Florida State University Libraries
Electronic Theses, Treatises and Dissertations The Graduate School
2017 Predictability and Dynamics of the Genoa Low: Case Study and Operational Considerations Michael Snyder
Follow this and additional works at the DigiNole: FSU's Digital Repository. For more information, please contact [email protected] FLORIDA STATE UNIVERSITY
COLLEGE OF ARTS AND SCIENCES
PREDICTABILITY AND DYNAMICS OF THE GENOA LOW: CASE STUDY AND
OPERATIONAL CONSIDERATIONS
By
MICHAEL SNYDER
A Thesis submitted to the Department of Earth, Ocean, & Atmospheric Science in partial fulfillment of the requirements for the degree of Master of Science
2017 Michael Snyder defended this thesis on March 30, 2017. The members of the supervisory committee were:
Jeffrey Chagnon Professor Directing Thesis
Robert Hart Committee Member
Vasu Misra Committee Member
The Graduate School has verified and approved the above-named committee members, and certifies that the thesis has been approved in accordance with university requirements.
ii
TABLE OF CONTENTS
List of Figures ...... iv Abstract ...... vi
1. INTRODUCTION ...... 1
2. METHODOLOGY ...... 9
3. RESULTS AND DISCUSSION ...... 16
3.1. PREDICTABILITY OF THE GENOA LOW ...... 16
3.2. GENOA LOW FORMATION AND THE MISTRAL WIND ...... 25
4. CONCLUSIONS ...... 34
References ...... 38
Biographical Sketch ...... 40
iii
LIST OF FIGURES
1 Figure 1. Manual analysis of surface potential temperature (C) (dashed) and MSLP (mb) with surface observations of sky cover, wind (kts), temperature (C), and dew point temperature (C) from 12Z 15 Nov 2007 (Figure 4 from McTaggart-Cowan, et al. 2009) ...... 2
2 Figure 2. Terrain map of southwestern Europe and the Mediterranean Sea...... 3
3 Figure 3. Time versus the natural logarithm of pressure in the center of the Genoa low broken into three developmental stages (Figure 12 from Buzzi and Tibaldi, 1977) ...... 4
4 Figure 4. The spatial domain of COSMO-LEPS as indicated by the green box. The contours indicate the elevation in meters that was simulated in the numerical simulations (COSMO, 2008)...... 10
5 Figure 5. A schematic of the ensemble method to assessing the likelihood of a weather event and the spread of numerical solutions which arise from small perturbations in the initial conditions. The ensemble spread is analogous to the forecast uncertainty of the specific weather phenomenon shaded in grey. The ensemble spread then can be projected to a set probabilities that the particular phenomenon will occur represented by the map on the right. (Figure 3, Bauer et al. 2015) ...... 11
6 Figure 6. Example MSLP (mb) field (top) and 10 meter speed (kts) (bottom) demonstrating the locations of the subdomains (boxed regions) used to isolate the Genoa low (top) and the mistral winds (bottom)...... 14
7 Figure 7. Surface analysis (Crown copyright ©) for case 1 (top) and case 2 (bottom) (UKMET) ...... 17
8 Figure 8. Synoptic evolution for Case 1 (top) and case 2 (bot). 500mb geopotential height contours with 850mb isotachs (kts) (top) and shaded MSLP (mb) (bot) for each case...... 18
9 Figure 9. MSLP (mb) (shaded) for each ensemble initializations of case 1 at 1, 2, 3, and 4 days prior to cyclogenesis with time fixed at the cyclogenesis time. Individual ensemble members are numbered. Missing members were not archived...... 20
10 Figure 10. Spaghetti plots of the minimum MSLP (mb) through the forecast period for case 1 ensemble simulations with event lead times of 1-4 days. The red line is the mean of all ensemble members, and the black dashed line marks the time of cyclogenesis...... 22
11 Figure 11. Maps of the time evolution of the ensemble mean MSLP (mb) for the 3 day forecast in case 1 with time progressing from the 69-hour forecast time to the 81 hour forecast time from left to right...... 23
iv 12 Figure 12. Shaded is the ensemble variance in MSLP (mb) for case 1 at the cyclogenesis time for each ensemble forecast lead time. Note: The scale for the 1 and 2 day lead times ranges from 0 to 4, and the scale for the 3 and 4 day lead times ranges from 0 to 12. Values below .5 mb are white...... 24
13 Figure 13 Ensemble standard deviation in minimum MSLP (mb) (top) and 10m maximum mistral wind (bot) for simulations initialized 1-4 days prior to Genoa low development. The black dashed line marks the time of cyclogenesis...... 26
14 Figure 14. Maps of MSLP (mb) (left) and 10 m wind speed (kts) (right) for case 1 for all ensemble members. This ensemble run was initialized 3 days prior to the event. Time is fixed at the cyclogenesis time...... 28
15 Figure 15. Ensemble spaghetti plots of minimum MSLP (mb) (top) and maximum 10m mistral winds (kts) (bot) for case 1 (top) and case 2 (bot) for lead times at 3 days (left) and 4 days (right). The solid black line is the ensemble mean for each EPS run, and the vertical dashed black line marks the start of cyclogenesis. The green lines are the ensemble members chosen because they produced deep Genoa lows. The red lines are ensemble members chosen because they did not generate a Genoa low. The same 4 ensemble members in each case are highlighted in the same colors in the spaghetti plots of the maximum mistral wind field...... 29
16 Figure 16. Correlation of minimum MSLP and maximum mistral winds at a 3 day event lead time for case 1 (top) and at a 4 day event lead time for case 2 (bot). The correlation plots on the left side are at the time of cyclogenesis, and the plots on the right are correlations between the minimum MSLP and the time-lagged maximum wind speed. The time lad is 6 hours for case 1 and 12 hours for case 2. The line is the least squares regression line, and the correlation coefficient is labeled on each plot...... 31
17 Figure17. Time evolution of vorticity (shaded) ( 10 ^(-4) )(s^(-1)) and MSLP (contours) with 10m wind (vectors) from case 2, ensemble member 9. Time proceeds from left to right at 3-hourly segments...... 33
v ABSTRACT
The rapid development and sub-synoptic scale nature of the Genoa low in the Mediterranean Sea poses a forecasting challenge for United States Air Force (USAF). The Genoa low is a high-impact event for several Department of Defense (DOD) locations located in southern Europe, especially in the Po River Valley of northern Italy. This study evaluates the predictability and dynamics of the Genoa low extending to a 4-day event lead time as is required by the mission protocols at the affected locations. Two Genoa low case studies are analyzed: 16 Feb 2015 (case 1), and 13 July 2016 (case 2), using the COnsortium for Small-scale MOdeling Limited-area Ensemble Prediction System (COSMO-LEPS). Ensemble prediction systems provide a range of possible forecast outcomes given the uncertainty in initial conditions, boundary conditions, as well as model physics. As such, ensembles are used to assess and analyze the predictability of the Genoa low. The analysis demonstrates several key findings concerning the Genoa low. The Genoa low is only weakly predictable at a lead time of 4 days. It is shown that only a small fraction of ensemble members (approximately 25%) met the Genoa low verification thresholds at this lead time. Ensemble spaghetti plots and maps of the ensemble variance show that the possibility of low formation at longer lead times is most effectively visualized using maps of ensemble variance. Traditional postage-stamp plots and minimum MSLP plots contain too much noise and variability to permit a forecaster to extract a signal indicating possible low formation. The formation of the Genoa low is associated with strong mistral winds. It is demonstrated that all ensemble simulations that were successful in identifying cyclogenesis also produce strong mistral winds, i.e., the strength of the mistral winds is anti-correlated to the minimum MSLP of the Genoa low. This linkage implies a potential dynamical connection between the two features. Further investigation shows that the mistral jet may exert an organizing influence on the Genoa low via a vorticity seeding mechanism. Time-lagged correlations show that the mistral jet amplifies several hours prior to cyclogenesis. The amplification is associated with mesoscale vorticity generation on the eastern periphery of the jet. These vorticity centers were subsequently shed into the target region where cyclogenesis occurs. Such a small-scale and rapid-developing dynamical link between the mistral winds and the Genoa low implies a limit on the predictability of the Genoa low. This study concludes that weather forecasting operations in the USAF would
vi benefit from expansion of current ensemble prediction systems, not only for the purpose of improving the Genoa low forecast process and performance but also to better inform the mission planners of the limitations and uncertainties of predicting the Genoa low.
