with general underprediction despite a relative low incidence year. Potential opportunities for improvement include refining mechanistic approaches, integrating additional data sources, and considering different approaches for areas with and without previous cases. Graphical Abstract
The expected peak wind speed of the day is an important forecast element in the 45th Weather Squadron's (45 WS) daily 24-Hour and Weekly Planning Forecasts. The forecasts are used for ground and space launch operations at the Kennedy Space Center (KSC) and Cape Canaveral Air Force Station (CCAFS). The 45 WS also issues wind advisories for KSC/CCAFS when they expect wind gusts to meet or exceed 25 kt, 35 kt and 50 kt thresholds at any level from the surface to 300 ft. The 45 WS forecasters have indicated peak wind speeds are challenging to forecast, particularly in the cool season months of October - April. In Phase I of this task, the Applied Meteorology Unit (AMU) developed a tool to help the 45 WS forecast non-convective winds at KSC/CCAFS for the 24-hour period of 0800 to 0800 local time. The tool was delivered as a Microsoft Excel graphical user interface (GUI). The GUI displayed the forecast of peak wind speed, 5-minute average wind speed at the time of the peak wind, timing of the peak wind and probability the peak speed would meet or exceed 25 kt, 35 kt and 50 kt. For the current task (Phase II ), the 45 WS requested additional observations be used for the creation of the forecast equations by expanding the period of record (POR). Additional parameters were evaluated as predictors, including wind speeds between 500 ft and 3000 ft, static stability classification, Bulk Richardson Number, mixing depth, vertical wind shear, temperature inversion strength and depth and wind direction. Using a verification data set, the AMU compared the performance of the Phase I and II prediction methods. Just as in Phase I, the tool was delivered as a Microsoft Excel GUI. The 45 WS requested the tool also be available in the Meteorological Interactive Data Display System (MIDDS). The AMU first expanded the POR by two years by adding tower observations, surface observations and CCAFS (XMR) soundings for the cool season months of March 2007 to April 2009. The POR was expanded again by six years, from October 1996 to April 2002, by interpolating 1000-ft sounding data to 100-ft increments. The Phase II developmental data set included observations for the cool season months of October 1996 to February 2007. The AMU calculated 68 candidate predictors from the XMR soundings, to include 19 stability parameters, 48 wind speed parameters and one wind shear parameter. Each day in the data set was stratified by synoptic weather pattern, low-level wind direction, precipitation and Richardson Number, for a total of 60 stratification methods. Linear regression equations, using the 68 predictors and 60 stratification methods, were created for the tool's three forecast parameters: the highest peak wind speed of the day (PWSD), 5-minute average speed at the same time (A WSD), and timing of the PWSD. For PWSD and A WSD, 30 Phase II methods were selected for evaluation in the verification data set. For timing of the PWSD, 12 Phase\I methods were selected for evaluation. The verification data set contained observations for the cool season months of March 2007 to April 2009. The data set was used to compare the Phase I and II forecast methods to climatology, model forecast winds and wind advisories issued by the 45 WS. The model forecast winds were derived from the 0000 and 1200 UTC runs of the 12-km North American Mesoscale (MesoNAM) model. The forecast methods that performed the best in the verification data set were selected for the Phase II version of the tool. For PWSD and A WSD, linear regression equations based on MesoNAM forecasts performed significantly better than the Phase I and II methods. For timing of the PWSD, none of the methods performed significantly bener than climatology. The AMU then developed the Microsoft Excel and MIDDS GUls. The GUIs display the forecasts for PWSD, AWSD and the probability the PWSD will meet or exceed 25 kt, 35 kt and 50 kt. Since none of the prediction methods for timing of the PWSD performed significantly better thanlimatology, the tool no longer displays this predictand. The Excel and MIDDS GUIs display forecasts for Day-I to Day-3 and Day-I to Day-5, respectively. The Excel GUI uses MesoNAM forecasts as input, while the MIDDS GUI uses input from the MesoNAM and Global Forecast System model. Based on feedback from the 45 WS, the AMU added the daily average wind speed from 30 ft to 60 ft to the tool, which is one of the parameters in the 24-Hour and Weekly Planning Forecasts issued by the 45 WS. In addition, the AMU expanded the MIDDS GUI to include forecasts out to Day-7.
