Steven L Mullen

Interests

Research

Construction of ensemble forecasting systems. Improving precipitation forecasts.

Teaching

Weather analysis and forecasting. General Education. Online courses.

Courses

Scholarly Contributions

Journals/Publications

• Wandishin, M. S., Stensrud, D. J., Mullen, S. L., & Wicker, L. J. (2010). On the predictability of mesoscale convective systems: Three-dimensional simulations. Monthly Weather Review, 138(3), 863-885.
  More info

  Abstract: Mesoscale convective systems (MCSs) are a dominant climatological feature of the central United States and are responsible for a substantial fraction of warm-season rainfall. Yet very little is known about the predictability of MCSs. To help address this situation, a previous paper by the authors examined a series of ensemble MCS simulations using a two-dimensional version of a storm-scale (Δx 5 1 km) model. Ensemble member perturbations in the preconvective environment, namely, wind speed, relative humidity, and convective instability, are based on current 24-h forecast errors from the North American Model (NAM). That work is now extended using a full three-dimensional model. Results from the three-dimensional simulations of the present study resemble those found in two dimensions. The model successfully produces anMCSwithin 100 km of the location of the control run in around 70% of the ensemble runs using perturbations to the preconvective environment consistent with 24-h forecast errors, while reducing the preconvective environment uncertainty to the level of current analysis errors improves the success rate to nearly 85%. This magnitude of improvement in forecasts of environmental conditions would represent a radical advance in numerical weather prediction. The maximum updraft and surface wind forecast uncertainties are of similar magnitude to their two-dimensional counterparts. However, unlike the two-dimensional simulations, in three dimensions, the improvement in the forecast uncertainty of storm features requires the reduction of preconvective environmental uncertainty for all perturbed variables. The MCSs in many of the runs resemble bow echoes, but surface winds associated with these solutions, and the perturbation profiles that produce them, are nearly indistinguishable from the nonbowing solutions, making any conclusions about the bowlike systems difficult. © 2010 American Meteorological Society.
• W., S., Castro, C. L., Mullen, S. L., Comrie, A. C., & Pytlak, E. (2009). The relationship of transient upper-level troughs to variability of the North American monsoon system. Journal of Climate, 22(15), 4213-4227.
  More info

  Abstract: Relationships between transient upper-tropospheric troughs and warm season convective activity over the southwest United States and northern Mexico are explored. Analysis of geopotential height and vorticity fields from the North American Regional Reanalysis and cloud-to-ground lightning data indicates that the passage of mobile inverted troughs (IVs) significantly enhances convection when it coincides with the peak diurnal cycle (1800-0900 UTC) over the North American monsoon (NAM) region. The preferred tracks of IVs during early summer are related to the dominant modes of Pacific sea surface temperature (SST) variability. When La Niña-like (El Niño-like) conditions prevail in the tropical Pacific and the eastern North Pacific has a horseshoe-shaped negative (positive) SST anomaly, IVs preferentially track farther north (south) and are slightly (typically one IV) more (less) numerous. These results point to the important role that synoptic-scale disturbances play in modulating the diurnal cycle of precipitation over the NAM region and the significant impact that the statistically supported low-frequency Pacific SST anomalies exert on the occurrence and track of these synoptic transients. © 2009 American Meteorological Society.
• Wandishin, M. S., & Mullen, S. J. (2009). Multiclass ROC analysis. Weather and Forecasting, 24(2), 530-547.
  More info

  Abstract: Receiver operating characteristic (ROC) curves have become a common analysis tool for evaluating forecast discrimination: the ability of a forecast system to distinguish between events and nonevents. As is implicit in that statement, application of the ROC curve is limited to forecasts involving only two possible outcomes, such as rain and no rain. However, many forecast scenarios exist for which there are multiple possible outcomes, such as rain, snow, and freezing rain. An extension of the ROC curve to multiclass forecast problems is explored. The full extension involves high-dimensional hypersurfaces that cannot be visualized and that present other problems. Therefore, several different approximations to the full extension are introduced using both artificial and actual forecast datasets. These approximations range from sets of simple two-class ROC curves to sets of three-dimensional ROC surfaces. No single approximation is superior for all forecast problems; thus, the specific aims in evaluating the forecast must be considered. © 2009 American Meteorological Society.
• Hamill, T. M., Whitaker, J. S., & Mullen, S. L. (2008). Reply. Bulletin of the American Meteorological Society, 89(9), 1376-1378.
• Kursinski, A. L., & Mullen, S. L. (2008). Spatiotemporal variability of hourly precipitation over the Eastern contiguous United States from stage IV multisensor analyses. Journal of Hydrometeorology, 9(1), 3-21.
  More info

  Abstract: The statistical character of precipitation events from hourly stage IV analyses is documented for the eastern United States during the cool [December February (DJF)] and the warm [June-August (JJA)] seasons for the four years of 2002-05. Isotropic e-folding distances and in situ e-folding times are computed for mesh sizes that vary from 4 km (the minimal stage IV pixel size) to 32 km for two thresholds: light (1 mm h-1) and heavy (5 mm h-1) precipitation rates. Marked seasonal variability characterizes the e-folding times. They typically run between 2 and 3 h during winter and 1 and 2 h during summer for light events, and they run an hour shorter for heavy rainfall during both seasons. Spatial decorrelation estimates also reveal considerable seasonal and geographical variability; e-folding distances typically lie between 60 and 180 km during the winter and between 30 and 60 km during the summer for light episodes, and they are approximately a factor of 2 to 3 shorter for heavy events. Anisotropic statistics are estimated by a simple geometric model. Hourly precipitation patterns show a preference for a southwest-northeast orientation during both seasons with greater elongation during the winter. Mean propagation velocities of precipitating systems are faster and are more closely aligned with the dilatation axis during the winter. These statistics should provide useful guidance for diagnosing and improving the spatiotemporal variance characteristics of precipitation for downscaling algorithms and numerical models of hydrometeorological prediction systems. © 2008 American Meteorological Society.
• Wandishin, M. S., Stensrud, D. J., Mullen, S. L., & Wicker, L. J. (2008). On the predictability of mesoscale convective systems: Two-dimensional simulations. Weather and Forecasting, 23(5), 773-785.
  More info