vii CHAPTER 1
INTRODUCTION
Genoa lows form in the Ligurian Sea near the Gulf of Genoa which is located near Genoa, Italy along the Mediterranean coast. They have long posed a difficult forecast challenge to meteorologists in the region. Genoa lows produce severe weather impacts including heavy rain, floods, landslides, heavy snow, and strong winds. For example, between 30 May 2013 and June 2, 2013, a Genoa low formed and slowly moved eastward as it produced rainfall amounts in excess of 11 inches across a large area in Germany. Widespread flooding occurred as many of the rivers reached 100-year flood levels which forced tens of thousands to evacuate. Despite the evacuations, there were 25 fatalities as a result of the extreme floods. Air Worldwide – a catastrophic modeling center - estimated that insured losses in Germany from this flooding event was approximately $5.5 billion, and that the economic loss was much higher (Insurance Journal, 2013). Another Genoa low event on the 4 Nov 2011 produced a third of the average annual rainfall in a single day over Liguria, Italy which generated flooding that killed 6 people and caused over $100 million in damage (Fiori et al. 2014). Figure 1 is a manual analysis of a typical Genoa low which formed on 15 Nov 2007. The low pressure center is located in the Genoa low target zone between Corsica and the Mediterranean coast of France. There are strong northerly winds advecting cold air from the north which is reflected in the bowing of the potential temperature contours and the wind observations to the west of the low pressure center. These strong winds are called mistral winds, and the potential link that they have to the development of the Genoa low is examined in later sections of this thesis. Climatology of Mediterranean cyclones performed by Trigo et al. (2001) indicates that cyclogenesis occurs in the Gulf of Genoa more than any of the other Mediterranean low regime regions. Additionally, the phenomenon occurs year-round, and lows tend to be deeper but less frequent in the winter than lows that develop in the summer. In the winter, low level baroclinicity over the northern Mediterranean coast is stronger, and the synoptic scale upper level troughs moving over the Alpine region are deeper and more prevalent. These features aid in producing the stronger Genoa lows in the winter. Conversely in the summer, the frequent Genoa low formation relies less on the meridional temperature gradient between the
1 Mediterranean Sea and the northern coast and more on the diurnal fluctuations of surface temperature inland, south of the Alps. The diurnal warming aids in forming thermal lows which become enhanced by lee side cyclogenesis and relatively weaker upper level troughs propagating through the upper level flow.
Figure 1. Manual analysis of surface potential temperature (C) (dashed) and MSLP (mb) with surface observations of sky cover, wind (kts), temperature (C), and dew point temperature (C) from 12Z 15 Nov 2007 (Figure 4 from McTaggart-Cowan, et al. 2009).
Part of what makes the Genoa low such a difficult event to forecast is the complex terrain in the region. Figure 2 shows various mountain chains that are located near the Mediterranean Sea. The most notable terrain influence on Genoa cyclogenesis is the Alps. A northerly flow across the Alps induces lee side cyclogenesis in close proximity to the area of interest involved with a Genoa low. Numerical experiments (Egger, 1972) demonstrated that the shape of the mountain chain is critical to cyclogenesis. His idealized experiments simulated the Alps as a horizontal east to west barrier. The results were compared to model runs with the actual arc shape of the Alps. Eggar’s work showed that the early, intense stages of cyclogenesis only
2 Figure 2. Terrain map of southwestern Europe and the Mediterranean Sea.
occurred when the arc shape of the Alps was simulated. He concluded that this was partially due to the deformation of the thermal field caused by the partial blocking of cold air advection as well as the channeling effect which occurs in the gap between the western edge of the Alps and the Pyrennes. The initial phase of development of the Genoa low is characterized by rapid cyclogenesis in the Ligurian Sea (Egger, 1972). Buzzi and Tibaldi (1977) used a climatology of the Genoa low to quantify its developmental stages. Figure 3 shows the rate of pressure decrease with time, and three clear stages of development can be deduced. Note that a drop in MSLP corresponds to a positive change in the natural logarithm of ∆P in Figure 3. Stage A, the first 12 hours of cyclone development, marks the initial stage of cyclogenesis which is also the period with the most rapid pressure drop. The growth during the initial phase happens at a time scale that is too fast and on a spatial scale that is too small to be explained by quasi geostrophic theory. During the initial development phase, the upper-level, synoptic support for cyclogenesis is out of phase with the cyclone at the surface because those mechanisms are slowed by the Alps. The initial
3 development is linked to the enhanced baroclinic zone south of the Alps brought about by the northerly winds funneling between the Alps and the Pyrenees as well as the convective feedback brought about by developing thunderstorms in the beginning stage of cyclogenesis that acts to organize the low level convergence. It is not until phase B when the upper level trough synchs with the surface cyclone to further mature the system through quasi-geostrophic processes and it takes on the character of a typical mid-latitude cyclone. In this way, the Genoa low can be considered a hybrid type of cyclone that does not align with typical development seen in either tropical or extra-tropical cyclones. This initial phase of rapid growth and its link to the local northerly low level jet that is formed between the Alps and the Pryenees is explored further in later chapters of this thesis as it is a potential source in limiting the predictability of Genoa lows in numerical models.
A B C
Figure 3. Time versus the natural logarithm of pressure in the center of the Genoa low broken into three developmental stages (Figure 12 from Buzzi and Tibaldi, 1977)
4 The northerly winds channeled between the Alps and the Pyrenees are called mistral winds. The United States Naval severe weather guide for Mediterranean ports characterizes mistral winds as a rapid onset of winds which can increase from 20 knots to 40 knots within 30 minutes, and waves grow to 3-4 meters. Severe mistral wind events have resulted in wind gusts that reach 85 knots (Englebretson, 1989). This thesis will explore a potential link between the vorticity created by mistral winds and the trigger for the development of the Genoa low. Such a small-scale feature poses difficulties for regional and global numerical forecasts. Genoa lows are of great interest to the United States Department of Defense (DoD) because there are numerous locations which are affected by the associated severe weather impacts, especially in the Po Valley of Italy. These locations include Aviano Air Base, Italy, Camp Darby, Italy, Ghedi, Italy, Vicenza, Italy, and Camp Bondsteel, Kosovo. These DOD sites house important missions and assets which includes the 31st Fighter Squadron, the only Air Force fighter squadron south of the Alps, the North Atlantic Treaty Organization (NATO) headquarters for peacekeeping in Kosovo, and other multinational coalition forces that are critical to supporting U.S. and NATO interests. The combined resources at these sites amount to five thousand personnel and over two billion dollars’ worth of DOD assets (21 OWS, 2014). The organization that is responsible for forecasting and tracking the weather impacts for the DOD locations in the region affected by Genoa lows is the USAF’s 21st Operational Weather Squadron (OWS) located in Kapaun Air station, Germany. This unit produces the terminal aerodrome forecasts and issues all of the resource protection weather advisories, watches, and warnings for these sites. Additionally, it produces the flight weather hazard charts consisting of icing, turbulence, and thunderstorm forecasts and tailors them into the flight weather briefings for every flight in the area of responsibility. In order to maintain mission readiness at all times, the most sensitive units in the region affected by the Genoa low require a four-day lead time of a Genoa low event. This allows enough time for contingency procedures to take place so that the critical missions in those areas are uninterrupted. The 21st OWS has identified the Genoa low as a high priority forecast challenge that needs to be improved. Given the transient nature of Air force personnel and the perishable nature of forecasting experience in this area of responsibility, the current forecast performance for the Genoa low does not meet the standards set forth by the units that the 21st OWS support. Currently, forecasters are routinely having difficulty diagnosing Genoa cyclogenesis that is