The path toward realizing the potential of seasonal forecasting and its socioeconomic benefits relies on improving general circulation model (GCM) based dynamical forecast systems. To improve dynamical seasonal forecasts, it is crucial to set up forecast benchmarks, and clarify forecast limitations posed by model initialization errors, formulation deficiencies, and internal climate variability. With huge costs in generating large forecast ensembles, and limited observations for forecast verification, the seasonal forecast benchmarking and diagnosing task proves challenging. Here, we develop a probabilistic deep learning-based statistical forecast methodology, drawing on a wealth of climate simulations to enhance seasonal forecast capability and forecast diagnosis. By explicitly modeling the internal climate variability and GCM formulation differences, the proposed Conditional Generative Forecasting (CGF) methodology enables bypassing crucial barriers in dynamical forecast, and offers a top-down viewpoint to examine how complicated GCMs encode the seasonal predictability information. We apply the CGF methodology for global seasonal forecast of precipitation and 2 m air temperature, based on a unique data set consisting 52,201 years of climate simulation. Results show that the CGF methodology can faithfully represent the seasonal predictability information encoded in GCMs. We successfully apply this learned relationship in real-world seasonal forecast, achieving competitive performance compared to dynamical forecasts. Using this CGF as benchmark, we reveal the impact of insufficient forecast spread sampling that limits the skill of the considered dynamical forecast system. Finally, we introduce different strategies for composing ensembles using the CGF methodology, highlighting the potential for leveraging the strengths of multiple GCMs to achieve advantgeous seasonal forecast.
This paper presents a modeling framework for real-time decision support for irrigation scheduling using the National Oceanic and Atmospheric Administration's (NOAA's) probabilistic rainfall forecasts. The forecasts and their probability distributions are incorporated into a simulation-optimization modeling framework. In this study, modeling irrigation is determined by a stochastic optimization program based on the simulated soil moisture and crop water-stress status and the forecasted rainfall for the next 1-7 days. The modeling framework is applied to irrigated corn in Mason County, Illinois. It is found that there is ample potential to improve current farmers practices by simply using the proposed simulation-optimization framework, which uses the present soil moisture and crop evapotranspiration information even without any forecasts. It is found that the values of the forecasts vary across dry, normal, and wet years. More significant economic gains are found in normal and wet years than in dry years under the various forecast horizons. To mitigate drought effect on crop yield through irrigation, medium- or long-term climate predictions likely play a more important role than short-term forecasts. NOAA's imperfect 1-week forecast is still valuable in terms of both profit gain and water saving. Compared with the no-rain forecast case, the short-term imperfect forecasts could lead to additional 2.4-8.5% gain in profit and 11.0-26.9% water saving. However, the performance of the imperfect forecast is only slightly better than the ensemble weather forecast based on historical data and slightly inferior to the perfect forecast. It seems that the 1-week forecast horizon is too limited to evaluate the role of the various forecast scenarios for irrigation scheduling, which is actually a seasonal decision issue. For irrigation scheduling, both the forecast quality and the length of forecast time horizon matter. Thus, longer forecasts might be necessary to evaluate the role of forecasts for irrigation scheduling in a more effective way.