  Abstract: Mesoscale convective systems (MCSs) are a dominant climatological feature of the central United States and are responsible for a substantial fraction of warm season rainfall. Yet very little is known about the predictability of MCSs. To help alleviate this situation, a series of ensemble simulations of an MCS are performed on a two-dimensional, storm-scale (Δx = 1 km) model. Ensemble member perturbations in wind speed, relative humidity, and instability are based on current 24-h forecast errors from the North American Model (NAM). The ensemble results thus provide an upper bound on the predictability of mesoscale convective systems within realistic estimates of environmental uncertainty, assuming successful convective initiation. The simulations are assessed by considering an ensemble member a success when it reproduces a convective system of at least 20 km in length (roughly the size of two convective cells) within 100 km on either side of the location of the MCS in the control run. By that standard, MCSs occur roughly 70% of the time for perturbation magnitudes consistent with 24-h forecast errors. Reducing the perturbations for all fields to one-half the 24-h error values increases the MCS success rate to over 90%. The same improvement in forecast accuracy would lead to a 30%-40% reduction in maximum surface wind speed uncertainty and a roughly 20% reduction in the uncertainty in maximum updraft strength, and initially slower growth in the uncertainty in the size of the MCS. However, the occurrence of MCSs drops below 50% as the midlayer mean relative humidity falls below 65%. The response of the model to reductions in forecast errors for instability, moisture, and wind speed is not consistent and cannot be easily generalized, but each can have a substantial impact on forecast uncertainty. © 2008 American Meteorological Society.
• Yuan, H., Gao, X., Mullen, S. L., Sorooshian, S., Jun, D. u., & Juang, H. H. (2007). Calibration of probabilistic quantitative precipitation forecasts with an artificial neural network. Weather and Forecasting, 22(6), 1287-1303.
  More info

  Abstract: A feed-forward neural network is configured to calibrate the bias of a high-resolution probabilistic quantitative precipitation forecast (PQPF) produced by a 12-km version of the NCEP Regional Spectral Model (RSM) ensemble forecast system. Twice-daily forecasts during the 2002-2003 cool season (1 November-31 March, inclusive) are run over four U.S. Geological Survey (USGS) hydrologic unit regions of the southwest United States. Calibration is performed via a cross-validation procedure, where four months are used for training and the excluded month is used for testing. The PQPFs before and after the calibration over a hydrological unit region are evaluated by comparing the joint probability distribution of forecasts and observations. Verification is performed on the 4-km stage IV grid, which is used as "truth." The calibration procedure improves the Brier score (BrS), conditional bias (reliability) and forecast skill, such as the Brier skill score (BrSS) and the ranked probability skill score (RPSS), relative to the sample frequency for all geographic regions and most precipitation thresholds. However, the procedure degrades the resolution of the PQPFs by systematically producing more forecasts with low nonzero forecast probabilities that drive the forecast distribution closer to the climatology of the training sample. The problem of degrading the resolution is most severe over the Colorado River basin and the Great Basin for relatively high precipitation thresholds where the sample of observed events is relatively small. © 2007 American Meteorological Society.
• Yuan, H., Mullen, S. L., Gao, X., Sorooshian, S., Jun, D. u., & Juang, H. H. (2007). Short-range probabilistic quantitative precipitation forecasts over the southwest United States by the RSM ensemble system. Monthly Weather Review, 135(5), 1685-1698.
  More info

  Abstract: The National Centers for Environmental Prediction (NCEP) Regional Spectral Model (RSM) is used to produce twice-daily (0000 and 1200 UTC), high-resolution ensemble forecasts to 24 h. The forecasts are performed at an equivalent horizontal grid spacing of 12 km for the period 1 November 2002 to 31 March 2003 over the southwest United States. The performance of 6-h accumulated precipitation is assessed for 32 U.S. Geological Survey hydrologic catchments. Multiple accuracy and skill measures are used to evaluate probabilistic quantitative precipitation forecasts. NCEP stage-IV precipitation analyses are used as "truth," with verification performed on the stage-IV 4-km grid. The RSM ensemble exhibits a ubiquitous wet bias. The bias manifests itself in areal coverage, frequency of occurrence, and total accumulated precipitation over every region and during every 6-h period. The biases become particularly acute starting with the 1800-0000 UTC interval, which leads to a spurious diurnal cycle and the 1200 UTC cycle being more adversely affected than the 0000 UTC cycle. Forecast quality and value exhibit marked variability over different hydrologic regions. The forecasts are highly skillful along coastal California and the windward slopes of the Sierra Nevada Mountains, but they generally lack skill over the Great Basin and the Colorado basin except over mountain peaks. The RSM ensemble is able to discriminate precipitation events and provide useful guidance to a wide range of users over most regions of California, which suggests that mitigation of the conditional biases through statistical postprocessing would produce major improvements in skill. © 2007 American Meteorological Society.
• Hamill, T. M., Whitaker, J. S., & Mullen, S. L. (2006). Reforecasts: An important dataset for improving weather predictions. Bulletin of the American Meteorological Society, 87(1), 33-46.
  More info

  Abstract: A "reforecast" (retrospective forecast) dataset has been developed. This dataset is comprised of a 15-member ensemble run out to a 2-week lead. Forecasts have been run every day from 0000 UTC initial conditions from 1979 to the present. The model is a 1998 version of the National Centers for Environmental Prediction's (NCEPs) Global Forecast System (GFS) at T62 resolution. The 15 initial conditions consist of a reanalysis and seven pairs of bred modes. This dataset facilitates a number of applications that were heretofore impossible. Model errors can be diagnosed from the past forecasts and corrected, thereby dramatically increasing the forecast skill. For example, calibrated precipitation forecasts over the United States based on the 1998 reforecast model are more skillful than precipitation forecasts from the 2002 higher-resolution version of the NCEP GFS. Other applications are also demonstrated, such as the diagnosis of the bias for model development and an identification of the most predictable patterns of week-2 forecasts. It is argued that the benefits of reforecasts are so large that they should become an integral part of the numerical weather prediction process. Methods for integrating reforecast approaches without seriously compromising the pace of model development are discussed. Users wishing to explore their own applications of reforecasts can download them through a Web interface. ©2005 American Meteorological Society.
• Wandishin, M. S., Baldwin, M. E., Mullen, S. L., & Cortinas Jr., J. V. (2005). Short-range ensemble forecasts of precipitation type. Weather and Forecasting, 20(4), 609-626.
  More info

