- Email: [email protected]
Separation and classification of extreme wind events from anemometric records
J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Contents lists available at ScienceDirect
Journal of Wind Engineering and Industrial Aerodynamics journal homepage: www.elsevier.com/locate/jweia
Separation and classification of extreme wind events from anemometric records Patrizia De Gaetano, Maria Pia Repetto n, Teresa Repetto, Giovanni Solari Department of Civil, Chemical and Environmental Engineering, University of Genova, Italy
art ic l e i nf o
a b s t r a c t
Article history: Received 12 July 2013 Received in revised form 9 December 2013 Accepted 19 January 2014 Available online 14 February 2014
The separation and classification of intense wind events into homogeneous families is a key topic to study the wind-excited response of structures and to determine the distribution of extreme wind velocities and extreme wind-induced effects. This paper deals with the management of large sets of wind velocity data, in order to separate and classify independent extreme wind events through a semiautomated procedure involving a suitable mix of systematic quantitative controls and specific qualitative judgments. The proposed method is applied to an extensive dataset of continuous wind measurements provided by a wide and high quality monitoring network realized in the five main ports of the Northern Tyrrhenian Sea – namely Genova, Savona, La Spezia, Livorno and Bastia – in the framework of the European Project “Wind and Ports”. & 2014 Elsevier Ltd. All rights reserved.
Keywords: Dataset Depression Extreme wind events Gust factor Gust front Monitoring network Thunderstorm
1. Introduction In mixed wind climates, the separation and classification of intense wind events into homogeneous families is a crucial aspect to determine sound distributions of the extreme wind velocity and to carry out homogenous analyses of the wind-excited response of structures. With reference to statistical analysis, Thom (1967) first proposed to deal with the mixed populations of extra-tropical and tropical cyclones by two combined distributions, then showed that one-third of the yearly peak wind velocities in the United States occur during thunderstorms (Thom, 1968). Gomes and Vickery (1976) carried out a study of the extreme wind velocities in Australia, in which they separated thunderstorm from nonthunderstorm winds, determined the distributions of these two phenomena and derived a so-called mixed distribution. Also Riera et al. (1977) first represented thunderstorms and cyclones by different distributions, then they combined such distributions into a unique mixed distribution. Gomes and Vickery (1977/1978) extended the above formulations to mixed climates including several phenomena of different nature, namely extra-tropical pressure systems, thunderstorms, hurricanes and tornadoes; in addition to increasing the accuracy, they noted that the separate analysis of different phenomena enables the use of statistical
n
Corresponding author. Tel.: +39 0103532121.
0167-6105/$ - see front matter & 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jweia.2014.01.006
models especially suited to each phenomenon. Twisdale and Vickery (1992), Choi (1999), Choi and Hidayat (2002a) and Lombardo et al. (2009) separated thunderstorm from nonthunderstorm winds, determined the distributions of the maximum wind velocity of each phenomenon, and combined them into a mixed distribution. Choi and Tanurdjaja (2002) performed the same operation by separating large scale phenomena, identified with monsoons, from small scale phenomena, including squall lines and thunderstorms. Kasperski (2002) first noted that thunderstorms cannot be clearly separated from frontal depressions, since a third family of phenomena exist, called as gust fronts by author, with intermediate properties; therefore he applied mixed statistics to these three families of events. Cook et al. (2003) extended the theory of Gomes and Vickery (1977/1978) in a wider and more general analytical and operative framework. As far as concerns the wind-excited response of structures, the separation of the analyses related to different phenomena is mainly associated with their stationary or non-stationary and Gaussian or non-Gaussian properties. Choi and Hidayat (2002b) first discussed the different behaviour of single-degree-of-freedom (SDOF) systems subjected to non-stationary thunderstorms and stationary synoptic winds. Analogous studies were carried out later by Chen and Letchford (2004) and Chay and Albermani (2005). Kwon and Kareem (2009, 2013) proposed a method through which the dynamic response and the equivalent static force due to non-stationary thunderstorm winds are evaluated as a correction of the corresponding quantities due to stationary
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
