Forecast Projections Research Articles

Time series analysis and forecasting of the number of canine rabies confirmed cases in Thailand based on national-level surveillance data

IntroductionRabies, a deadly zoonotic viral disease, accounts for over 50,000 fatalities globally each year. This disease predominantly plagues developing nations, with Thailand being no exception. In the current global landscape, concerted efforts are being mobilized to curb human mortalities attributed to animal-transmitted rabies. For strategic allocation and optimization of resources, sophisticated and accurate forecasting of rabies incidents is imperative. This research aims to determine temporal patterns, and seasonal fluctuations, and project the incidence of canine rabies throughout Thailand, using various time series techniques.MethodsMonthly total laboratory-confirmed rabies cases data from January 2013 to December 2022 (full dataset) were split into the training dataset (January 2013 to December 2021) and the test dataset (January to December 2022). Time series models including Seasonal Autoregressive Integrated Moving Average (SARIMA), Neural Network Autoregression (NNAR), Error Trend Seasonality (ETS), the Trigonometric Exponential Smoothing State-Space Model with Box-Cox transformation, ARMA errors, Trend and Seasonal components (TBATS), and Seasonal and Trend Decomposition using Loess (STL) were used to analyze the training dataset and the full dataset. The forecast values obtained from the time series models applied to the training dataset were compared with the actual values from the test dataset to determine their predictive performance. Furthermore, the forecast projections from January 2023 to December 2025 were generated from models applied to the full dataset.ResultsThe findings revealed a total of 4,678 confirmed canine rabies cases during the study duration, with apparent seasonality in the data. Among the models tested with the test dataset, TBATS exhibited superior predictive accuracy, closely trailed by the SARIMA model. Based on the full dataset, TBATS projections suggest an annual average of approximately 285 canine rabies cases for the years 2023 to 2025, translating to a monthly average of 23 cases (range: 18–30). In contrast, SARIMA projections averaged 277 cases annually (range: 208–214).DiscussionThis research offers a new perspective on disease forecasting through advanced time series methodologies. The results should be taken into consideration when planning and conducting rabies surveillance, prevention, and control activities.

Predicting Peak Wind Gusts during Specific Weather Types with the Meteorologically Stratified Gust Factor Model

Abstract Wind gusts present challenges to operational meteorologists, both to forecast accurately and also to verify. Strong wind gusts can damage structures and create costly risks for diverse industrial sectors. The meteorologically stratified gust factor (MSGF) model incorporates site-specific gust factors (the ratio of peak wind gust to mean wind speed) with wind speed and direction forecast guidance. The MSGF model has previously been shown to be a viable operational tool that exhibits skill (improvement over climatology) in forecasting peak wind gusts. This study assesses the performance characteristics of the MSGF model by evaluating peak gust predictions during several types of gust-producing weather phenomena. Peak wind gusts were prepared and verified for seven specific weather conditions over an 8-yr period at 16 sites across the United States. When coupled with two forms of model output statistics (MOS) wind guidance, the MSGF model generally shows skill in predicting peak wind gusts at forecast projections ranging from 6 to 72 h. The model performed best during high pressure and nocturnal conditions and was also skillful during conditions involving snow. The model did not perform well during the “rain with thunder” weather type. The MSGF model is a viable tool for the operational prediction of peak gusts for most gust-producing weather types. Significance Statement Wind gusts are an important and potentially costly environmental hazard. Wind gusts affect many industrial sectors, including transportation, power generation, forestry, construction, and insurance, but predicting gusts remains a challenging component of weather forecasting. Recent studies have demonstrated that the meteorologically stratified gust factor (MSGF) model shows skill in predicting peak gusts. This study shows that the MSGF model is skillful at predicting peak gusts during specific types of gust-producing weather phenomena at forecast projections up to 72 h, providing further confirmation that the MSGF model is a viable tool for the operational prediction of peak gusts.