  Abstract: Short-range ensemble forecasting is extended to a critical winter weather problem: forecasting precipitation type. Forecast soundings from the operational NCEP Short-Range Ensemble Forecast system are combined with five precipitation-type algorithms to produce probabilistic forecasts from January through March 2002. Thus the ensemble combines model diversity, initial condition diversity, and postprocessing algorithm diversity. All verification numbers are conditioned on both the ensemble and observations recording some form of precipitation. This separates the forecast of type from the yes-no precipitation forecast. The ensemble is very skillful in forecasting rain and snow but it is only moderately skillful for freezing rain and unskillful for ice pellets. However, even for the unskillful forecasts the ensemble shows some ability to discriminate between the different precipitation types and thus provides some positive value to forecast users. Algorithm diversity is shown to be as important as initial condition diversity in terms of forecast quality, although neither has as big an impact as model diversity. The algorithms have their individual strengths and weaknesses, but no algorithm is clearly better or worse than the others overall. © 2005 American Meteorological Society.
• Yuan, H., Mullen, S. L., Gao, X., & Sorooshian, S. (2005). Calibration of probabilistic quantitative precipitation forecasts from the RSM ensemble forecasts over hydrologic regions. 85th AMS Annual Meeting, American Meteorological Society - Combined Preprints, 1879-1883.
• Yuan, H., Mullen, S. L., Gao, X., & Sorooshian, S. (2005). Calibration of probabilistic quantitative precipitation forecasts from the RSM ensemble forecasts over hydrologic regions. 85th AMS Annual Meeting, American Meteorological Society - Combined Preprints, 199-203.
• Yuan, H., Mullen, S. L., Gao, X., Sorooshian, S., Jun, D. u., & Juang, H. H. (2005). Verification of probabilistic quantitative precipitation forecasts over the southwest United States during winter 2002/03 by the RSM ensemble system. Monthly Weather Review, 133(1), 279-294.
  More info

  Abstract: The National Centers for Environmental Prediction (NCEP) Regional Spectral Model (RSM) is used to generate ensemble forecasts over the southwest United States during the 151 days of 1 November 2002 to 31 March 2003. RSM forecasts to 24 h on a 12-km grid are produced from 0000 and 1200 UTC initial conditions. Eleven ensemble members are run each forecast cycle from the NCEP Global Forecast System (GFS) ensemble analyses (one control and five pairs of bred modes) and forecast lateral boundary conditions. The model domain covers two NOAA River Forecast Centers: the California Nevada River Forecast Center (CNRFC) and the Colorado Basin River Forecast Center (CBRFC). Ensemble performance is evaluated for probabilistic forecasts of 24-h accumulated precipitation in terms of several accuracy and skill measures. Differences among several NCEP precipitation analyses are assessed along with their impact on model verification, with NCEP stage IV blended analyses selected to represent "truth." Forecast quality and potential value are found to depend strongly on the verification dataset, geographic region, and precipitation threshold. In general, the RSM forecasts are skillful over the CNRFC region for thresholds between 1 and 50 mm but are unskillful over the CBRFC region. The model exhibits a wet bias for all thresholds that is larger over Nevada and the CBRFC region than over California. Mitigation of such biases over the Southwest will pose serious challenges to the modeling community in view of the uncertainties inherent in verifying analyses. © 2005 American Meteorological Society.
• Bright, D. R., & Mullen, S. L. (2002). Short-range ensemble forecasts of precipitation during the Southwest monsoon. Weather and Forecasting, 17(5), 1080-1100.
  More info

  Abstract: The skill and potential value of fifth-generation Pennsylvania State University-National Center for Atmospheric Research Mesoscale Model (MM5) ensembles are evaluated for short-range (24 h) probabilistic quantitative precipitation forecasts over Arizona during the Southwest monsoon. The sensitivity of different ensemble constructs is examined with respect to analysis uncertainty, model parameterization uncertainty, and a combination of both. Model uncertainty is addressed through different cumulus and planetary boundary layer parameterizations and through stochastic forcing representative of a component of subgrid-scale uncertainty, in which a first-order autoregression model adds a stochastic perturbation to the Kain-Fritsch cumulus scheme and Medium-Range Forecast Model PBL scheme. The results indicate that the precipitation forecasts are skillful and may assist operational weather forecasters during the monsoon; however, the forecasts are highly dependent on the cumulus parameterization. The addition of a stochastic element in the Kain-Fritsch cumulus scheme produces a small increase in skill and dispersion. Ensembles that incorporate mixed physics and perturbed analyses are the most skillful. A simple cost-loss model reveals that the monsoon ensembles can aid decision makers. Operational application is demonstrated for a heavy rain event over southern Arizona.
• Bright, D. R., & Mullen, S. L. (2002). The sensitivity of the numerical simulation of the southwest monsoon boundary layer to the choice of PBL turbulence parameterization in MM5. Weather and Forecasting, 17(1), 99-114.
  More info

  Abstract: Summertime convection over Arizona typically begins in the early afternoon and continues into the night. This suggests that the evolution of the daytime planetary boundary layer is important to the development of Arizona convection. If numerical models are to provide useful guidance for forecasting convection during the monsoon, then the planetary boundary layer must be simulated as accurately as possible through utilization of the appropriate physical parameterizations. This study examines the most appropriate Pennsylvania State University-National Center for Atmospheric Research fifth-generation Mesoscale Model (MM5) planetary boundary layer parameterization(s) for deterministic and ensemble modeling of the monsoon. The four MM5 planetary boundary layer parameterizations tested are the Blackadar, Burk-Thompson, Eta, and medium-range forecast (MRF) schemes. The Blackadar and MRF planetary boundary layer schemes correctly predict the development of the deep, monsoon planetary boundary layer, and consequently do a better job of predicting the convective available potential energy and downdraft convective available potential energy, but not the convective inhibition. Because the convective inhibition is not accurately predicted, it is possible that the MM5's ability to initiate or "trigger" convection might be a limiting factor in the model's ability to produce accurate quantitative precipitation forecasts during the monsoon. Since the MM5 planetary boundary layer predicted by the Burk-Thompson and Eta schemes does not accurately reproduce the basic structure of the monsoon planetary boundary layer, their inclusion in a mixed physics ensemble is discussed.
• Mullen, S. L., & Buizza, R. (2002). The impact of horizontal resolution and ensemble size on probabilistic forecasts of precipitation by the ECMWF ensemble prediction system. Weather and Forecasting, 17(2), 173-191.
  More info