5 already occurring, and predicting the event at a relevant lead time is even more difficult. There are several forecast factors involved in the genesis of a Genoa low which make forecasting the Genoa low difficult. The complex terrain in the region consisting of several prominent mountain ranges alters the behavior of the atmospheric flow patterns. The winds at many different levels are subject to funneling, blocking, upslope, and downslope which can aid in creating localized atmospheric features like, jets, regions of increased shear, pressure changes, and enhanced temperature advection. Additionally, the Mediterranean Sea creates a temperature gradient with the nearby landmass. It can also play a role in altering the composition of the atmosphere through heat and moisture fluxes off the water. The mistral winds advect cold air from the north over the Mediterranean Sea on the west side of the Alps. This results in an enhanced baroclinic zone in the region. Convective elements form near this region of frontogenesis, and the latent heat release aids in dropping the pressure and enhancing low level convergence in the region. The convective feedback generates more vorticity on top of the existing environmental vorticity associated with the upper level trough in the synoptic regime (McTaggart-Cowan, et al. 2009). These features occur on small spatial and time scales, and accounting for all of them makes for a difficult forecast challenge, especially for unseasoned forecasters. Because of the consistent forecast performance deficiency of the Genoa low, the 21st OWS has proposed a Ph.D. dissertation project that would culminate in the development of a Genoa low forecast tool that would serve as a lasting solution that would persist through the constant personnel changes. The desired end state of the proposal is a statistically based, interactive decision aid that would allow a forecaster to input current weather conditions and derive the anticipated timing, track, and intensity of the event. However, before jumping right into an extensive data review for formation, intensity, and track projections, it is important to consider the predictive limitations involved with the Genoa low which is the intent of this thesis research. The current forecast process involves mostly global forecast models, primarily the Unified Model (UM) developed by the United Kingdom Met Office. The UM’s 17 kilometer grid spacing is effective in resolving the larger mesoscale to synoptic scale features, but the resolution is inadequate in resolving a small, rapidly developing feature like a Genoa low which depends on many small, local mechanisms. Air Force Weather has recently started to implement the use of ensemble forecasting. This strategy allows for the uncertainty of initial conditions and
6 parameterization schemes to be accounted for. It also allows for predictability to be assessed by analyzing the growth in the ensemble spread. Ensemble forecasts indicate the sensitivity of the numerical prediction to small perturbations in the calculations which is important to an operational forecaster who must consider the range of possible solutions and the range of impacts that may result. Compared to the use of global weather models, the operational use of ensemble weather models will better equip the forecaster with a range of possible scenarios without the need to compare multiple models. Given the nature of the cyclogenetic processes involved with the genoa low, especially given the complex terrain, the grid spacing and time step in the numerical model must be small enough to adequately resolve rapid growth in the early stage of development. In addition to the limitations to the predictive skill of numerical modeling systems brought about by the imperfect representation of complex physical and dynamical processes, the dynamical system is subject to inherent instabilities and sensitivities to initial conditions. Errors from chaotic noise at small scales can project upscale and influence the larger scale features that forecasters expect the numerical models to have the ability to accurately resolve (Bauer, at al. 2015). Lorenz (1963) demonstrated that two states in any real system which differ by imperceptible amounts may eventually evolve into two considerably different states. Therefore, if there is any error in observing the initial state, accurately predicting the conditions in the distant future may well be impossible. There are intrinsic limits to the accuracy of numerically calculating the evolution of the chaotic, non-linear dynamical system that is the atmosphere. The limit to the predictive skill for synoptic scale, high impact events is about 1-2 weeks, and the limit for small-scale events is between hours and days (Bauer et al. 2015). The spatial and time scales involved in the Genoa low are relatively small and rapidly evolving. This poses a challenge to the forecaster trying to accurately forecast the event out to a 4-day lead time per the USAF requirement. The primary purpose of this investigation is to characterize the predictability of Genoa cyclogenesis for two cases. A secondary purpose is to identify possible physical mechanisms of predictability degradation. The analysis of the predictability is performed with the intention of providing guidance to operational forecasting activities of the USAF, particularly with the respect to the utility of ensemble prediction systems. To achieve these goals, two separate Genoa low cases are analyzed in a regional ensemble prediction system. Each ensemble member is
7 initialized with small differences in the boundary conditions, initial conditions, and parameterization schemes, thus allowing for the assessment of the predictability of the event at different lead times. It also allows for the close analysis and comparison of members which successfully predict the event to the members which did not. From the comparison, indications to the key physical processes which may degrade the predictability of the Genoa low can be drawn. The remainder of this thesis is organized as follows. Section 2 covers the research method and experimental design. Section 3 presents the results. Finally, section 4 discusses the results, draws conclusions, and sets the path forward for the continuation of the Genoa low research prescribed by the 557th Weather Wing and the 21st OWS.
8 CHAPTER 2
METHODOLOGY
In order to investigate the predictability of Genoa cyclogenesis, two Genoa low events were analyzed: 16 Feb 2015 (case 1), and 13 July 2016 (case 2). These two cases were chosen because they were both classic Genoa lows with significant impacts, and they formed in two different meteorological seasons. The persistent heavy precipitation produced by case 1 contributed to fatal floods and landslides in Italy (Davies, 2015). The strong winds produced by case 2 altered the 2016 Tour de France based on the forecasted 100 km/hour winds along the southwestern edge of the Alps (Wynn, 2016). Although the sample size is too small to make robust conclusions, analyzing an event from the winter as well as the summer also offers an opportunity to investigate the seasonal dependence of the predictability of the Genoa low. Not every Genoa low has such damaging impacts as these two events, but the choice of stronger, high impact cases made the analysis of the defining features of a Genoa low easier to distinguish. Both cases were analyzed in depth using an operational class regional ensemble prediction system (EPS), the COnsortium for Small-scale MOdeling Limited-area Ensemble Prediction System (COSMO-LEPS) COSMO-LEPS is operated and maintained jointly by the European Center for Medium- Range Weather Forecasts (ECMWF) and partners in Germany, Greece, Italy, and Switzerland. The purpose of the COSMO project is to improve the short-to-medium range forecasts for extreme and localized weather events. Specifically, COSMO-LEPS combines the ability of a global EPS to synoptic scale weather with the ability of a limited area model to resolve the regional details of atmospheric features on the mesoscale, primarily in regions with complex orography (Marsigli, et al. 2004). COSMO-LEPS has a 144-hour forecast range, and it operates with a 10 km horizontal resolution with a 511 by 415 point grid arranged on rotated latitude- longitude coordinates (COSMO, 2008). Figure 4 shows the domain of the EPS as well as the terrain that the model takes into account. The COSMO-LEPS EPS consists of 16 members which are uniquely initialized with random perturbations to the initial conditions and adjustments to the parameterization schemes. The boundary conditions are perturbed by four global model operational runs of the ECMWF, Deutscher Wetterdienst, National Centers for Environmental
9 Prediction, and the Japan Meteorological Agency (Marsigli et al. 2014). This EPS is suitable for this research because it is designed specifically to resolve the small-scale features which aid in early rapid development stage of Genoa cyclogenesis. The complex local terrain is widely accepted to have a major contribution to the evolution of the Genoa low, so the terrain in the model is an important component to the validity of the model’s solution.
Figure 4. The spatial domain of COSMO-LEPS as indicated by the green box. The contours indicate the elevation in meters that was simulated in the numerical simulations (COSMO, 2008).