The association between high temperature and hospitalisations for cardiovascular diseases under climate change scenarios in Ho Chi Minh City

Global warming is anticipated to induce an increase in the frequency and intensity of hot days and heatwaves, which ultimately have effects on public health. The study aimed to identify the impacts of high temperature and developing climate forecasting projections focused on cardiovascular causes in Ho Chi Minh city (HCMC). The projections were built up based on updated climate scenarios in HCMC. Poisson regression model combined with a distributed lag non-linear model (DLNM) was applied. The forecasting model for cardiovascular causes was performed by using the Global Climate Model (GCM) and Yasushi Honda model. Result showed that the average daily temperature induces an increase in the risk of hospitalisation, in which temperature below 25.7°C reduced number of patients due to cardiovascular disease, meanwhile temperature above 25.7°C has increased hospitalisations. Heat waves over 31°C had the strongest impact on the > 60 years old elderly people after 5 days lag and decreased its impact consecutively to age groups of 41 to 60, 16 to 40, and less than 15 years old. The incremental prediction for the hospitalised cardiovascular disease cases based on the RCP4.5 scenario was 79,713 cases and based on the RCP8.5 scenario was 81,362 cases, respectively.

Investigating the Application of a Transportation Energy Consumption Prediction Model for Urban Planning Scenarios in Machine Learning and Shapley Additive Explanations Method

Accurate forecasts of future energy usage are an important step towards reaching carbon mitigation commitments for city policymakers. Beyond identifying sources of emission intensity for a region, the forecast mechanism must be capable of compensating for gaps in available data and of accounting for the uncertainties behind the dynamics of an urban system. By considering a range of possible scenarios, the prediction model can identify recurring sources of high energy consumption and fine-tune areas of priority with incoming data. This paper considers the impact of predicted shifts in demographic and economic trends for the region on transportation energy consumption. The transportation energy use model is formulated from the Delaware Valley Regional Planning Commission (DVRPC) open-source Household Travel Survey (HTS). Based on these data inputs, a Machine Learning (ML) algorithm is implemented in the form of an Extreme Gradient Boosting (XGBoost) model to estimate energy consumption with a corresponding SHapley Additive exPlanations (SHAP) analysis of feature contribution. From this, a synthetic population is produced using the ML outputs and marginal sums with data from the Census Bureau’s American Community Survey (ACS) to estimate energy consumption for the region. The results indicate that shifting dominant travel modes and income distribution in accordance with the Enduring Urbanism forecast projections led to a decrease in household transportation energy use. Moreover, additional analysis of the model output demonstrates that changes in energy use depend strongly on geographic area and income group.

Meteorological and Climate Modelling Services Tailored to Viticulturists

Grape production is likewise inherently interconnected to climate and weather, and, although grapes may grow worldwide, premium wine-grape production occurs in Mediterranean-like climate ranges. Changes in climate and weather patterns are threatening premium wine-grapes, directly affecting the European wine industry. This is because grapevines are extremely sensitive to their surrounding environment, with seasonal variations in yield much higher than other common crops, such as cereals. With a view to making South European wine industry resilient to climate change, VISCA (Vineyards Integrated Smart Climate Application) project has deployed a Climate Service (CS) Decision Support System (DSS) tool that provides to wine producers with well-founded information to be able to apply correctly adaptation strategies on specific grape varieties and locations, and to achieve optimum production results (e.g., yield and quantity). In this paper we show the meteorological, seasonal and climatic models and data sets used to answer the viticulturist needs; from short-term and mid-term forecast to seasonal forecast and climate projections.

Influence of habitat availability and fire disturbance on a northern range boundary