  Abstract: The effect of horizontal resolution and ensemble size on the ECMWF Ensemble Prediction System (EPS) is assessed for probabilistic forecasts of 24-h accumulated precipitation. Two sets of experiments are analyzed. The primary experiment compares two spectral truncations (total wavenumbers 159 and 255) for 30 summer and 57 winter dates. An auxiliary experiment compares three truncations (total wavenumbers 159, 255, and 319) for 16 initial dates (8 cool- and 8 warm-season events) during which heavy precipitation (>50 mm) occurred over the eastern United States at day 5 of the forecast. Rain gauge data from the River Forecast Centers of NOAA are used for verification. Skill is measured relative to long-term climatic frequencies, and the statistical significance of differences in the accuracy among the forecasts is estimated. Finer model resolution produces statistically significant improvements in EPS performance for ensemble configurations with the same number of members, especially for lighter thresholds (1 and 10 mm day-1). Performance changes somewhat when ensemble configurations with different resolutions and ensemble sizes, but equivalent computational costs, are compared for the heavier amounts (20 and 50 mm day-1). Coarser-resolution, larger-member ensembles can outperform higher-resolution, smaller-member ensembles in terms of ability to predict rare (in terms of climatic frequency of occurrence) precipitation events. The overall conclusion is that probabilistic forecasts of precipitation from large ensemble sizes at lower resolution can be more valuable to users and decision makers than probabilistic forecasts from smaller ensemble sizes at higher resolution, particularly when heavy precipitation occurs.
• Mullen, S. L., & Buizza, R. (2001). Quantitative precipitation forecasts over the United States by the ECMWF ensemble prediction system. Monthly Weather Review, 129(4), 638-663.
  More info

  Abstract: The performance of the ECMWF Ensemble Prediction System (EPS) is assessed for probabilistic forecasts of 24-h accumulated precipitation over the eastern United States. Daily forecasts for the priod 1 January 1997 to 31 January 1999 are verified for projections of 1-10 days. Verification is performed separately for the cool and warm seasons, and the impact of changes to the EPS that occurred during the study period is assessed. Analyses of rain gauge data from the River Forecast Centers of NOAA are used for verification. Skill is measured relative to long-term climatic frequencies, and the statistical significance of differences in the accuracy and skill among forecasts is estimated. Overall, EPS forecasts are more skillful during the winter than the summer. The EPS produces significantly skillful forecasts to past 1 week for a threshold of 1 mm in both seasons. Accuracy decreases as the threshold increases, until forecasts of 50 mm are not significantly skillful at 1 day. The implementation of evolved singular vectors in the EPS appears to have minimal impact on skill during the summer. The addition of both evolved singular vectors and stochastic processes in the EPS appears to improve short-range performance for thresholds between 1 and 20 mm during the winter, but results for higher thresholds (50 mm) are equivocal.
• Wandishin, M. S., Mullen, S. L., Stensrud, D. J., & Brooks, H. E. (2001). Evaluation of a short-range multimodel ensemble system. Monthly Weather Review, 129(4), 729-747.
  More info

  Abstract: Forecasts from the National Centers for Environmental Prediction's experimental short-range ensemble system are examined and compared with a single run from a higher-resolution model using similar computational resources. The ensemble consists of five members from the Regional Spectral Model and 10 members from the 80-km Eta Model, with both in-house analyses and bred perturbations used as initial conditions. This configuration allows for a comparison of the two models and the two perturbation strategies, as well as a preliminary investigation of the relative merits of mixed-model, mixed-perturbation ensemble systems. The ensemble is also used to estimate the short-range predictability limits of forecasts of precipitation and fields relevant to the forecast of precipitation. Whereas error growth curves for the ensemble and its subgroups are in relative agreement with previous work for large-scale fields such as 500-mb heights, little or no error growth is found for fields of mesoscale interest, such as convective indices and precipitation. The difference in growth rates among the ensemble subgroups illustrates the role of both initial perturbation strategy and model formulation in creating ensemble dispersion. However, increase spread per se is not necessarily beneficial, as is indicated by the fact that the ensemble subgroup with the greatest spread is less skillful than the subgroup with the least spread. Further examination into the skill of the ensemble system for forecasts of precipitation shows the advantage gained from a mixed-model strategy, such that even the inclusion of the less skillful Regional Spectral Model members improves ensemble performance. For some aspects of forecast performance, even ensemble configurations with as few as five members are shown to significantly outperform the 29-km Meso-Eta Model.
• Burke, E. J., Shuttleworth, W. J., Yang, Z. -., Mullen, S. L., & Arain, M. A. (2000). The impact of the parameterization of heterogeneous vegetation on the modeled large-scale circulation in CCM3-BATS. Geophysical Research Letters, 27(3), 397-400.
  More info

  Abstract: This letter reports evidence of an unexpected large change in the planetary-scale circulation as a consequence of modifying the land cover within the Community Climate Model (CCM3). Three 10-year simulations were analyzed using (i) default land cover, i.e., a single dominant vegetation in each land grid cell; (ii) aggregate land cover that includes a representation of subpixel heterogeneity; and (iii) aggregate land cover for aerodynamic properties, but default land cover otherwise. Simulations (ii) and (iii) were similar, indicating that aerodynamic properties are influential at the planetary scale. Comparing these two runs with (i), there are significant differences in the boreal summer. These include a northward shift of the Northern Hemisphere jet that relates to a decrease in the zonally averaged aerodynamic roughness around 60°N, and a perturbation of the Southern Hemisphere jet. Field significance tests suggest these changes are likely a remote influence of the Northern Hemisphere perturbation, not a sampling fluctuation.
• Du, J., Mullen, S. L., & Sanders, F. (2000). Removal of distortion error from an ensemble forecast. Monthly Weather Review, 128(9), 3347-3351.
  More info