synoptic phenomena. Solari et al. (2013) determined the dynamic response of SDOF systems to different families of wind events, focusing on the main differences among the behaviour of structures. Yeo (2011) and Lombardo (2012) separated thunderstorm from non-thunderstorm winds in order to carry out a dataset assisted design of structures in mixed climates. The methods for the separation and classification of intense wind events may be subdivided into two families mainly associated with the meteorological and wind engineering sectors. The first family of methods is aimed at examining specific phenomena of relevant interest through detailed inspections and reconstructions of the meteorological conditions that occurred in such events. For this reason, it relies on the surface measurement of the main meteorological parameters, radar and satellite images, soundings, and any other suitable data to clarify the atmospheric conditions. Charba (1974) studied an intense gust front that occurred in Oklahoma on 31st May 1969 by anemometers, thermometers, barometers and hygrometers mounted above or around a 444 m high transmission tower; he also used radar images. Goff (1976) carried out the analysis of the outflows of 20 thunderstorms that occurred in Oklahoma between 1971 and 1973; he used the meteorological instruments put on a 461 m high tower and the radar images acquired by the National Severe Storm Laboratory. Wakimoto (1982) examined the life cycle of thunderstorms in the framework of the Project NIMROD; the data is provided by 3 doppler radars and 27 surface stations collecting temperature, pressure and wind velocity; rawinsondes were also launched in serial ascents. Sherman (1987) provided a detailed description of a downburst that struck Bald Hills in Queensland on 5th November 1977; for this, he used the wind velocity records acquired at 4 levels of a transmission tower, surface measurements of temperature, pressure and humidity, images provided by the meteorological radar in the airport of Brisbane, 13.5 km far from the tower. Hjelmfelt (1988) examined the morphology of the downbursts acquired for the project JAWS, adopting doppler radar images, surface measurements of the wind velocity and soundings. Fujita (1990) described the results of the experiments carried out for the projects NIMROD, JAWS and MIST, respectively in 1978, 1982 and 1986; he used between 27 and 81 surface anemometers and 5 doppler radars. Gast and Schroeder (2003) studied the wind records caused by a super-cell that produced a rear-flank downdraft that passed over 7 monitored towers in Lubbock, Texas, on 4th June 2002; this phenomenon was also examined through doppler radars data and meteorological soundings. A detailed analysis of the same phenomenon was carried out by Holmes et al. (2008). Gunter and Schroeder (2013) recently started a research project aimed at furnishing a detailed description of thunderstorms based on surface meteorological measurements and images produced by couples of mobile doppler radars with high space and time resolution. The second family of methods consists in the systematic separation and classification of the measurements belonging to large wind datasets, with the aim of developing statistical analyses of the extreme wind velocities and their effects on structures. In the light of the huge amount of the examined data, this family of methods rejects the prohibitive idea of providing a detailed meteorological representation of all the wind phenomena that take place. Therefore, synthetic information is adopted, based on the available data, to make the separation and classification process as automatic as possible. Gomes and Vickery (1976) identified the peak values of the thunderstorm winds with the daily peak values of the wind velocity during thunderstorm days. Riera and Nanni (1989) separated thunderstorm from synoptic wind events depending on the duration of the intense event, the occurrence of thunder and lightning, rainfalls and abrupt temperature drops. Twisdale and Vickery (1992) assumed that the
133
maximum daily velocity has thunderstorm origins during thunderstorm days. Holmes (1999) noted that this assumption is often wrong. Choi (1999) and Choi and Hidayat (2002a) distinguished wind phenomena into thunderstorm and non-thunderstorm winds by defining thunderstorms as those events that occur with thunders and rainfalls; they noted also that the gust factor of thunderstorms is always much greater than the gust factor of nonthunderstorm winds. Choi and Tanurdjaja (2002) separated wind phenomena into large- and small-scale events through visual examinations of long wind velocity records. Kasperski (2002) subdivided the data belonging to frontal depressions, thunderstorms and intermediate events, or gust fronts, based on three parameters: the mean wind velocity, the peak wind velocity and the gust factor. Cook et al. (2003) classified as thunderstorm events the daily maximum wind velocities that occurred in the course of days tagged as “thunderstorms seen or thunderstorms heard”. Duranona et al. (2006) defined as non-synoptic those events that satisfy the following four criteria: the peak velocity is greater than 15 m/s; the ratio between the peak and the mean wind velocity is greater than 