AbstractAimNon‐climatic constraints on species northern range boundaries are often overlooked in attempts to predict climate‐induced range shifts. Here, we examined the effects of habitat availability and fire disturbance on the distribution of a species that transitions from being common to being found only in marginal populations at the northern boundary of its range.LocationNorth‐western Quebec, Canada (46–51° N and 74–79° W).Taxon : Eastern white cedar (Thuja occidentalis L.)MethodsWe used forest inventory data (n = 4,987) to characterize white‐cedar habitat based on edaphic and topographic conditions at sampled sites along a 600‐km latitudinal gradient. Non‐metric multidimensional scaling was used to assess habitat similarity of sites in the south, where white‐cedar stands are abundant, and sites in the north, where white‐cedar stands are rare. We constructed ensemble white cedar distribution models based on habitat variables in the south and compared ensemble forecast projections of white cedar in the north with observed occurrences to determine if habitat availability was limiting. We independently estimated the age of white‐cedar stands and adjacent stands without white cedar along the gradient. ANOVA was performed to test the age difference between white‐cedar and adjacent stands to determine whether the location of white‐cedar stands was influenced by disturbance, primarily stand‐replacing fire.ResultsHabitat availability was not limiting the distribution of eastern white cedar at its northern range boundary. White cedar did not occupy most sites with suitable habitat in the north, suggesting that other factors prevent white cedar from establishing more stands northward. White‐cedar stands were older than adjacent stands without white cedar all along the gradient, but the difference was more pronounced in the north. This suggests that white‐cedar stands in the north are restricted to undisturbed areas.Main conclusionsFire disturbance, more than habitat availability, limits the distribution of white cedar at its northern range boundary. Projections of white cedar distribution under climate change that ignore fire could overestimate the ability of warming temperatures to extend its northern range limit.

Forecasting Peak Wind Gusts Using Meteorologically Stratified Gust Factors and MOS Guidance

Abstract Gust prediction is an important element of weather forecasting services, yet reliable methods remain elusive. Peak wind gusts estimated by the meteorologically stratified gust factor (MSGF) model were evaluated at 15 locations across the United States during 2010–17. This model couples gust factors, site-specific climatological measures of “gustiness,” with wind speed and direction forecast guidance. The model was assessed using two forms of model output statistics (MOS) guidance at forecast projections ranging from 1 to 72 h. At 11 of 15 sites the MSGF model showed skill (improvement over climatology) in predicting peak gusts out to projections of 72 h. This has important implications for operational wind forecasting because the method can be utilized at any location for which the meteorologically stratified gust factors have been determined. During particularly windy conditions the MSGF model exhibited skill in predicting peak gusts at forecast projections ranging from 6 to 72 h at roughly half of the sites analyzed. Site characteristics and local wind climatologies were shown to exert impacts on gust factor model performance. The MSGF method represents a viable option for the operational prediction of peak wind gusts, although model performance will be sensitive to the quality of the necessary wind speed and direction forecasts.

Short-Term Forecasting in the Trading Markets: Suggested Capture Intervals for Bloomberg™ FA-Downloads

General Context: Forecasting in the market trading world has been recognized as the core decision-making imperative. Archimedes remarked: Give Me a Fulcrum, and I Shall Move the World; the simile in the market trading world is; Give Me an Effective Forecasting Model and I can make Bill Gates and Sam Walton look like Paupers. An essential aspect of creating forecasts that can inform decision-makers are confidence intervals on forecasting projections.Research Focus: The Bloomberg™ terminals [BBT] provide a treasure-trove of historical data and also, on a selected basis, single-valued forecasting projections. However, these forecasts are single-value estimates, usually one or two periods-ahead only for the Income Statement variables. The focus of this research report is to produce an interval that can be: (1) simply oriented from reported BBT-information, and (2) used to better develop a decision-context for the information accessible through the FA-link of the BBT.Results: Using forecasting models in the OLS regression-class for: Time-Series, Y = f(X) versions, and a Moving Average Model, seven forecasting capture intervals were developed from randomly selected firms offered by the BBT. These capture intervals were used to produce standardized-capture-scalar-coefficients that were applied to the reported values offered by the BBT to create one-period-ahead capture intervals. This modeling protocol is referred to as the Empirical Interval Capture Model [EICM]. The EICM produces a capture interval with precision of about 30% where the capture rate was on the order of 90%.