The experimental design of this research treated the 16 members of the COSMO-LEPS EPS as a set of experiment as a set of equally likely realizations of the system evolution given the uncertainty in the initial state of the atmosphere. There are many advantages to using an EPS for this study. First, the EPS provides a virtual laboratory for investigating the predictability of the Genoa low. Second, with respect to identifying dynamical mechanisms regulating the
10 evolution and the predictability of the lows, the use of an EPS eliminates the need to catalogue multiple cases to be reviewed in a single numerical model. The sixteen unique numerical simulations provides a substantial sample size for determining statistics. This strategy permits the analysis of the predictability of the numerical solution for each of the cases at different lead times. The random perturbations in the initial conditions and parameterization schemes represent the uncertainty of numerical modelers to the initial conditions and the lack of understanding or the means of numerically representing the complex physical processes involved. Figure 5 is a schematic of the EPS forecast strategy and its application to measuring the forecast uncertainty. The ensemble members are created using perturbations to the initial conditions and the parameterization schemes executed by the dynamical core of the numerical prediction system which are representative of the errors and uncertainties within them. With increasing time, the individual ensemble member solutions diverge into a range of differing solutions to the forecasted event. The statistics of the ensemble spread of solutions then can be projected onto a map of the probabilities that the event will occur which is a useful tool for the forecaster. The main focus of this methodology is the statistical and physical nature of this ensemble member spread specific to the Genoa low when the EPS is initialized at different lead times.
Figure 5. A schematic of the ensemble method to assessing the likelihood of a weather event and the spread of numerical solutions which arise from small perturbations in the initial conditions. The ensemble spread is analogous to the forecast uncertainty of the specific weather phenomenon shaded in grey. The ensemble spread then can be projected to a set probabilities that the particular phenomenon will occur represented by the map on the right. (Figure 3, Bauer et al. 2015)
11 The method of analysis employed was as follows. The USAF desires a 4-day lead time for threat-based-operations forecasting, so the longest lead time chosen for each Genoa low was 4 days. Model initializations at 3 days, 2 days, and 1 day prior to each event were also used to test how the predictability changes as a function of lead time to the event. MSLP and 10 meter wind maps were plotted so that subjective identification of the successful and unsuccessful members could be accomplished. The subjective analysis of these surface weather fields consisted of identifying the ensemble members which were successful in the Genoa simulation and which were not. Distinctive similarities within these subsets of ensemble members as well as the contrasting features between the subsets were identified. These features were then investigated to assess whether they contributed to the predictive limitations of the Genoa low. In order to identify errors in the forecast, the simulations were compared to a verification of the event comprised of the zero-hour COSMO-LEPS initializations through the forecast period. An ensemble member was deemed to have predicted a Genoa low if it was deemed a success if it simulated a low pressure center in the target area with a closed isobar within 4 mb of the low pressure in the verification. Furthermore, the success of the ensemble member was subject to manual analysis to determine if the simulated Genoa low exhibited the mesoscale nature of a Genoa low with a relatively tight, closed circulation in the target cyclogenesis region. Identifying the success rate of the EPS on a member-by-member basis allowed for the comparison of meteorological fields in the successful versus the unsuccessful simulations. Ensemble statistics were also computed for each run at each forecast lead time interval to characterize the predictability of the Genoa low. The forecast uncertainty within the EPS was quantified by analyzing the differences between ensemble members. Maps of ensemble variance in MSLP as a function of location in the domain was used to identify the geographical and feature-dependent nature of the forecast uncertainty. The analysis of forecast uncertainty focuses on a target area in which genoa low formation occurs. Another purpose for the computation of ensemble variance as a function of location was to test that the EPS was within the predictive limits of its dynamical core with respect to time at the longest event lead time of 4 days. This allows for verification that the variance found in the ensemble solutions specific to the target region are reflective of the predictive limits of the Genoa low rather than the predictive limits of the EPS across the entire domain. The ensemble mean was also mapped for each case. The extent to which many forecasters tend to use an EPS does not go beyond an analysis of the time
12 evolution of the ensemble mean, so the mean was compared to the control for each case to assess its accuracy and relevance to the forecaster. In addition to analyzing the overall predictive skill of the Genoa low, the potential role of specific regional flow features in determining the predictability of the Genoa low is investigated. Specifically, this research focused on characterizing the link between the mistral winds and Genoa cyclogenesis. In order to quantify the link between the initial phase of Genoa cyclogenesis and the mistral winds, the two features needed to be diagnosed in isolation. A subdomain was used to isolate the respective features during their evolution for each ensemble member. Figure 6 shows an example of the subdomains used in this analysis. They were devised by observing the regions in which the mistral winds and the Genoa low develop and assigning a new grid domain in which calculate the statistics of the EPS simulations. In order to properly quantify the strength of the northerly mistral winds, only the northerly component of the 10 meter wind field was calculated in the subdomain. The mistral amplitude is defined here as the maximum northerly component of winds funneled through the Rhine River valley into the northern Mediterranean Sea. From the data in each subdomain, ensemble variability and standard deviation in the MSLP and 10 meter wind fields were computed at each forecast lead time interval. This allowed for the identification of abrupt shifts in the forecast uncertainty specific to each flow feature. Additionally, it incorporated some perspective to the link between the mistral wind feature and the cyclogenesis. Spaghetti plots of MSLP and mistral wind strength within the masks revealed that the ensemble members which produced deeper Genoa lows tended to also produce stronger mistral winds, and the ensemble members which produced weaker Genoa lows or failed entirely to produce a low tended to have weak mistral winds. The MSLP and mistral wind masks provided a simple means of measuring the variability of the EPS simulations relative to the control in these fields. Comparisons of the standard deviations for each event lead time gave an additional quantified measure of the predictive nature of the Genoa low. To quantify the link between the mistral winds and Genoa cyclogenesis within the subdomain, correlation coefficients across all ensemble members at the time of cyclogenesis were generated. Furthermore, to test the causality of the mistral wind strength to Genoa cyclogenesis, correlations between MSLP at the time of cyclogenesis to the time lagged strength if the mistral winds prior to Genoa cyclogenesis were calculated for every 3-hour interval previous to cyclogenesis out to 24 hours. This served as a test to determine a potential causal
13 relationship between the mistral winds and the Genoa low in an effort to identify a predictable precursor to a Genoa low event.
Figure 6. Example MSLP (mb) field (top) and 10 meter speed (kts) (bottom) demonstrating the locations of the subdomains (boxed regions) used to isolate the Genoa low (top) and the mistral winds (bottom).
14 Further exploration into the predictability of the Genoa low required a more in-depth analysis into some of the smaller scale features involved in the process. The statistical analysis performed on the mistral winds and Genoa cyclogenesis isolated a number of successful and unsuccessful model simulations. These cases were used for further analysis of small scale processes which would give numerical models difficulty, even at short forecast lead times. It was identified that the Genoa low forms on the eastern fringe of the strong northerly mistral jet. Along the boundary of the mistral jet exists a stark wind speed gradient. The shear vorticity along that boundary of the mistral jet results in small cyclonic circulations which may act to trigger or aid in the rapid development of the Genoa low. To provide evidence that this process is a potential limiting factor on predictability of Genoa cyclogenesis in numerical models, surface vorticity maps were generated and compared between the successful and unsuccessful EPS members.
15 CHAPTER 3
RESULTS AND DISCUSSION
3.1. PREDICTABILITY OF THE GENOA LOW
The two Genoa low cases chosen for this research are synoptically summarized in Figures 7 and 8. The surface analysis from the United Kingdom national weather service (UKMET) for case 1 shows an occluded front bowed out over Corsica and Sardinia from a 1007mb low centered between Corsica and the southern coast of France. The occlusion wraps northwestward over Spain and over the Atlantic coast of western Europe. The surface map for case 2 was analyzed 12 hours before cyclogenesis started occurring in the Genoa low region, but the cold front is located at the leading edge of the mistral winds which precede the beginning of the event. There is a prominent bow in the cold front in that specific location which is indicative of the enhanced cold air advection resulting from the funneled winds between the Alps and the Pyrenees. There is a 1009mb low pressure center stamped over central Italy which was just one of the lows to eject out of the Gulf of Genoa during this pattern. Neither case features a warm front as is typically seen the classic Norwegian cyclone model; the absence of the warm front is l likely due to the Genoa lows not yet being fully developed, and the only temperature advection mechanism was the northerly mistral winds.