1.5; the velocity increases in less than 3 min; the velocity diminishes in less than 10 min. Lombardo et al. (2009) developed an automated procedure, applied to the US datasets acquired through the ASOS network, in order to separate thunderstorms from non-thunderstorm events; after observing that the manual separation of the data is prohibitive, they formulated a method framed into the following steps: (1) the peak wind velocities are extracted from each dataset; (2) the beginning and the end of each thunderstorm are registered; (3) the peak wind velocities that occur between the beginning and the end of a thunderstorm are associated to thunderstorms, all the other peaks being associated to non-thunderstorm events; (4) two datasets including independent thunderstorm and nonthunderstorm events are created, by defining as independent thunderstorm events those separated by at least 4 h, as independent non-thunderstorm events those separated by at least 4 days. Rowcroft (2011) extracted thunderstorm events from measurements carried out on monitored Australian towers for over 20 years; first, he selected the peak wind velocities greater than 40 m/s; then, he defined as thunderstorms those events that satisfy the following four conditions: (1) they last between 5 and 30 min; (2) the temperature has a drop of 1.5 1C or more; (3) there is an increase in wind speed at more than one height of the towers; (4) the wind speed differential is greater than 10 m/s. This paper belongs to this second family of methods, i.e. those methods that systematically separate and classify the anemometric records of large datasets without providing any specific meteorological survey of the examined data. It is also part of a wide research activity carried out in the framework of the European Project “Wind and Ports”, aimed at investigating extreme wind events in port areas (Solari et al., 2012). Section 2 describes the monitoring network that has been realized in the main ports of the High Tyrrhenian Sea for the project “Wind and Ports”; thanks to this network the raw wind velocity data is continuously acquired with a sampling rate of 10 Hz; this data is processed in order to extract the main statistical parameters referred to subsequent 10-min periods; such parameters are ordered into a dataset that represents the starting point for the separation and classification procedure described in the following sections. Section 3 examines the above wind velocity records and the related statistical parameters, classifying them into three families corresponding to stationary and Gaussian, nonstationary and non-Gaussian, and stationary and non-Gaussian events associated, respectively, with extra-tropical depressions, thunderstorms and intermediate phenomena called gust fronts by Kasperski (2002). Coherently with this classification, Section 4 illustrates the properties of the sub-datasets into which the wind
134
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Fig. 1. Anemometric stations installed in the port areas involved in the project “Wind and Ports”.
velocity records and the related statistical parameters are separated. Section 5 describes the algorithm used to carry out such separation; it involves a suitable mix of systematic quantitative controls of synthetic statistical parameters and specific qualitative judgments based on visual inspections of continuous records. Section 6 illustrates the criteria by means of which independent extreme wind velocities are extracted for depressions, thunderstorms and intermediate events, respectively. Section 7 summarizes the main conclusions and points out some perspectives.
Table 1 Main properties of the anemometers of the monitoring network of the project “Wind and Ports”. Port
Anemometer no.
h (m)
Type
Sampling rate (Hz)
Savona
0 1 2 3 4 5 1 2 1 2 3 4 1 2 3 4 5 1 2 3 4 5
84 33 18 27 32 42 61.4 13.3 15.5 13 10 11 20 20 20 20 75 10 10 13 10 10
Tri-axial
10
Bi-axial
10
Bi-axial
10
Tri-axial
10
Bi-axial
2
Genova
2. Monitoring network and main datasets La Spezia
Fig. 1 shows the anemometric monitoring network realized for the project “Wind and Ports – The forecast of the wind for the safety and the management of port areas” (Solari et al., 2012). The project, financed by the European Territorial Cooperation Objective Crossborder program “Italy–France Maritime 2007–2013” and carried out in the period 2010–2012, involved the port authorities of the five main ports in the High Tyrrhenian Sea, namely Genova, La Spezia, Livorno, Savona (Italy) and Bastia (France). The Department of Civil, Chemical and Environmental Engineering (DICCA) of the University of Genova was the scientific actuator. The network is constituted by 22 ultrasonic anemometers (circles) whose most important properties are listed in Table 1; h is the height of the anemometers above ground. Nine new anemometers co-financed by the Port Authority of Genova have been recently installed in the Port of Genova (squares in Fig. 1). In mid-2013, the European Territorial Cooperation Objective Cross-border program “Italy–France Maritime 2007–2013” has financed a new project, “Wind, Ports and Sea – The monitoring and forecast of weather and sea conditions for safe access to the port areas” – which will be developed by the same partners in the period 2013–2015. In the framework of this new project, the monitoring network will be enhanced by the addition of 6 more anemometers, a set of thermometers, barometers and hygrometers and, above all, 3 lidars installed in the Port of Genova, Savona and Livorno. The position of the anemometric instruments has been chosen in order to cover homogeneously the port areas and to register undisturbed wind velocity histories. The instruments are mounted