Sectoral Energy Demand Forecasting under an Assumption-Free Data-Driven Technique

In order to implement sustainable economic policies, realistic and high accuracy demand projections are key to drawing and implementing realizable environmentally-friendly energy policies. However, some core energy models projections depict considerably high forecast inaccuracies in their previous projections. The inaccuracies are due to the massive assumption-driven variables whose assumptions and scenarios typically deviate from their realized levels. Here, we propose a high-accuracy assumption-free own-data-driven technique that utilizes zero of the traditional determinants as well as assumptions or scenarios for sectorial energy demand forecasting; and implement it in the United States (U.S.). The results show that the forecast accuracy of our gated recurrent network presents an enormous improvement on Annual Energy Outlook 2008 forecast projections. With evidence that our proposed sequential algorithm outperformed Annual Energy Outlook 2008 forecast projections, our proposed algorithm will guide policymakers in making sustainable energy-related policies in the near future. Although future realized consumption levels are unknown, we present our estimated projections along with Annual Energy Outlook 2018 projections to inform policymakers on future energy demands for the commercial sector, industrial sector, residential sector, and transportation.

La planification des médecins en Europe : une revue de la littérature des modèles de projection

BackgroundHealthcare is a labor-intensive sector in which half of the expenses are dedicated to human resources. Therefore, policy makers, at national and internal levels, attend to the number of practicing professionals and the skill mix. This paper aims to analyze the European forecasting model for supply and demand of physicians. MethodsTo describe the forecasting tools used for physician planning in Europe, a grey literature search was done in the OECD, WHO, and European Union libraries. Electronic databases such as Pubmed, Medine, Embase and Econlit were also searched. ResultsQuantitative methods for forecasting medical supply rely mainly on stock-and-flow simulations and less often on systemic dynamics. Parameters included in forecasting models exhibit wide variability for data availability and quality. The forecasting of physician needs is limited to healthcare consumption and rarely considers overall needs and service targets. Besides quantitative methods, horizon scanning enables an evaluation of the changes in supply and demand in an uncertain future based on qualitative techniques such as semi-structured interviews, Delphi Panels, or focus groups. Finally, supply and demand forecasting models should be regularly updated. Moreover, post-hoc analyze is also needed but too rarely implemented. ConclusionMedical human resource planning in Europe is inconsistent. Political implementation of the results of forecasting projections is essential to insure efficient planning. However, crucial elements such as mobility data between Member States are poorly understood, impairing medical supply regulation policies. These policies are commonly limited to training regulations, while horizontal and vertical substitution is less frequently taken into consideration.

Analisis Trend Produksi Padi Organik dan Anorganik pada Lahan Rawa di Kabupaten Hulu Sungai Utara

This study aims to find out (i) fluctuations in production trends, (ii) forecasting projections, and (iii) volumes of organic and inorganic paddy production from 2016-2021 at wetland in Hulu Sungai Utara Regency. The study was conducted from March to July 2016. The research method used was descriptive with secondary data obtained from the Office of Food Crops and Horticulture of Hulu Sungai Utara Regency, data obtained from 2010 to 2015. The analysis used in this study was linear trend with the least squares method which aims to predict or predict the development of organic and inorganic rice production in swamps in North Hulu Sungai Regency, South Kalimantan Province. The results showed fluctuations in production trends, forecasting projections and production of organic and inorganic paddy each year has increased

ASKBILL as a web-based program to enhance sheep well-being and productivity

ASKBILL is a web-based program that uses farm measurements, climate data and information on genetics to predict pasture growth, animal performance and animal health and climate risks. The program uses several biophysical models, which are customised by user inputs, localised daily weather updates and a dynamical probabilistic 90-day climate forecast to enhance sheep well-being and productivity. This approach can minimise the requirement for manual, auto and remote measurements, thus reducing labour requirements and complexity. In this article, the animal growth model provides an example of a biophysical model used to provide predictions. This is an energy-based model and the model parameterisation is designed to be physiologically meaningful and able to be customised for the genetic merit of the animal using a growth coefficient that calibrates growth of body components and energy requirements. A key feature of the animal growth model is its forecast projections, which are based on an ensemble of simulations. The model can estimate supplementary feeding rates required to achieve target liveweights and body condition scores and stocking rates required to achieve target pasture levels. The model can be customised for a farm and its livestock and is updated daily in response to climate data. This dynamic feature enables it to provide early stage alerts to users when animal production targets are unlikely to be met.