Figure 8 demonstrates the evolution of the two Genoa low cases analyzed control data at 12-hour intervals. Owing to the rapid intensification, the beginning stages of the Genoa low are captured by the control run data because some of the stages of growth occur between the 12- hourly initializations. Nevertheless, the resulting structure and strength of the Genoa low is captured for each case. Figure 8 also presents the 500mb geopotential height overlayed with the 850mb isotachs. For both cases, there is a prominent 500mb trough over western Europe which deepens as it moved southeastward over the Ligurian Sea. The pressure at the surface drops and organizes into a closed cyclonic circulation when this trough makes its way to the south. At the same time, the winds on the upstream (western) side of the trough were oriented generally from north to south. These strong low level winds indicate the strong mistral winds which occurred in both cases. Specifically for case 2, there are two pronounced upper level short waves which
16 Surface Analysis – 00Z 13 Jul 2016
Surface Analysis – 00Z 16 Feb 2015
Figure 7. Surface analysis (Crown copyright ©) for case 1 (top) and case 2 (bottom) (UKMO)
17
) )
mb Pressure (
Figure 8. Synoptic evolution for Case 1 (top) and case 2 (bot). 500mb geopotential height contours with 850mb isotachs (kts) (top) and shaded MSLP (mb) (bot) for each case.
18 revolve around the exterior of the broader upper level low centered farther north. Additionally, there were persistent winds on the lee side of the Alps which helped to lower pressure locally. As each trough swung over the region, this area of lower pressure on the lee of the Alps was kicked south and east. The focus of this research was the secondary development seen in the 00Z 15 July 2015 and 12Z 15 July 2016 panels for case 2 in Figure 8 which resulted in the Genoa low which slid down the west coast of Italy.
The ensemble simulations indicated that, as expected, the variability and predictability of the forecasts were highly dependent on the lead time to cyclogenesis. Figure 9 presents the variety of ensemble solutions for case 1 for simulations initialized at various lead times to the event. A wide range of scenarios was depicted at the 4-day lead time, and only 4 of the 16 members successfully simulated the occurrence of a Genoa low event. A large majority of the members showed a general decrease in the pressure in the Mediterranean near Italy, but most of these members did not drop the pressure past 1010 mb. Furthermore, the area enclosed by the lowest isobar was typically too broad to be characterized as a Genoa low. The ensemble runs which were initialized closer to the time of the event saw more success. For the ensemble initialized 3 days prior to the event, closer to half of its members successfully resolve the Genoa low. Ensemble members 1, 2, 3, 5, 7, and 10 successfully generated a Genoa low resembling that seen in the control for case 1, but the remaining cases failed in this regard. This trend of increased forecast accuracy continued for the ensembles initialized 1 and 2 days prior to the event.
A full breakdown of the subjective verification for the success of the ensemble model for both cases presented in Table 1 which shows that the EPS success rate increases monotonically from 38% at the 4-day lead time to 88% at the 1-day lead time. The success rate for case 1 is similar with a 25% success rate at the 4-day lead time that increases with the shorter event lead times to 80%. However, for case 1, there is a bigger jump in the success rate between the 3-day and 2-day lead times. This suggests that some factor initial atmospheric conditions present for the 2-day lead time EPS run made the Genoa low more predictable than were present for the initialization of the EPS at a 3-day lead time.
19 Figure 9. MSLP (mb) (shaded) for each ensemble initializations of case 1 at 1, 2, 3, and 4 days prior to cyclogenesis with time fixed at the cyclogenesis time. Individual ensemble members are numbered. Missing members were not archived.
20 Table 1. The successful and unsuccessful ensemble members for both cases across all 4 event lead times. Case 11 Day Lead Time 2 Day Lead Time 3 Day Lead Time 4 Day Lead Time Successes 1213 64 Failures 338 12 Success Rate 80%81%43%25%
Case 2 Successes 1410 95 Failures 2577 Success Rate 88%67%56%38%
Figure 10 shows the ensemble spaghetti plots of the minimum MSLP field for case 1 across all four lead times. These plots demonstrate the sensitivity of the MSLP field in the target area to the initial conditions of the numerical modeling system. As expected, the unique model solutions diverge with time. The minimum pressure curves in the spaghetti plots show a persistent gradual deepening of the pressure leading up to the time of the event. This is a reflection of the synoptic environment being influenced by the trough aloft and a synoptic level deepening of the pressure across a large part of the domain. At the time of the event, however, the pressure drops more rapidly than the surrounding environment. This secondary deepening, when it occurs, is the signal associated with Genoa cyclogenesis. The minimum pressure associated with Genoa low is anomalous to the synoptic pressure decrease. The spread in the ensemble simulations remains low and relatively concentrated about the mean during the deepening phase for both the 1 and 2-day lead times. For both of those forecast ranges, every ensemble member exhibits a defined drop in MSLP starting at the cyclogenesis time. It is also seen that while the ensemble spread is relatively small for those two lead times, there is an observable increase in the spread after the time of cyclogenesis. Furthermore, the ensemble mean for the first two lead time intervals reflects the characteristics of Genoa cyclogenesis.
21 Figure 10. Spaghetti plots of the minimum MSLP (mb) through the forecast period for case 1 ensemble simulations with event lead times of 1-4 days. The red line is the mean of all ensemble members, and the black dashed line marks the verification time of cyclogenesis.
In contrast, for the ensemble simulations initialized 3 and 4 days before cyclogenesis, the spread in the ensemble members and the deviation from the mean is significantly greater. There are several members which do not indicate the deepening pressure associated with Genoa cyclogenesis. The ensemble spread increases rapidly near the cyclogenesis time for both the 3 and 4-day lead times. The ensemble means for both 3-day and 4-day lead times completely miss cyclogenesis. In both of these ensembles, the minimum MSLP mean actually increases slightly at the time of the Genoa low cyclogenesis. Similar characteristics are also reflected in the ensemble spaghetti plots of case 2. Unfortunately, many forecasters are inclined to look at the ensemble
22 mean and consider it a consensus. Figure 11 shows the ensemble mean solution for the 3-day forecast for day 1. When compared to the control for case 1, the broad, weak low pressure system depicted in the ensemble mean fails to capture the size, structure, and strength of the Genoa low. For this Genoa low event, the EPS mean shows a broad low that never deepens past 1010mb. The wide range of solutions that the ensemble members uniquely indicated (as seen in Figure 4 and Figure 5) are the reason why the ensemble mean is not a reliable solution. There are members which deepen the low to 1004mb while others do not drop the pressure below 1018mb. The mean has the added deficiency of potentially being an average of a bifurcated set of solutions, and therefore it may not be a solution of either regime represented by the EPS. The spatial scale of the event is also missed by the ensemble mean. The Genoa low is characterized as a mesoscale cyclone, but the mean misses out on this small structure with a small closed circulation near the pressure center.
Figure 11. Maps of the time evolution of the ensemble mean MSLP (mb) for the 3 day forecast in case 1 with time progressing from the 69-hour forecast time to the 81 hour forecast time from left to right.
A key factor in the assessment of the Genoa low predictability is determining whether the 4-day desired lead time reaches into a forecast range (as described in chapter 1) in which the unique numerical solutions lose the identity of the initial conditions. At such point, the numerical solutions become statistically indistinguishable from randomly selected observations of the MSLP field at any given time. The ensemble spread for the 4 day lead time ensemble is close to
23 the point of this deterministic limit in predicting the MSLP field in the target area associated with the Genoa low. Forecast hours beyond 100 hours show a generally evenly distributed range of solutions ranging from 1002mb to 1020mb. This is a broad range of MSLP that contains little information above climatology.
12 12
10 10
8 8
6 6
4 4
2 2
0 0
4 4
3.5 3.5
3 3
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
Figure 12. Shaded is the ensemble variance in MSLP (mb) for case 1 at the cyclogenesis time for each ensemble forecast lead time. Note: The scale for the 1 and 2-day lead times ranges from 0 to 4, and the scale for the 3 and 4 day lead times ranges from 0 to 12. Values below .5 mb are white.