  Abstract: Large errors developed by 24 h during a 25-member ensemble forecast of quantity of precipitation. The errors could be attributed to an insufficient northeastward motion of the area of precipitation and excessive amounts. This was determined by partitioning of the root-mean-square error into a distortion error, the sum of contributions from incorrect position and magnitude, and a residual error. The distortion error accounted for more than half of the total error. The distortion error occurs on the synoptic scale and can likely be somewhat ameliorated by future improvements in analysis-forecast systems. The residual error occurs at smaller, less predictable scales, and prospects for its deterministic improvement are not so sanguine.
• Hamill, T. M., Mullen, S. L., Snyder, C., Toth, Z., & Baumhefner, D. P. (2000). Ensemble Forecasting in the Short to Medium Range: Report from a Workshop. Bulletin of the American Meteorological Society, 81(11), 2653-2664.
• Sanders, F., Mullen, S. L., & Baumhefner, D. P. (2000). Ensemble simulations of explosive cyclogenesis at ranges of 2-5 days. Monthly Weather Review, 128(8 II), 2920-2934.
  More info

  Abstract: Ensemble simulations of explosive cyclogenesis are examined in a lengthy run of a global general circulation model with the perfect ensemble context. Attention is focused on the day when the deepest low appeared. An ensemble of 31 members is obtained by integrating 30 additional runs strating from slightly perturbed initial conditions. The perturbations are randomly selected to represent equal approximations to the truth, given typical analysis differences between major centers. Ensembles are generated starting two, three, four, and five days prior to maximum depth. Two lows are contrasted, the deepest low near Kamchatka and a marginally explosive low over the central Pacific. The early development of both systems was suppressed by their presence in the confluent entrance region of the Pacific winter jet. An intense low near Kamchatka eventually developed in each member of the ensemble at all projections, but the details of development varied from member to member and were related to the involvement of a surface perturbation coming up into the system from low latitudes. In contrast, cyclogenesis over the central Pacific occurred in some members of the ensemble but not at all in others. The difference in behavior of the two systems is reflected in a localized enhancement of the error growth of the planetary and s synoptic scales for the central Pacific low and is related to the smaller horizontal scale of the central Pacific low. Probabilistic estimates of precipitation quantity and surface wind speeds produced by the ensemble showed moderate skill at day 5 with respect to climatology, mainly away from the reginos of most vigorous synoptic activity, when verified against individual ensemble members. Skill would be reduced if the ensemble mean proved to be more seriously in error as is the case for a forecast verified against observations.
• Mullen, S. L., Jun, D. u., & Sanders, F. (1999). The dependence of ensemble dispersion on analysis-forecast systems: Implications to short-range ensemble forecasting of precipitation. Monthly Weather Review, 127(7), 1674-1686.
  More info

  Abstract: The impact of differences in analysis-forecast systems on dispersion of an ensemble forecast is examined for a case of cyclogenesis. Changes in the dispersion properties between two 25-member ensemble forecasts with different cumulus parameterization schemes and different initial analyses are compared. The statistical significance of the changes is assessed. Error growth due to initial condition uncertainty depends significantly on the analysis-forecast system. Quantitative precipitation forecasts and probabilistic quantitative precipitation forecasts are extremely sensitive to the specification of physical parameterizations in the model. Regions of large variability tend to coincide with a high likelihood of parameterized convection. Analysis of other model fields suggests that those with relatively large energy in the mesoscale also exhibit highly significant differences in dispersion. The results presented here provide evidence that the combined effect of uncertainties in model physics and the initial state provides a means to increase the dispersion of ensemble prediction systems, but care must be taken in the construction of mixed ensemble systems to ensure that other properties of the ensemble distribution are not overly degraded.
• Mullen, S. L., Schmitz, J. T., & Rennó, N. O. (1998). Intraseasonal variability of the summer monsoon over southeast Arizona. Monthly Weather Review, 126(11), 3016-3035.
  More info

  Abstract: Intraseasonal fluctuations associated with the Mexican monsoon system are examined for the semiarid Sonoran Desert region. Daily rain gauge accumulations, radiosonde reports, satellite imagery, and global analyses are all analyzed. Composite wet and dry periods during July and August of 1985-92 are compared, and the statistical significance of differences between the composite fields are assessed. Significant differences exist between many of the wet and dry fields over the Sonoran Desert. As the monsoon shifts from dry to wet conditions, the subtropical ridge moves ∼5° lat northward, and the middle- and upper-tropospheric (700-300 mb) winds back from southwesterly to southeasterly. The midtropospheric transport of water vapor from the southeast strengthens, and the precipitable water values increase by as much as ∼1.2 cm (∼0.5 in.). Middle-tropospheric air parcels ascend into the region from the southeast, while low-level air parcels continue to stream inland from the Gulf of California and up the slopes of the Sierra Madre Occidental. The surface and midtropospheric air parcels rise at an average rate of ∼50-75 mb per day and would saturate within 2 days if undiluted. This combination of conditions leads to a more unstable atmosphere and an increase in convective activity. Spectral analysis of precipitation data for southeast Arizona indicates that 75% of the temporal variance is contained in fluctuations longer than 7 days and that a statistically significant peak exists in the 12-18-day band. These results suggest that large-scale, low-frequency dynamics might play an important role in modulating the variability of convective activity over the Sonoran Desert.
• Nutter, P. A., Mullen, S. L., & Baumhefner, D. P. (1998). The impact of initial condition uncertainty on numerical simulations of blocking. Monthly Weather Review, 126(9), 2482-2502.
  More info