Acceleration in U.S. Mean Sea Level? A New Insight using Improved Tools

ABSTRACT Watson, P.J., 2016. Acceleration in U.S. mean sea level? A new insight using improved tools. The detection of acceleration in mean sea level around the data-rich margins of the United States has been a keen endeavour of sea-level researchers following the seminal work of Bruce Douglas in 1992. Over the past decade, such investigations have taken on greater prominence given mean sea level remains a key proxy by which to measure a changing climate system. The physics-based climate projection models are forecasting that the current global average rate of mean sea-level rise (≈3 mm/y) might climb to rates in the range of 10–20 mm/y by 2100. Most research in this area has centred on reconciling current rates of rise with the significant accelerations required to meet the forecast projections of climate models. The analysis in this paper is based on a recently developed analytical package titled “msltrend,” specifically designed to enhance estimates of trend, real-time velocity and acceleration in the ...

Estimating credit risk parameters using ensemble learning methods: an empirical study on loss given default

In credit risk modeling, banks and insurance companies routinely use a single model for estimating key risk parameters. Combining several models to make a final prediction is not often considered. Using an ensemble or a collection of models rather than a single model can improve the accuracy and robustness of prediction results. In this study, we investigate two well-established ensemble learning methods (stochastic gradient boosting and random forest) and propose two new ensembles (ensemble by partial least squares and bag-boosting) in the application of predicting the loss given default. We demonstrate that an ensemble approach significantly increases the discriminatory power of the model compared with a single decision tree. In addition, the ensemble learning methods can be applied directly to predicting the exposure at default and probability of default with some simple modifications. The proposed approaches introduce a novel modeling framework that banks and other financial institutions can use to estimate and validate credit risk parameters based on the internal data of different portfolios. Moreover, the proposed approaches can be readily extended to general portfolio risk modeling in the areas of regulatory capital and economic capital management, loss forecasting, stress testing and pre-provision net revenue projections.

SPACEBORNE QUANTITATIVE ASSESSMENT OF PRIMARY PRODUCTION VARIATIONS IN THE ARCTIC OCEAN OVER THE PREVIOUS DECADE

Abstract. Spaceborne one month averaged data, predominantly from the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) and partly from the Moderate Resolution Imaging Spectroradiometer (MODIS), were used to investigate changes in primary production (PP) by phytoplankton in the Arctic Ocean from 1998 till 2010. Several PP retrieval algorithms were tested against the collected in situ data, and it was shown that the algorithm by Behrenfeld and Falkowski gave the best results (with the coefficient of correlation, r equal to 0.8 and 0.75, respectively, for the pelagic and shelf zones. Based on the performed test, the Behrenfeld and Falkowski algorithm was further applied for determining both the annual PP in the Arctic and the PP trend over the aforementioned time period. The results of our analysis indicate that PP in the Arctic has increased by ~ 15.9% over 13 years. This finding, as well as the absolute annual values of PP remotely quantified in the present study, is at odds with analogous numerical assessments by other workers. These disagreements are thought to be due to differences in the applied methodologies of satellite data processing, such as cloud masking and determination of phytoplankton concentration within (i) overcast areas, and (ii) areas of massive growth of coccolithophore algae, as well as (iii) in the shelf zone prone to a significant influence of land and river runoff. Hindcast (a decadal long) and forecast projections of PP variations are performed.