24 Although the minimum MSLP fields at 4-day lead time do not indicate a pronounced signal associated with Genoa low formation, maps of ensemble variance provide a stronger indication to the forecaster that Genoa low formation is possible. Figure 12 shows maps of the ensemble variance of the MSLP for each set of ensemble runs initialized 1-4 days before Genoa low formation. Note that the scale for the 1-day and 2-day lead time ranges from 0 to 4 mb but the scale for the 3 and 4 day lead time ranges from 0 to 12 mb. From these maps, it is evident that the predictability limitations of the Genoa low expressed by the range of solutions in figure 1 and the standard deviation of the MSLP field seen in figure 2 is associated specifically with the Genoa low. For lead times up to 3 days, ensemble variance in the Mediterranean and the surrounding area is small outside of the Genoa low. It can be concluded that the variability among the ensemble solutions out to the 3-day lead time is not simply associated with the synoptic-scale low, but is directly related to a potential Genoa low formation. At 4-day lead time, there is a large anomaly in the variance in the target region of the Genoa low. The anomalous variance in the Genoa low region indicates that the EPS shows some potential for indicating a Genoa low out to the 4-day range, even if the uncertainty remains large. The predictive nature of the Genoa low in the EPS beyond that point rapidly decreases with the increase of the variability of the numerical solutions.
3.2. GENOA LOW FORMATION AND THE MISTRAL WIND
It has thus far been established that a Genoa low is a highly uncertain but weakly predictable event at a 4-day lead time. The following analysis will focus on whether Genoa cyclogenesis may be linked to other features in the ambient environment and whether such features may serve to limit or enhance the predictability of cyclogenesis. Specifically, this analysis will focus on the relationship between the Genoa low and the mistral wind.
Figure 13, presents the evolution of the ensemble standard deviation of both the Genoa low minimum MSLP and the maximum mistral wind speed. The standard deviation of the minimum MSLP field remains near 1mb leading up to cyclogenesis time for the 1-day and 2-day lead times. However, for the 3-day and 4-day lead times, the ensemble standard deviation increases to near 3mb at cyclogenesis time, and the standard deviation increases afterward. It can be seen that cyclogenesis time for the 1-day and 2-day lead times is when the standard deviation is maximized. The evolution of the standard deviation for each event lead time demonstrates a
25 correspondence between the predictability of the Genoa low and the mistral wind intensity. The standard deviation of the maximum mistral winds sharply increases near the time of the event for each of the ensemble lead times, just as did the minimum MSLP standard deviation. This correspondence revealed by the evolution of the standard deviation suggests that the mistral winds and the Genoa low share a similar predictability limit. It also suggests the possibility for a physical link between the mistral winds and the Genoa low, as will be explored further below.
Ensemble Standard Day Lead Deviation Ti e in Min MSLP (mb) (top) Day and Lead Max Ti e 10m Wind (kt)
7 Day Lead Ti e 7 Day Lead Ti e
MSLP MSLP
Fore ast Hour Fore ast Hour Day Lead Ti e Day Lead Ti e 9 Day Lead Ti e 9 Day Lead Ti e 7 Mistral Wi d 7 Mistral Wi d kts kts
Fore ast Hour Fore ast Hour 7 7 Day Lead Ti e Day Lead Ti e MSLP MSLP
Fore ast Hour Fore ast Hour Day Lead Ti e Day Lead Ti e 9 Day Lead Ti e 9 Day Lead Ti e
7 Mistral Wi d 7 Mistral Wi d
kts kts
Fore ast Hour Fore ast Hour Figure 13 Ensemble standard deviation in minimum MSLP (mb) (top) and 10m maximum mistral wind (bot) for simulations initialized 1-4 days prior to Genoa low development. The black dashed line marks the time of cyclogenesis.
26 Figure 14 shows the MSLP and 10m wind speed for all ensemble members initialized 3 days prior to the Genoa low event at the time of the beginning stages of the Genoa low event. The postage stamps in Figure 9 demonstrated a clear association between the strength of the mistral winds and the Genoa low. At a 3-day lead time to the event, there are several successes and failures available for comparison. The subjective analysis of the two plots in Figure 9 indicate a negative correlation between the strength of the mistral winds and the MSLP of the Genoa low. The ensemble members which produced a deeper low also feature stronger mistral winds. Conversely, the ensemble members which did not generate a Genoa low or a weaker low pressure center feature weak mistral winds. For example, ensemble member 10 rapidly generated a low of 1004mb. The mistral winds for that case at the same time were nearly 40kts. Ensemble member 9 did not produce a Genoa low. The lowest MSLP value was 1014mb over a broad area in the Mediterranean, and the mistral winds were slight across a relatively small region of northerly winds at 20kts. This relationship occurs in both cases across all event lead times, so it is likely not an artifact of the EPS, but rather a real feature of the Genoa low.
Figure 15 presents spaghetti plots for the 3 and 4-day event leads time from both cases with the minimum MSLP on the top and the maximum mistral winds on the bottom. The curves highlighted in green are the ensemble members which produced the deepest Genoa lows, and the red curves are the ensemble members which did not produce a Genoa low. In order to clarify the link between the mistral winds and the Genoa low found in the subjective analysis of Figure 9, the same ensemble members identified for either the deepest pressure or the highest pressure in the target area were also highlighted in the maximum mistral wind spaghetti plots. The results in Figure 10 are consistent with the side-by-side postage stamp comparison presented in Figure 9. The green curves in the mistral wind spaghetti charts are all characterized by a sharp increase in the wind speeds specifically in the 12 hours previous to the marked cyclogenesis time. For both of the event lead times, the mistral wind speed increases from 10kts to 40kts during this critical time leading to cyclogenesis. The associated MSLP in these members decreases from 1008mb to 1000mb or below in the 12 hours following the marked cyclogenesis time. The same relationship is seen for case 1. Conversely, the red curves in some cases have a rise in MSLP after the cyclogenesis time. The same ensemble members highlighted in the mistral wind spaghetti plots
27
Figure 14. Maps of MSLP (mb) (left) and 10 m wind speed (kts) (right) for case 1 for all ensemble members. This ensemble run was initialized 3 days prior to the event. Time is fixed at the cyclogenesis time.
28
Figure 15. Ensemble spaghetti plots of minimum MSLP (mb) (top) and maximum 10m mistral winds (kts) (bot) for case 1 (top) and case 2 (bot) for lead times at 3 days (left) and 4 days (right). The solid black line is the ensemble mean for each EPS run, and the vertical dashed black line marks the start of cyclogenesis. The green lines are the ensemble members chosen because they produced deep Genoa lows. The red lines are ensemble members chosen because they did not generate a Genoa low. The same 4 ensemble members in each case are highlighted in the same colors in the spaghetti plots of the maximum mistral wind field.
29 characterized by a weaker and more flat trend. The ensemble members which do not simulate a Genoa low also lack the strong mistral winds. Peak intensity of the MSLP and the mistral winds occur together in the ensemble simulations just as the weakest MSLP and mistral wind features occur in the same ensemble simulations. The negative correlation between the minimum MSLP in the Genoa low and the maximum winds of the mistral winds within the ensemble members is quantified in figure 11, and all of the members between the 2 extremes were included. Correlations were generated across all ensemble members for both cases. Figure 11 shows 2 examples of the negative correlation that exists between the minimum MSLP and the maximum mistral winds at the time of cyclogenesis on the left for the 3 and 4-day event lead time ensemble runs. The negative correlation is explicit in these plots. The correlation coefficients are -0.85 and -0.76 for case 1 and case 2 respectively. These plots confirm the negative correlation that links the strength of the mistral winds to the strength of the Genoa low. An association between the mistral winds and the Genoa low may not be surprising at first glance. After all, a deeper low implies a larger pressure gradient and stronger winds. Intriguingly, the time-lagged correlation remains large and negative when the mistral winds precede the Genoa low. In other words, it appears as though the mistral wind intensifies first, followed by a deepening of the Genoa low second. Such a behavior is evident in the time sequences of MSLP and wind speed. This results section concludes by proposing a potential mechanism by which mistral winds play a role in the deepening of the Genoa low and, consequently, play a controlling role in the predictability of cyclogenesis. In addition to being of theoretical interest, a meteorologist responsible for forecasting the Genoa low may exploit such a link in predicting cyclogenesis. If the MSLP field itself proves difficult to forecast but the mistral wind is consistent across multiple forecasting tools, then the forecaster can lean on the negative correlation between the two features to weigh the probability of the Genoa low forming. The right side of figure 16 is the same correlation plot as the left side, but the time of the mistral wind is lagged. The results of the lagged correlations show that for case 1, there was a relevant negative correlation of the Genoa low cyclogenesis to the mistral winds out to 6 hours prior to cyclogenesis time. For case 2, there was a correlation of Genoa cyclogenesis to the mistral winds out to 12 hours prior to the cyclogenesis time. There are a great number of factors in play during a scenario in which a Genoa low forms, so this research cannot conclusively prove that there is a
30 significant causal relationship of the mistral winds to the genesis of a Genoa low, but the correlation suggests that the mistral winds could serve as a predictive precursor to the start of Genoa low. This is potentially valuable to a forecaster who may be able to make a forecast decision based on the observed behavior of the mistral winds.