  Abstract: The impact of initial condition uncertainty (ICU) on the onset and maintenance of eastern North Pacific blocking is examined within the framework of a general circulation model (GCM) and the perfect model assumption. Comparisons are made with the contrasting zonal flow regime. Twenty-member ensembles of perturbed simulations are run out to 15 days for the zonal flow, and for blocking at lead times of 8, 4, 2, and 0 days. Blocking occurs in 95% of the 0-day lead simulations and declines monotonically to 65% for the 8-day lead simulations. The uncertainty in the exact time of onset among those simulations that form blocks also increases with lead time. The synoptic-scale features in both the blocking and zonal ensembles saturate, relative to climatological variance, and decorrelate (anomaly correlation coefficient < 0.5) by 6 days. The planetary-scale features, however, maintain skill relative to climatology beyond 10 days. The zonal simulations are generally the first to saturate and decorrelate, followed by simulations of blocking maintenance (0-day lead) and onset (2-, 4-, and 8-day lead), respectively. Thus, initial flows that project negatively (zonal flows) on the GCM's Pacific-North American teleconnectivity pattern are more sensitive to ICU, and thus are less predictable than positive (blocking flows) projections. While the results for this study demonstrate that error growth due to ICU ultimately imposes limits on the predictability of blocking, they also suggest that skillful ensemble predictions of transitions to a blocked state are possible at long lead times if the model error is held to a minimum. The majority of the perturbed simulations make the transition into a blocked state with an associated sustenance of skill even after the loss of skill in the synoptic-scale waves. The results are consistent with the hypothesis that the planetary-scale waves may need to be preconditioned toward the formation of blocking events. They also may, in part, help explain the poor performance of operational models in forecasts of blocking onset.
• Jun, D. u., Mullen, S. L., & Sanders, F. (1997). Short-range ensemble forecasting of quantitative precipitation. Monthly Weather Review, 125(10), 2427-2459.
  More info

  Abstract: The impact of initial condition uncertainty (ICU) on quantitative precipitation forecasts (QPFs) is examined for a case of explosive cyclogenesis that occurred over the contiguous United States and produced widespread, substantial rainfall. The Pennsylvania State University-National Center for Atmospheric Research (NCAR) Mesoscale Model Version 4 (MM4), a limited-area model, is run at 80-km horizontal resolution and 15 layers to produce a 25-member, 36-h forecast ensemble. Lateral boundary conditions for MM4 are provided by ensemble forecasts from a global spectral model, the NCAR Community Climate Model Version I (CCMI). The initial perturbations of the ensemble members possess a magnitude and spatial decomposition that closely match estimates of global analysis error, but they are not dynamically conditioned. Results for the 80-km ensemble forecast are compared to forecasts from the then operational Nested Grid Model (NGM), a single 40-km/15-layer MM4 forecast, a single 80-km/29-layer MM4 forecast, and a second 25-member MM4 ensemble based on a different cumulus parameterization and slightly different unperturbed initial conditions. Large sensitivity to ICU marks ensemble QPF. Extrema in 6-h accumulations at individual grid points vary by as much as 3.00′. Ensemble averaging reduces the root-mean-square error (rmse) for QPF. Nearly 90% of the improvement is obtainable using ensemble sizes as small as 8-10. Ensemble averaging can adversely affect the bias and equitable threat scores, however, because of its smoothing nature. Probabilistic forecasts for five mutually exclusive, completely exhaustive categories are found to be skillful relative to a climatological forecast. Ensemble sizes of approximately 10 can account for 90% of improvement in categorical forecasts relative to that for the average of individual forecasts. The improvements due to short-range ensemble forecasting (SREF) techniques exceed any due to doubling the resolution, and the error growth due to ICU greatly exceeds that due to different resolutions. If the authors' results are representative, they indicate that SREF can now provide useful QPF guidance and increase the accuracy of QPF when used with current analysis-forecast systems.
• Sanders, F., & Mullen, S. L. (1996). The climatology of explosive cyclogenesis in two general circulation models. Monthly Weather Review, 124(9), 1948-1954.
  More info

  Abstract: The occurence of explosive cyclogenesis is studied in two 180-day cold-season simulations by the Community Climate Models (CCM1 and CCM2) developed at the National Center for Atmospheric Research. The CCM1 run was realistic in some respects but produced relatively few cases over the North Atlantic Ocean and failed to show concentration just off the east coasts of the continents and north of the warm ocean currents. The intensity of cyclogenesis was underestimated in the CCM1 run. All of these flaws were mended in the CCM2 run, in which the climatology of explosive cyclogenesis closely resembled that in the real atmosphere.
• Schmitz, J. T., & Mullen, S. L. (1996). Water vapor transport associated with the summertime North American monsoon as depicted by ECMWF analyses. Journal of Climate, 9(7), 1621-1634.
  More info

  Abstract: The origins and transport of water vapor into the semi-arid Sonoran Desert region of southwestern North America are examined for the July-August wet season. Vertically integrated fluxes and flux divergences of water vapor are computed for the 8 summers 1985-1992 from ECMWF mandatory level analyses possessing a spectral resolution of triangular 106 (T106). The ECMWF analyses indicate that transports of water vapor by the time-mean flow dominate the transports by the transient eddies. Most of the moisture at upper levels (above 700 mb) over the Sonoran Desert arrives from over the Gulf of Mexico, while most moisture at low levels (below 700 mb) comes from the northern Gulf of California. There is no indication of moisture entering the Sonoran Desert at low levels directly from the southern Gulf of California or the tropical East Pacific. Water vapor from the tropical East Pacific can enter the region at upper levels after upward transport from low levels along the western slopes of the Sierra Madre Occidental of Mexico and subsequent horizontal transport aloft. The T106 ECMWF analyses, when only the mandatory level analyses are used, do not possess sufficient resolution to yield accurate estimates of highly differentiated quantities such as the divergence of the vertically integrated flux of water vapor. Even at a T106 resolution, the northern Gulf of California and the terrain of the Baja California peninsula are not adequately resolved.
• Stensrud, D. J., Gall, R. L., Mullen, S. L., & Howard, K. W. (1995). Model climatology of the Mexican monsoon. Journal of Climate, 8(7), 1775-1794.
  More info

  Abstract: The Mexican monsoon is a significant feature in the climate of the southwestern United States and Mexico during the summer months. Rainfall in northwestern Mexico during the months of July through September accounts for 60% to 80% of the total annual rainfall, while rainfall in Arizona for these same months accounts for over 40% of the total annual rainfall. Deep convection during the monsoon season produces frequent damaging surface winds, flash flooding and hail and is a difficult forecast problem. The Pennsylvania State University/National Center for Atmospheric Research mesoscale model is used to simulate 32 successive 24-h periods during the monsoon season. Mean fields produced by the model simulations are compared against observations to validate the ability of the model to reproduce many of the observed features. -from Authors
• Stoss, L. A., & Mullen, S. L. (1995). The dependence of short-range 500-mb height forecasts on the initial flow regime. Weather & Forecasting, 10(2), 353-368.
  More info