Technoeconomic Projections with Artificial Neural Networks using an Ensemble of Sparsely-Sampled Bootstrapped Data

Aims: The present study refers to developing an artificial neural network (ANN) that can be designed exclusively for ex ante forecasting in technoeconomic contexts using an ensemble set of sparse and insufficient sampled-data availed ex post. Study Design: In general, the samples in a data set of technoeconomic structures would largely be limited in number due to sparse-sampling; also, availability of number of such sets is mostly inadequate for robust training of an ANN so as to obtain realistic inferences subsequently in the prediction phase. Hence, a sparsity-recovery strategy is advocated via a cardinality enhancement procedure (through Nyquist sampling) performed on the sparse data set in order to augment the number of samples in its sampled-data space. Further, the concept of statistical bootstrapping technique of resampling is invoked and applied on the cardinality-improved subset so as to obtain an enhanced number of data sets. This ensemble of data set is then adopted to facilitate robust training of the test Original Research Article British Journal of Economics, Management & Trade, 4(2): 228-263, 2014 229 ANN. Place and Duration of Study: The studies were conducted (2012-2013) at: Department of Computer and Electrical Engineering and Computer Science, College of Engineering & Computer Science, Florida Atlantic University, Boca Raton, Florida 33431, USA. Methodology: The study governs technoeconomic ex ante projections pertinent to a wind-power generation business complex elucidated via ANN-based forecasting. Relevant test ANN is designed to accommodate training with an ensemble of sampled set available ex post but, in limited numbers. The associated scarcity is recovered by artificially enhancing the data space to an adequate extent via Nyquist sampling and bootstrapping techniques. Further, the test ANN designed corresponds to a multilayer perceptron (MLP) supporting backpropagation of the perceived error at the output with respect to a supervisory value. It accommodates the bootstrapped data space at its input relevant to technoeconomic details on a practical wind-power system performance reported in the literature. The training and prediction exercises on the test ANN corresponds to optimally elucidating output predictions in the context of the technoeconomics framework of the power generation considered. Results: Using the test ANN trained with bootstrap-enhanced, scarcity-recovered sparse data on wind-power generation statistics and associated plant economics, reliable inference (in the prediction phase) is achieved on the system performance. That is, the ANN output obtained depicts forecast projections on the productivity of electric power generation in the ex ante regime. Simulation studies thereof and results obtained demonstrate the efficacy of the method proposed, bootstrapping algorithm developed and the use of MLP in the technoeconomic contexts.

Diagnosing Snowband Predictability Using a Multimodel Ensemble System

Abstract The forecast uncertainty of mesoscale snowband formation and evolution is compared using predictions from a 16-member multimodel ensemble at 12-km grid spacing for the 25 December 2002, 12 February 2006, and 14 February 2007 northeast U.S. snowstorms. Using these predictions, the case-to-case variability in the predictability of band formation and evolution is demonstrated. Feature-based uncertainty information is also presented as an example of what may be operationally feasible from postprocessing information from future short-range ensemble forecast systems. Additionally, the initial condition sensitivity of band location in each case is explored by contrasting the forecast evolutions of initial condition members with large differences in snowband positions. Considerable uncertainty in the occurrence, and especially timing and location, of band formation and subsequent evolution was found, even at forecast projections <24 h. The ensemble provided quantitative mesoscale band uncertainty information, and differentiated between high-predictability (14 February 2007) and low-predictability (12 February 2006) cases. Among the three cases, large (small) initial differences in the upper-level PV distribution and surface mean sea level pressure of the incipient cyclone were associated with large (small) differences in forecast snowband locations, suggesting that case-to-case differences in predictability may be related to the quality of the initial conditions. The complexity of the initial flow may also be a discriminator. Error growth was evident in each case, consistent with previous mesoscale predictability research, but predictability differences were not correlated to the degree of convection. Discussion of these results and future extensions of the work are presented.

What to Do with Less

In his column, AWWA Executive Director David LaFrance discusses the impacts of declining water sales on utilities. LaFrance presents an alternative to raising fixed fees in order to manage revenue risks. The alternative includes the following: dividing reduced water sales into two categories ‐ planned and unplanned; making sure revenues are meeting costs at today's water sales levels; develop revenue forecast projections covering at least three to five years; and, get a solid handle on the relationship of fixed and variable costs to trued‐up fixed and variable revenue, making sure there are reserves to absorb any reasonable demand risks.