Figure 16. Correlation of minimum MSLP and maximum mistral winds at a 3 day event lead time for case 1 (top) and at a 4 day event lead time for case 2 (bot). The correlation plots on the left side are at the time of cyclogenesis, and the plots on the right are correlations between the minimum MSLP and the time-lagged maximum wind speed. The time lad is 6 hours for case 1 and 12 hours for case 2. The line is the least squares regression line, and the correlation coefficient is labeled on each plot.
31 If the mistral winds have a causal relationship with the formation of the Genoa low, then the predictability of the Genoa low may also be tied to the predictability of the mistral winds. Genoa lows consistently form along the eastern fringe of the mistral jet. Along this boundary, there is a sharp gradient in the wind speeds. The Gulf of Genoa is shielded from the strong northerly winds by the Alps while all of the cold air from the north funnels southward between the Alps and the Pyrenees. The area where the intense early stages of cyclogenesis occur near this wind speed gradient where the region is subject to a lot of cyclonic shear vorticity. Figure 17 presents maps of the MSLP contours with the shaded surface vorticity and wind vectors for ensemble member 9 in the 4-day event lead time ensemble run for case 2. Each row depicts a separate time sequence within the same ensemble member. A persistent region of positive vorticity in the region southwest of the Alps near where the Genoa low forms is present in all three sequences, and this localized region of positive vorticity appears to trigger a new Genoa low in each sequence via a vorticity seeding mechanism. All three of the lows that formed through this ensemble member formed on top of the cyclonic vorticity on the eastern fringe of the mistral winds. The positive vorticity can then be traced along with the eastward track of the low after it forms in this favored region. This vorticity seeding mechanism alone is neither a complete picture of the dynamics involved with the mistral winds nor is it the only potential factor in the cyclogenesis of a Genoa low. But it is an intriguing potential dynamical link between the mistral winds and the Genoa low, and it is yet another small scale feature involved in the process which may limit the predictability of the Genoa low in numerical modeling systems.
32
Figure 17. Time evolution of vorticity (shaded) and MSLP (contours) with 10m wind (vectors) from case 2, ensemble member 9. Time proceeds from left to right−4 at 3-hourly−1 segments. �
33 CHAPTER 4
CONCLUSIONS
This thesis has presented an analysis of Genoa low predictability and dynamics using a case-study approach. Two cases were analyzed using the COSMO-LEPS regional ensemble prediction system. Simulations were performed at a range of lead times, from 1-day to 4-days before the respective cyclogenesis events. Ensemble prediction systems provide a range of possible forecast outcomes given the uncertainty in initial conditions, boundary conditions, as well as model physics. As such, ensembles may be used to assess and analyze the predictability of a feature such as the Genoa low. The analysis demonstrated several key findings concerning the Genoa low: 1) The Genoa low is only weakly predictable at a lead time of 4 days. It was shown that only a small fraction of ensemble members (approximately 25%) indicated cyclogenesis at this lead time. 2) The possibility of low formation at longer lead times is most effectively visualized using maps of ensemble variance. Traditional postage-stamp plots and minimum MSLP plots contained too much noise and variability to permit a bench forecaster to extract a signal indicating possible low formation. 3) The formation of the Genoa low is associated with strong mistral winds. All ensemble simulations that were successful in identifying cyclogenesis had also produced strong mistral winds. The strength of the mistral winds was strongly anti-correlated to the minimum MSLP. This linkage implies a potential dynamical connection between the two features. 4) The mistral jet may exert an organizing influence on the Genoa low via a vorticity seeding mechanism. It was shown that the mistral jet amplifies several hours prior to cyclogenesis. The amplification is associated with mesoscale vorticity generation on the eastern periphery of the jet. These vorticity centers were subsequently shed into the region where cyclogenesis occurred. The current expectation for the USAF weather forecasters who are responsible for predicting the genesis of a Genoa low and its impacts on the DOD locations in the region is to be able to provide a 4-day lead time for the impacts of a Genoa low. This research has shown that while the regional EPS has some predictive skill out to the 4-day range, the Genoa low regime is subject to predictability limitations which affect a numerical prediction system’s ability to resolve the evolution of the Genoa low. Specifically, the EPS which was designed to be able to
34 resolve the spatial and time scales involved with the Genoa low had only a 25% and 35% success rate for case 1 and case 2 respectively at a 4-day lead time. The success rate only improves to 43% and 56% respectively at the 3-day lead time. The processes which occur at the mesoscale and smaller likely have a profound effect on the genesis of the Genoa low, and the numerical modeling systems are limited in their ability to accurately resolve their interactions and evolutions, especially at longer forecast ranges. The 21st OWS has identified the Genoa low as a forecast challenge which forecasters have consistently failed to accurately predict. The expectation for a 4-day lead time is not realistic because of the predictive limits of the Genoa low, but there is an alternative. The results presented in this thesis low should motivate a change in the expectation from an absolute forecast of Genoa lows at 4-day lead time (i.e., yes or no) to a probabilistic-based forecast, and this can be done through the use of an EPS. Being able to communicate the uncertainties of the forecast is key to properly informing the end users, the USDOD and ally force mission planners. A single deterministic numerical model will not be able to provide any information about the sensitivity of the forecast to small errors in the initial conditions and in the physics of the model. In this way, the use of an EPS can be a far more useful forecasting tool, especially at the 3-day and 4-day forecast lead times. A difficult obstacle to overcome regarding DOD operations are the stringent “go/no-go” criteria for mission execution. Almost every military commander demands that personnel adhere to clear cut checklists which have absolute criteria. Such absolutes are difficult to apply to weather forecasts. Is it going to rain at a specific time? Will there be a cloud ceiling a certain level? Will the wind speed reach a certain threshold? Will there be lightning over a certain location? Weather forecasters are expected to provide “go/no-go” answers for questions tailored to the specific mission such as these. It is critically important for the mission planners to have the best information possible when making decisions that put lives, property, and information on the line. The expectation for weather forecasters is not appropriate given the uncertainties of the forecast, and the result is that certain critical conditions are missed when forecasters are forced to check a “go/no-go” box. In these cases, the criteria intended to ensure an ideal scenario for mission execution actually cause the mission to suffer. In contrast, if weather forecasters are able to better communicate the uncertainties involved with a particular weather forecast, then it would be up to the mission planners to weigh the weather-related risks when making a decision. This is a far more realistic and effective means of mission planning.