  Abstract: Forecast errors in the 500-mb geopotential height field over North America and adjacent ocean environs are calculated for the National Meteorological Center's Nested Grid Model (NGM). The eight winters 1985/96-1992/93 are examined. Errors are compared for the time-mean flow and for four recurring planetary-scale flow regimes, and the statistical significance of the differences is estimated. Overall, the NGM produces very accurate 500-mb height forecasts out to 48 h, with every forecast cycle of the study period exhibiting useful deterministic skill at 48 h when averaged over the study domain. During the first two winters of operational NGM implementation, the spatially averaged errors were noticeably greater than in subsequent winters. NGM error was essentially constant during the 1987/88-1992/93 winters. -from Authors
• Mullen, S. L. (1994). An estimate of systematic error and uncertainty in surface cyclone analysis over the North Pacific Ocean: some forecasting implications. Weather & Forecasting, 9(2), 221-227.
  More info

  Abstract: The AVN (aviation run) central pressures are, on average, 2.2 mb higher than the NH values. The AVN initial maps underestimate cyclone intensity at all central pressures values, but the error is biggest (3.9 mb) for deep systems (central pressures ≤980 mb). The absolute displacement error averages 210 km with standard deviation of 200 km. Position errors are largest for weak systems (central pressures ≥1000 mb), averaging 300 km and being more than double the 140 km value for deep lows. The aggregate mean vector displacement of 10 km is negligible. Cyclone positions for the AVN initial maps agree to within 130 km of the NH locations for rapid deepening lows (12 h pressure change ≤-12 mb), the smallest mean displacement error for any category examined. The life cycle of cyclones, as portrayed by the AVN initial charts, is too slow, with the AVN maps underestimating the 12-h deepening rate for rapid deepening lows by 5.4 mb, for all other deepening lows by 1.5 mb, and the 12-h filling rate for all cyclolytic lows by 2.4 millibars. -from Author
• Mullen, S. L. (1994). The impact of an envelope orography on low-frequency variability and blocking in a low-resolution general circulation model. Journal of Climate, 7(12), 1815-1826.
  More info

  Abstract: Sensitivity experiments with a perpetual January version of a low-resolution general circulation model (GCM) are conducted to investigate the influence of different, nonzero specifications of orography on low-frequency variability (LFV) and blocking in the Northern Hemisphere. The use of an envelope orography leads to significant changes in the distribution of the LFV (periods 10-90 days) over the Northern Hemisphere. The impact of envelope orography on blocking, as measured by an objective criterion, is then examined. The changes in blocking distribution over the North Atlantic make the model's climatology more consistent with observations. It is hypothesized that the changes in LFV and blocking over the North Atlantic are a response to differences in the orographic forcing downstream of the Rocky Mountains. -from Author
• Mullen, S. L., & Baumhefner, D. P. (1994). Monte Carlo simulations of explosive cyclogenesis. Monthly Weather Review, 122(7), 1548-1567.
  More info

  Abstract: The impact of initial condition uncertainty on short-range (0-48 h) simulations of explosive surface cyclogenesis is examined within the context of a perfect model environment. Eleven Monte Carlo simulations are performed on 10 cases of rapid oceanic cyclogenesis that occurred in a long-term, perpetual January integration of a global spectral model. The perturbations used to represent the initial condition error have a magnitude and spatial decomposition that closely matches estimates of global analysis error. -from Authors
• Mullen, S. L., & Smith, B. B. (1993). The dependence of short-range surface cyclone forecasts on the large- scale circulation: a preliminary assessment. Weather & Forecasting, 8(2), 235-247.
  More info

  Abstract: Sea level cyclone errors for two contrasting planetary-scale flow regimes, a long-wave trough verses a longwave ridge over western North America, are computed for the National Meteorological Center's Nested Grid Model (NGM) and "Aviation Run' of the Global Spectral Model (AVN). The study is performed for the 1987/8 and 1989/90 cool seasons (1 December-31 March). Statistically significant differences in forecast skill are found between the two flow patterns. This finding suggests that the utility of cyclone forecasts can be improved if model performance is documented for other recurrent, persistent flow regimes. -from Authors
• Smith, B. B., & Mullen, S. L. (1993). An evaluation of sea level cyclone forecasts produced by NMC's nested-grid model and global spectral model. Weather & Forecasting, 8(1), 37-56.
  More info

  Abstract: Aggregate errors can be summarized as follows: NGM forecasts of central pressure are too low (forecast pressure lower than analyzed) by 0.72 mb at 24 h and 0.66 mb at 48 h, while AVN forecasts are too high by 2.06 mb at 24 h and 2.50 mb at 48 h. Variance statistics for the pressure error indicate that AVN forecasts possess less variability than those of the NGM. Both mean absolute displacement errors and mean vector displacement errors are smaller for the AVN. The NGM moves surface cyclones too slowly and places them too far poleward into the cold air; the AVN possesses a smaller, slow bias only. Both models contain a weak cold bias as judged from the 1000-500-mb thickness over the cyclone center. -from Authors
• Mullen, S. L., & Baumhefner, D. P. (1989). The impact of initial condition uncertainty on numerical simulations of large-scale explosive cyclogenesis. Monthly Weather Review, 117(12), 2800-2821.
  More info

  Abstract: Compared to the control simulations, the inclusion of initial error produces a composite cyclone with maximum deepening rate that is slightly reduced and a 24 h period of most rapid deepening that is somewhat delayed. The absolute position error in the surface cyclone is approximately 100 km the first +36 h of the forecast then abruptly increases to 300 km by +48 h. We estimate that, on the average, the forecast error due to initial condition uncertainty is as large as that due to the modeling error associated with today's best operational models, whereas five years ago modeling error was much more important. Short-range forecasts of explosive cyclogenesis are much more sensitive to initial error than those for ordinary flows. -from Authors
• Mullen, S. L., & Baumhefner, D. P. (1988). Sensitivity of numerical simulations of explosive oceanic cyclogenesis to changes in physical parameterizations. Monthly Weather Review, 116(11), 2289-2329.
  More info