Prior to the COVID-19 pandemic, US hospitals relied on static projections of future trends for long-term planning and were only beginning to consider forecasting methods for short-term planning of staffing and other resources. With the overwhelming burden imposed by COVID-19 on the health care system, an emergent need exists to accurately forecast hospitalization needs within an actionable timeframe. Our goal was to leverage an existing COVID-19 case and death forecasting tool to generate the expected number of concurrent hospitalizations, occupied intensive care unit (ICU) beds, and in-use ventilators 1 day to 4 weeks in the future for New Mexico and each of its five health regions. We developed a probabilistic model that took as input the number of new COVID-19 cases for New Mexico from Los Alamos National Laboratory’s COVID-19 Forecasts Using Fast Evaluations and Estimation tool, and we used the model to estimate the number of new daily hospital admissions 4 weeks into the future based on current statewide hospitalization rates. The model estimated the number of new admissions that would require an ICU bed or use of a ventilator and then projected the individual lengths of hospital stays based on the resource need. By tracking the lengths of stay through time, we captured the projected simultaneous need for inpatient beds, ICU beds, and ventilators. We used a postprocessing method to adjust the forecasts based on the differences between prior forecasts and the subsequent observed data. Thus, we ensured that our forecasts could reflect a dynamically changing situation on the ground. Forecasts made between September 1 and December 9, 2020, showed variable accuracy across time, health care resource needs, and forecast horizon. Forecasts made in October, when new COVID-19 cases were steadily increasing, had an average accuracy error of 20.0%, while the error in forecasts made in September, a month with low COVID-19 activity, was 39.7%. Across health care use categories, state-level forecasts were more accurate than those at the regional level. Although the accuracy declined as the forecast was projected further into the future, the stated uncertainty of the prediction improved. Forecasts were within 5% of their stated uncertainty at the 50% and 90% prediction intervals at the 3- to 4-week forecast horizon for state-level inpatient and ICU needs. However, uncertainty intervals were too narrow for forecasts of state-level ventilator need and all regional health care resource needs. Real-time forecasting of the burden imposed by a spreading infectious disease is a crucial component of decision support during a public health emergency. Our proposed methodology demonstrated utility in providing near-term forecasts, particularly at the state level. This tool can aid other stakeholders as they face COVID-19 population impacts now and in the future.
In this study, we develop a forecast-based adaptive control framework for Oroville reservoir, California, to assess the value of seasonal and inter-annual forecasts for reservoir operation.We use an Ensemble Streamflow Prediction (ESP) approach to generate retrospective, one-year-long streamflow forecasts based on the Variable Infiltration Capacity hydrology model. The optimal sequence of daily release decisions from the reservoir is then determined by Model Predictive Control, a flexible and adaptive optimization scheme.We assess the forecast value by comparing system performance based on the ESP forecasts with that based on climatology and a perfect forecast. In addition, we evaluate system performance based on a synthetic forecast, which is designed to isolate the contribution of seasonal and inter-annual forecast skill to the overall value of the ESP forecasts.Using the same ESP forecasts, we generalize our results by evaluating forecast value as a function of forecast skill, reservoir features, and demand. Our results show that perfect forecasts are valuable when the water demand is high and the reservoir is sufficiently large to allow for annual carry-over. Conversely, ESP forecast value is highest when the reservoir can shift water on a seasonal basis.On average, for the system evaluated here, the overall ESP value is 35% less than the perfect forecast value. The inter-annual component of the ESP forecast contributes 20-60% of the total forecast value. Improvements in the seasonal component of the ESP forecast would increase the overall ESP forecast value between 15 and 20%.