High-Resolution GFS-Based MOS Quantitative Precipitation Forecasts on a 4-km Grid

Abstract The Meteorological Development Laboratory (MDL) of the National Weather Service (NWS) has developed high-resolution Global Forecast System (GFS)-based model output statistics (MOS) 6- and 12-h quantitative precipitation forecast (QPF) guidance on a 4-km grid for the contiguous United States. Geographically regionalized multiple linear regression equations are used to produce probabilistic QPFs (PQPFs) for multiple precipitation exceedance thresholds. Also, several supplementary QPF elements are derived from the PQPFs. The QPF elements are produced (presently experimentally) twice per day for forecast projections up to 156 h (6.5 days); probability of (measurable) precipitation (POP) forecasts extend to 192 h (8 days). Because the spatial and intensity resolutions of the QPF elements are higher than that for the currently operational gridded MOS QPF elements, this new application is referred to as high-resolution MOS (HRMOS) QPF. High spatial resolution and enhanced skill are built into the HRMOS PQPFs by incorporating finescale topography and climatology into the predictor database. This is accomplished through the use of specially formulated “topoclimatic” interactive predictors, which are formed as a simple product of a climatology- or terrain-related quantity and a GFS forecast variable. Such a predictor contains interactive effects, whereby finescale detail in the topographic or climatic variable is built into the GFS forecast variable, and dynamics in the large-scale GFS forecast variable are incorporated into the static topoclimatic variable. In essence, such interactive predictors account for the finescale bias error in the GFS forecasts, and thus they enhance the skill of the PQPFs. Underlying the enhanced performance of the HRMOS QPF elements is extensive use of archived fine-grid radar-based quantitative precipitation estimates (QPEs). The fine spatial scale of the QPE data supported development of a detailed precipitation climatology, which is used as a climatic predictive input. Also, the very large number of QPE sample points supported specification of rare-event (i.e., ≥1.50 and ≥2.00 in.) 6-h precipitation exceedance thresholds as predictands. Geographical regionalization of the PQPF regression equations and the derived QPF elements also contributes to enhanced forecast performance. Limited comparative verification of several 6-h model QPFs in categorical form showed the HRMOS QPF with significantly better threat scores and biases than corresponding GFS and operational gridded MOS QPFs. Limited testing of logistic regression versus linear regression to produce the 6-h PQPFs showed the feasibility of applying the logistic method with the very large HRMOS samples. However, objective screening of many candidate predictors with linear regression resulted in slightly better PQPF skill.

A Comparison between Raw Ensemble Output, (Modified) Bayesian Model Averaging, and Extended Logistic Regression Using ECMWF Ensemble Precipitation Reforecasts

Abstract Using a 20-yr ECMWF ensemble reforecast dataset of total precipitation and a 20-yr dataset of a dense precipitation observation network in the Netherlands, a comparison is made between the raw ensemble output, Bayesian model averaging (BMA), and extended logistic regression (LR). A previous study indicated that BMA and conventional LR are successful in calibrating multimodel ensemble forecasts of precipitation for a single forecast projection. However, a more elaborate comparison between these methods has not yet been made. This study compares the raw ensemble output, BMA, and extended LR for single-model ensemble reforecasts of precipitation; namely, from the ECMWF ensemble prediction system (EPS). The raw EPS output turns out to be generally well calibrated up to 6 forecast days, if compared to the area-mean 24-h precipitation sum. Surprisingly, BMA is less skillful than the raw EPS output from forecast day 3 onward. This is due to the bias correction in BMA, which applies model output statistics to individual ensemble members. As a result, the spread of the bias-corrected ensemble members is decreased, especially for the longer forecast projections. Here, an additive bias correction is applied instead and the equation for the probability of precipitation in BMA is also changed. These modifications to BMA are referred to as “modified BMA” and lead to a significant improvement in the skill of BMA for the longer projections. If the area-maximum 24-h precipitation sum is used as a predictand, both modified BMA and extended LR improve the raw EPS output significantly for the first 5 forecast days. However, the difference in skill between modified BMA and extended LR does not seem to be statistically significant. Yet, extended LR might be preferred, because incorporating predictors that are different from the predictand is straightforward, in contrast to BMA.