  Abstract: The total diabatic heating accounts for about one-half of the cyclone's deepening rate, with baroclinic dynamics accounting for the remaining part. The absence of diabatic heating also leads to a systematic error in the position of the cyclone. Surface fluxes of sensible heat are responsible for about one-half of the deepening rate due to diabatic processes, while latent heating due to grid-scale resolvable precipitation in conjunction with surface latent heat flux accounts for most of the remaining half. An increase in the surface drag over the ocean to its larger land value was found to be of comparable importance to both surface sensible heat flux and latent heat release, but was only half as important as the total diabatic heating. The case-to-case variability exhibited by the 11-member ensemble is examined. The potential danger in attempting to generalize results from a single case of explosive cyclogenesis as being representative of those for the ensemble average is illustrated. -from Authors
• Mullen, S. L. (1987). Transient eddy forcing of blocking flows.. Journal of the Atmospheric Sciences, 44(1), 3-22.
  More info

  Abstract: The net forcing of blocking flows by transient eddies having synoptic time scales is examined within the framework of quasi-geostrophic theory. Temporal filtering is used to distinguish the effects of transients having synoptic time scales. The eddy flux convergences for individual blocking events are then composited according to the locations of the blocks. The quasi-geostrophic potential vorticity equation relating geopotential tendency and the eddy flux convergences of heat and vorticity is solved for composite blocking conditions. The diagnosis is performed for both observed blocks and blocks simulated by a general circulation model. The geopotential tendencies due to transient eddies exhibit a quadrature relationship with the blocking pattern throughout the troposphere, with anticyclonic eddy forcing being located about one-quater wavelength upstream of the blocking anticyclone. The effects of transient disturbances on blocking flows as depicted by our diagnostic technique are contrasted with those for earlier studies. Reasons for the inconsistencies in the results of earlier studies are discussed.-from Author
• Blackmon, M. L., Mullen, S. L., & Bates, G. T. (1986). The climatology of blocking events in a perpetual January simulation of a spectral general circulation model.. Journal of the Atmospheric Sciences, 43(13), 1379-1405.
  More info

  Abstract: A variety of statistical comparisons is made between the fluctuations occurring in a 1200-day perpetual January simulation of a spectral general circulation model and those occurring in a 20-winter data set. Attention is focused on the persistent anomalies with lifetimes greater than one week. Over the Atlantic and Pacific oceans, we show that there is generally good agreement between modeled and observed persistent anomalies in their frequency of occurrence and mean lifetime. However, we note a striking deficiency in the simulation of persistent anomalies over the Soviet Union. We focus further on a subset of persistent anomalies, blocking highs, and show that the model blocks have vertical structure in agreement with observations. We also show an example of the development of a blocking event which follows an episode of explosive cyclogenesis. Finally, we show an example of the interaction of low- and high-frequency eddies during a period of blocking and at the termination of model blocking. The results of this study demonstrate that the internal dynamics and physics of the model by themselves are able to generate quite realistic blocking episodes over the wintertime North Atlantic and Pacific oceans. We conclude that blocking is a naturally occurring, internally generated phenomenon of the model and believe that model data can be used as a reasonable proxy for observations to study this phenomenon. -Authors
• Mullen, S. L. (1986). The local balances of vorticity and heat for blocking anticyclones in a spectral general circulation model.. Journal of the Atmospheric Sciences, 43(13), 1406-1441.
  More info

  Abstract: Blocking anticylones that appear in perpetual January simulations of a spectral general circulation model are examined. Blocks in three geographical regions are studied: the North Pacific, the North Atlantic and western North America. Local time-averaged balances of vorticity and heat are evaluated for composite cases of blocking. A number of common relationships emerged from these budgets, and are discussed. A quasi-geostrophic diagnosis of the ageostrophic motion field suggests that dynamical processes which strongly affect the vorticity balance may be more important to the maintenance of model blocks than processes which strongly affect the heat balance. The mountains appear capable of influencing the shape of the model blocks, but preliminary results indicate that orographic forcing may not be absolutely essential for the blocking process to occur in the model. -from Author
• Mullen, S. L. (1983). Computer simulation of atmospheric ' blocking'.. Weatherwise, 36(5), 232-233.
  More info

  Abstract: Climate models can help meteorologists understand 'blocking'. The newest generation of general circulation models is able to simulate blocking quite realistically. The influence of abnormal sea surface temperature conditions on blocking can be investigated with 'sensitivity experiments'.-D.G.Tout
• Mullen, S. L. (1983). Explosive cyclogenesis associated with cyclones in polar air streams.. Monthly Weather Review, 111(8), 1537-1553.
  More info

  Abstract: Sanders and Gyakum's 24-hr criterion (a central pressure fall that averages at least 1 mb h-1 for 24 h) of a rapidly deepening extratropical surface cyclone is adopted to investigate explosive surface cyclogenesis associated with cyclones in polar air streams and frontal waves induced by cyclones in polar air streams. A case is described in detail in which both the cold air vortex and frontal wave meet the criterion. Other examples of rapid cyclogenesis in which only the cold air cyclone or frontal wave satisfy the criterion are given.-from Author
• Mullen, S. L. (1982). Cyclone development in polar air streams over the wintertime continent ( North America).. Monthly Weather Review, 110(11), 1664-1676.
  More info

  Abstract: A polar air stream cyclogenesis over the wintertime North American continent is documented. The main difference between the land case and oceanic polar lows is that the land cyclone develops in an air mass with negligible water vapor content whereas oceanic ones clearly do not since they are characterized by enhanced cumulus activity during their early stage of development.-from Author
• Mullen, S. L. (1979). An investigation of small synoptic-scale cyclones in polar air streams.. Monthly Weather Review, 107(12), 1636-1647.
  More info

  Abstract: The large-scale environment of small synoptic-scale cyclones that occur in polar air streams over the wintertime North Pacific behind or poleward of major frontal bands is objectively documented by the compositing of meridional (latitude-height) cross sections for 22 cases. The cold air mass cyclones are found to be associated with deep baroclinity throughout the troposphere and are located on the low-pressure side of well-developed jet streams in regions of strong cyclonic wind shear. The lower troposphere is conditionally unstable and in the early stage of development is strongly heated from below by the warmer ocean.- from Author