35 The potential of an EPS to be used as an effective forecasting tool is limited by the extent to which the forecaster exploits the many advantages of the EPS. Ensemble forecasting is a relatively new tool, and USAF weather has yet to develop a usefully packaged ensemble suite with which to easily exploit all of the advanced tools made possible by an EPS. Currently, EPSs are being used for the purpose of a first glance forecast product consisting of maps of the ensemble mean for the relevant weather variables for the area of responsibility. This method is designed to draw the attention of the forecaster towards the main forecast challenges within their region. From that point forward, the forecaster is trained to address these forecast challenges by investigating them further using relevant forecast tools other than the EPS. The results of this experiment showed that the ensemble mean is not a proper representation of the collection of ensemble members. The common misconception is that the ensemble mean equates to an ensemble consensus which represents the most likely numerical solution, but for the case of the Genoa lows studied in this experiment, the variance in the ensemble members was too large at the 3-day and 4-day lead times for the mean to be representative of the eventual outcome. The mesoscale features of the Genoa low are masked by mean of the wide range of unique numerical solutions. This thesis proposes that forecasters could find more value from an improved suite of EPS tools. Typically, the forecasters responsible for the region where Genoa lows develop are also responsible for maintaining the terminal aerodrome forecasts for a 3 to 5 DOD locations as well as performing resource protections through the dissemination of watches, warnings, and advisories for several other DOD locations. Understandably, the forecaster cannot afford to spend the time sorting through each ensemble member to assembling a forecast. Therefore, the USAF EPS product suite needs to be developed further so that the forecasters can easily take advantage of its utility. Ideally, the forecaster should be able to easily view the spaghetti plots of the ensemble members, the maps of the variance in the ensemble, and maps of the ensemble members on both sides of the Genoa low intensity spectrum in addition to the ensemble means. It is entirely possible to develop a Genoa low-specific EPS page which could feature all of these ensemble statistic products on one page. This way, forecasters will be able to assess the behavior of the statistics of the ensemble in a way that will save them time and better prepare them to communicate the potential weather impacts to the mission planners at their locations. For example, the forecast process could start with the identification of synoptic scale, upper level
36 features typical of Genoa low formation which are resolved well by global numerical models, such as a 500 mb trough in the geopotential height over western Europe. They could then apply the ensemble products to gage the probability of a Genoa low forming in the Mediterranean Sea. The final forecast product consisting of language which transforms the variance of unique ensemble solutions into a set of probable scenarios including an estimation of the worst case will provide the mission planners with more accurate assessment potential weather impacts in the region. Being able to communicate a 4-day forecast in terms of probability rather than binary absolutes would be beneficial both the forecaster and the end user, and that is made possible by an EPS. In addition to illustrating the usefulness of the EPS when diagnosing the predictability of the Genoa low and the potential operational impact that a more complete use of the ensemble statistics could create, this experiment also showed a correspondence between the strength of the mistral winds and the Genoa low. The negative correlation between the two features is clear. The ensemble members which simulated deeper Genoa lows featured stronger mistral winds than the ensemble member which did not. More importantly, there is a negative correlation between the pressure of the Genoa low and the time-lagged mistral wind speeds. This suggests a causal effect of the mistral winds on the Genoa low. Previous authors mentioned in the introduction have shown the importance of the shape of the Alps and its role on the generation of the Genoa low citing how the arced shape of the western edge of the Alps causes the winds to naturally turn cyclonically in the Genoa low target region. Previous authors have also attributed the formation of the Genoa low to the enhanced baroclinic zone created by funneled cold air advection from the north over the Mediterranean to the west of the Genoa low formation zone. Still others have analyzed the convective feedback during the early stages of development which act to enhance low level convergence and organize the low pressure circulation. However, not much has been done to analyze potential that the vorticity along the periphery of the mistral winds affects the genesis of a Genoa low through vortex shedding along the mistral wind boundary. Figure 17 illustrates the low level vorticity associated with the mistral wind which demonstrated an influence on cyclogenesis on a fast time scale and relatively small spatial scale. This feature needs to be explored further because it bears the implication that the Genoa low is inheriting predictive limitations from rapid, highly variable features of the mistral winds.
37 REFERENCES
Archiv der UKMO-Bracknell-Bodenanalysen. Wetter : Wetterzentrale. Web. 20 Mar. 2017
Bauer, Peter, Alan Thorpe, and Gilbert Brunet. The Quiet Revolution of Numerical Weather Prediction. Nature 525.7567 (2015): 47-55.
Buzzi, A. and Tibaldi, S. (1978), Cyclogenesis in the lee of the Alps: A case study. Q.J.R. Meteorol. Soc., 104: 271–287.
COSMO. Operations for the COSMO LEPS., August 2008. Web. 1 Mar. 2017.
Central Europe Hit by Rains, Floods and Landslides: AIR Analysis. Insurance Journal. N.p., 04 June 2013. Web. 1 Mar. 2017.
Davies, Richard. "One Dead After Heavy Rain Causes Floods and Landslides in Italy." FloodList., 26 Feb. 2015. Web. 1 Mar. 2017.
Egger, J., 1988: Alpine Lee Cyclogenesis—Verification of theories. J. Atmos. Sci., 45, 2187– 2203.
Englebretson, Ronald E. (1989), Severe Weather Guide – Mediterranean Ports – 27. Genoa. Naval Environmental Prediction Research Facility, Monterey, CA
Fiori, E., Comellas, A., Molini, L., Rebora, N., Siccardi, F., Gochis, D.J., Tanelli, S., Parodi, A., Analysis and Hindcast Simulations of an Extreme Rainfall Event In the Mediterranean Area: The Genoa 2011 Case, Atmospheric Research, Volume 138, 1 March 2014, 13-29.
Lorenz, E., 1963: Deterministic Nonperiodic Flow. J. Atmos. Sci., 20, 130–141.
Marsigli C., Montani A., Nerozzi F., Paccagnella T., and Tibaldi S., COSMO LEPS Limited Area Ensemble Prediction System. ECMWF. 10 March 2004. Web. 1 Mar. 2017.
Marsigli, C., Montani A., and Paccagnella T., Perturbation of Initial and Boundary Conditions for a Limited area Ensemble: Multi model Versus Single Model Approach." Quarterly Journal of the Royal Meteorological Society 140.678 (2014): 197-208. ‐ ‐ ‐ McTaggart-Cowan, R., T. Galarneau, L. Bosart, and J. Milbrandt, 2010: Development and Tropical Transition of an Alpine Lee Cyclone. Part I: Case Analysis and Evaluation of Numerical Guidance. Mon. Wea. Rev., 138, 2281–2307.
Trigo, I., G. Bigg, and T. Davies, 2002: Climatology of Cyclogenesis Mechanisms in the Mediterranean. Mon. Wea. Rev., 130, 549–569.
38 Wynn, Nigel. "Tour de France stage to Mont Ventoux is Shortened Due to High Winds. Cycling Weekly. N.p., 18 Oct. 2016. Web. 1 Mar. 2017.
39 BIOGRAPHICAL SKETCH
Like many meteorologists, my interest in weather began at an early age. I grew up in Swansea, IL where I was able to witness everything from tornado outbreaks to heavy snow storms. My passion for weather was forged by a weather unit in the 4th grade for which I prepared a weather forecast presentation that went above and beyond the expectations. I was blessed with the opportunity to pursue an education in meteorology at Saint Louis University. I graduated with a BS in Meteorology in May 2012. Prior to graduation, I was selected by the United States Air Force to attend Officer Training School (OTS). Later that year, I attended OTS at Maxwell AFB in Montgomery, AL and commissioned as a weather officer in February 2013. My first assignment was at the 25th Operational Weather Squadron at Davis Monthan AFB in Tucson AZ where I was named the 2014 Air Force Weather Agency’s Company Grade Weather Officer of the Year. In 2015, I was selected to pursue a MS in meteorology through the Air Force Institute of Technology civilian institute program, and I started as a graduate student at Florida State University in August 2015. After graduation in May 2017, I will be assigned to Air Force Technical Applications Center at Patrick AFB in Cape Canaveral, FL. My amazing family, my wife, Jenny and my daughter, Olivia (also my dog, Yadi) have motivated and supported me every step of the way. Every move brings an exciting new adventure for us. We are looking forward to moving to Cape Canaveral because we will be able to take advantage of the year-round warm weather to stay active outdoors, especially at the beach. Plus, we will be right down the road from our favorite baseball team’s spring training facility in Jupiter, FL. It is rare that we miss a Cardinals game, no matter where we are.
40