Livorno
Bastia
on high rise towers or at the top of buildings, at least at 10 m height above ground level, with particular attention to avoid local effects contaminating the measures. Wind measurements are collected with a precision of 0.01 m/s and 11 for intensity and direction, respectively. A set of local servers, placed in each port authority headquarter, receives the measures acquired by the anemometers in their own port area, and elaborates the basic statistics on 10-min periods, namely the mean and peak wind velocities and the mean wind direction. Each server automatically sends this information to the central server located in DICCA. Two files are sent every 10 min containing, for each anemometer, the raw data and the statistical values of the previous 10-min periods. The operational centre of
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
135
Fig. 2. Stationary Gaussian depression recorded on 22th November 2011 by the anemometer 3 of the Port of Savona: (a) 1 h wind speed time-history; (b) distribution of wind speed: histogram and Gaussian density function obtained by data parameters; (c) wind direction time-history; (d) polar histogram of the wind direction.
DICCA stores this data into a central dataset after having systematically checked and validated the data received. The real time transfer is crucial for the short term forecasting (Burlando et al., 2013). Starting from the set of the raw data, a new dataset is created, denoted by 1, which collects the main statistical parameters of the anemometric measures. One record for each subsequent 10-min period is stored, which gathers such parameters into three groups: (1) peak velocity Vp10 averaged on τ ¼ 1 s, mean velocity Vm10, mean direction αm10, gust factor G10 ¼Vp10/Vm10, turbulence intensity I10, skewness γ10 and kurtosis κ10 in T ¼10-min time intervals; (2) mean velocity Vm60, gust factor G60 ¼Vp10/Vm60, turbulence intensity I60, skewness γ60 and kurtosis κ60 in the 1-h interval centred around T; (3) maximum value in T of the mean wind velocity averaged over 1-min Vm1 and gust factor G1 ¼Vp10/Vm1. The dataset 1 represents the starting point for the separation and classification procedure described in the next sections.
3. Classification of intense wind events An examination of the huge amount of the collected data shows that intense wind events can be classified into three families characterized by different properties: (1) stationary (S) Gaussian (G) events, with relatively large mean velocities and small gust factors; such events usually
corresponds to neutral synoptic atmospheric conditions; they are referred to as extended pressure systems by Gomes and Vickery (1976, 1977/1978) and strong frontal depressions by Kasperski (2002, 2009); they are called herein extra-tropical depressions (D); (2) non-stationary (NS) non-Gaussian (NG) events, with large peak velocities and gust factors but relatively small mean velocities; coherently with current literature, they are referred herein to as thunderstorms (T); (3) stationary (S) non-Gaussian (NG) events, with relatively small mean velocities but large peaks and gust factors; they are referred to as intermediate events or gust fronts (F) by Kasperski (2002); in spite of an almost total lack of meteorological interpretations, it seems reasonable to assume that they are associated to strongly unstable atmospheric conditions. A record is referred herein to as stationary when it exhibits statistical regularity for a time period of 10 min. Coherently with the separation and classification procedure described in the next sections, all records reported in the present paper refer to the 1 h interval centred on the considered 10 min record, with the aim of providing a more general overview of the examined wind event. Figs. 2–4 show three typical events of the aforementioned type. For each event, the scheme (a) shows the time-history of the wind speed raw data, the mean value over 1-h periods (horizontal line), the mean values over 10-min subsequent periods (dotted line),
136
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Fig. 3. Non-stationary non-Gaussian thunderstorm recorded on 25th October 2011 by the anemometer 3 of the Port of La Spezia: (a) 1 h wind speed time-history; (b) distribution of wind speed: histogram and Gaussian density function obtained by data parameters; (c) wind direction time-history; (d) polar histogram of the wind direction.