Wind power forecasting (WPF) provides important inputs to power system operators and electricity market participants. It is therefore not surprising that WPF has attracted increasing interest within the electric power industry. In this report, we document our research on improving statistical WPF algorithms for point, uncertainty, and ramp forecasting. Below, we provide a brief introduction to the research presented in the following chapters. For a detailed overview of the state-of-the-art in wind power forecasting, we refer to [1]. Our related work on the application of WPF in operational decisions is documented in [2]. Point forecasts of wind power are highly dependent on the training criteria used in the statistical algorithms that are used to convert weather forecasts and observational data to a power forecast. In Chapter 2, we explore the application of information theoretic learning (ITL) as opposed to the classical minimum square error (MSE) criterion for point forecasting. In contrast to the MSE criterion, ITL criteria do not assume a Gaussian distribution of the forecasting errors. We investigate to what extent ITL criteria yield better results. In addition, we analyze time-adaptive training algorithms and how they enable WPF algorithms to cope with non-stationary data and, thus, to adapt to new situations without requiring additional offline training of the model. We test the new point forecasting algorithms on two wind farms located in the U.S. Midwest. Although there have been advancements in deterministic WPF, a single-valued forecast cannot provide information on the dispersion of observations around the predicted value. We argue that it is essential to generate, together with (or as an alternative to) point forecasts, a representation of the wind power uncertainty. Wind power uncertainty representation can take the form of probabilistic forecasts (e.g., probability density function, quantiles), risk indices (e.g., prediction risk index) or scenarios (with spatial and/or temporal dependence). Statistical approaches to uncertainty forecasting basically consist of estimating the uncertainty based on observed forecasting errors. Quantile regression (QR) is currently a commonly used approach in uncertainty forecasting. In Chapter 3, we propose new statistical approaches to the uncertainty estimation problem by employing kernel density forecast (KDF) methods. We use two estimators in both offline and time-adaptive modes, namely, the Nadaraya-Watson (NW) and Quantilecopula (QC) estimators. We conduct detailed tests of the new approaches using QR as a benchmark. One of the major issues in wind power generation are sudden and large changes of wind power output over a short period of time, namely ramping events. In Chapter 4, we perform a comparative study of existing definitions and methodologies for ramp forecasting. We also introduce a new probabilistic method for ramp event detection. The method starts with a stochastic algorithm that generates wind power scenarios, which are passed through a high-pass filter for ramp detection and estimation of the likelihood of ramp events to happen. The report is organized as follows: Chapter 2 presents the results of the application of ITL training criteria to deterministic WPF; Chapter 3 reports the study on probabilistic WPF, including new contributions to wind power uncertainty forecasting; Chapter 4 presents a new method to predict and visualize ramp events, comparing it with state-of-the-art methodologies; Chapter 5 briefly summarizes the main findings and contributions of this report.
Existing frameworks to forecast time-series photovoltaic (PV) output power and consumer load for microgrid operations and controls assume a near-continuous availability of real-time input features from the field assets such as PV inverters, energy meters, and weather station. These incoming data points are used to periodically retrain models and update forecast snapshots over a moving horizon window, be it one hour-ahead, one-day ahead, or one-week ahead. However, such frameworks are not resilient to disruptions in data availability caused by losses in communications between the field sensors and data loggers. Hence, there is a need for programs that assume no availability of real-time microgrid asset data and still make reliable forecasts that can be used for decision-making. Such programs would be apt to function in extreme weather events such as hurricanes and would use lightweight recursive time-series models to independently forecast solar irradiance and ambient temperature, then compute PV power from those forecasts, as well as independently forecast consumer load. The codebase performs forecasting for the scenario of when the microgrid does not have a reliable access to forecasts or real-time observations of solar irradiance (I) and ambient temperature (AT) and load (Load) to be able to adequately forecast, in real-time, the PV power production or a business' load. In this case, using historical values of PV power and load, a univariate forecasting of generation and consumption are respectively made. The use-case in particular has two sub-scenarios: one, a normal 7-day ahead forecast where the unavailability of real-time data is assumed due to infrastructure issues such as loss of communication or sensor maintenance or service downtimes. Whereas a hurricane-caused unavailability of real-time data requires a second model trained specifically on historical hurricane days to be able to capture the extreme day behavior of generation in particular, and load if applicable. A gradient boosted regression tree comprises an ensemble of additive models that map between the input of historical values (be it irradiance, temperature, or load) and their corresponding output forecasts of a given horizon such that the individual learner predictions are summed up over the total number of such learners in the ensemble to produce an aggregate forecast. A weighting mechanism is applied to the training data in each iteration, where actual and forecast values are compared to penalize incorrect forecasts by increasing the weight and reducing it to reward correct forecasts. The code's benefits are that it: (a) accounts for a contingency where communication loss renders newly measured real-time data unavailable for model tuning and snapshot updates; (b) presents blind forecasting that recursively determines the next time-step value in a horizon using the forecast of the same attribute from a prior step; and (c) employs lightweight models that, once trained, can reliably generalize for different horizons, which make them suitable for enhancing the resilience of field microgrids prone to extreme events that encounter disruptions to data availability.