and the 1-s peak (circle) (obviously smaller than the instantaneous peak); the scheme (b) shows the histogram of the wind speed measures, compared with the ideal Gaussian density function matching the mean wind speed and the standard deviation; the scheme (c) shows the time-history of the wind direction raw data, the mean value over 1-h periods (horizontal line) and the mean values over 10-min subsequent periods (dotted line); the scheme (d) shows the polar histogram of the wind direction. Fig. 2 shows a typical 1-h record of a S G event corresponding to a depression (D). The record is registered by the anemometer 3 of the Port of Savona (Fig. 1 and Table 1). The measured event is characterized by a relatively high mean wind velocity (Vm10 ¼ 14.61 m/s, Vm60 ¼15.03 m/s) and by a 1-s gust peak Vp10 ¼22.46 m/s; the turbulence intensity (I10 ¼0.18, I60 ¼ 0.19) and the gust factor (G10 ¼ 1.53, G60 ¼ 1.49) are typical of neutral atmospheric conditions. The skewness (γ10 ¼0.37, γ60 ¼0.02) and the kurtosis (κ10 ¼ 3.42, κ60 ¼3.08) denote a rather typical Gaussian process, especially with reference to a 1-h period; the Gaussian density function fV(v) is close to the histogram of the recorded velocity v. The wind direction has stationary features. Fig. 3 shows a typical 1-h record of a NS NG event corresponding to a thunderstorm (T). The record is registered by the anemometer 3 of the Port of La Spezia (Fig. 1 and Table 1). The measured event is characterized by a relatively low mean velocity (Vm10 ¼9.09 m/s, Vm60 ¼ 7.33 m/s), a very high 1-s gust peak Vp10 ¼33.36 m/s and a very high gust factor (G10 ¼3.67, G60 ¼4.55). The skewness (γ10 ¼1.55, γ60 ¼1.20) and the kurtosis
(κ10 ¼4.47, κ60 ¼5.60) denote a typical non-Gaussian distribution. The time-history of the wind direction shows a sudden rotation of about 901 just in correspondence of the gust peak. The polar histogram of the wind direction points out this aspect through a bi-modal distribution. Fig. 4 shows a typical 1-h record of a S NG event corresponding to a gust front (F). The record is registered by the anemometer 3 of the Port of La Spezia (Fig. 1 and Table 1). The measured event is characterized by a low mean velocity (Vm10 ¼8.12 m/s, Vm60 ¼ 8.77 m/s) and by a relatively intense 1-s gust peak Vp10 ¼ 18.66 m/s; the turbulence intensity (I10 ¼ 0.38, I60 ¼ 0.31) and the gust factor (G10 ¼2.30, G60 ¼2.13) are much higher than those which are typical of neutral atmospheric conditions. The skewness (γ10 ¼0.42, γ60 ¼0.33) and the kurtosis (κ10 ¼2.60, κ60 ¼ 2.89) denote a moderately non-Gaussian distribution. Likewise the wind intensity, also the time-history of the wind direction is characterized by a relatively large variability, clearly pointed out by the spread in the polar histogram of the wind direction.
4. Separation flowchart into sub-datasets Fig. 5 shows the conceptual flowchart of the procedure by means of which the original set of the raw data and the related statistical parameters are transformed into more sub-datasets associated with the different wind events described in Section 3.
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
137
Fig. 4. Stationary non-Gaussian gust front recorded on 16th December 2011 by the anemometer 3 of the Port of La Spezia: (a) 1 h wind speed time-history; (b) distribution of wind speed: histogram and Gaussian density function obtained by data parameters; (c) wind direction time-history; (d) polar histogram of the wind direction.
α
γ
κ
Fig. 5. Conceptual flowchart of the procedure by means of which the original base of the raw data is transformed into more sub-datasets associated with different wind events.
(a) The series of the mean wind velocities Vm10 stored in the dataset 1 (Section 3) is processed by the method of storms proposed by Cook (1982). The mean wind velocities over subsequent 10 h periods is first calculated; such velocities are searched sequentially from the beginning, in order to find the first occurrence on which the wind speed drops below a given threshold Vn, defined as the start of a lull; between the start of two subsequent lulls there is a wind storm; the threshold Vn is chosen in such a way as to identify, for each anemometer, about 100 storms per year.
Table 2 Number of records in each dataset, with reference to the anemometer 3 of the Port of La Spezia and the anemometer 4 of the Port of Livorno. Dataset
Anemometer 3 Port of La Spezia
Anemometer 4 Port of Savona
1 2 3 4 5 6 7 8 9 10 11 12 13
105,264 695 244 24 17 450 17 228 200 35 12 32 15
131,472 3275 4559 71 83 3230 24 21 4506 94 21 17 81
(b) Starting from the dataset 1, two datasets denoted by 2 and 3 are created. The dataset 2 contains the records for which the peak wind velocities satisfy the condition Vp10 ZVnp10; the dataset 3 contains the records for which the mean wind velocities satisfy the condition Vm10 ZVnm10; in this case, Vnp10 ¼15 m/s and Vnm10 ¼10 m/s. (c) Starting from the datasets 2 and 3, two datasets denoted by 4 and 5 are created. They contain, respectively, the sub-set of the records associated with the maximum values of Vp10 and Vm10 for each storm, irrespective of any separation into families of homogeneous events.
138
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Fig. 6. (a) Roughness length z0 of the area of the Port of Livorno; reference turbulence intensity I0 at 20 m height above ground, when the wind blows from: (b) α ¼ 1251; (c) α ¼ 3051.
(d) Starting from the dataset 2, three datasets denoted by 6, 7 and 8 are extracted. The dataset 6 contains the sub-set of the records associated with all the Vp10 values corresponding to S G depressions (D). The dataset 7 contains the sub-set of the records associated with all the Vp10 values corresponding to NS NG thunderstorms (T). The dataset 8 contains the sub-set of the records associated with all the Vp10 values corresponding to S NG gust fronts (F). The classification of each event into one of the three categories D, T or F is carried out by means of the algorithm illustrated in Section 5. (e) Starting from the dataset 3, one dataset denoted by 9 is extracted. It contains the sub-set of the records associated with all the Vm10 values due to S G depressions (D) and corresponds, in terms of mean wind velocities, to the dataset 6 in terms of peak wind velocities. (f) Starting from the datasets 6, 7, 8 and 9, four datasets denoted by 10, 11, 12 and 13 are extracted. The datasets 10 and 13 contain, for each subsequent storm, the sub-set of the records associated with the maximum independent Vp10 and Vm10 values corresponding, respectively, to S G depressions (D). The datasets 11 and 12 contain the sub-set of the records associated with the maximum independent Vp10 and Vm10 values corresponding, respectively, to NS NG thunderstorms (T) and S NG gust fronts (F); the concept of independence is discussed in Section 6. Table 2 shows the number of records belonging to each of the above sub-datasets, with reference to the anemometer 3 of the Port of La Spezia and to the anemometer 4 of the Port of Livorno (Fig. 1 and Table 1). The smaller percentage of the records belonging to the datasets 2 and 3 in the Port of La Spezia with respect to the Port of Livorno depends on the less intense velocities measured, due to the lower height of the anemometer and to the major complexity of its surrounding.
5. Separation and classification algorithm Each anemometer is qualified via a card that reports its position and the main technical characteristics. This card also provides, at height h over ground and as a function of the direction αm10 of the oncoming wind, the turbulence intensity I010 and the three reference gust factors G010, G060 and G01; the apex 0 indicates that such values are evaluated numerically assuming, as it is classical for synoptic events, that intense wind velocities occur in neutral atmospheric conditions and are stationary Gaussian processes. Fig. 6 shows the turbulence intensity I010 in the Port of Livorno, at 10 m height above ground, when the wind blows from
αm10 ¼1251 (a) and 3051 (b). These results are obtained by the procedure formulated by ESDU (1993), in an advanced version developed at DICCA (Solari et al., 2012). Fig. 7 shows the diagrams of the reference gust factors G010, G060 and G01 as functions of the mean wind direction αm10 at the anemometers 1, 2 and 3 in the Port of La Spezia (Fig. 1 and Table 1). These quantities are evaluated through the method proposed by Solari (1993). It is apparent that their values decrease when the wind blows from the sea. The separation among depressions (D), thunderstorms (T) and gust fronts (F), indicated in Fig. 5 as the extraction of the datasets 6, 7 and 8 from the dataset 2, and of the dataset 9 from the dataset 3, is carried out through a semi-automated procedure shown by the flow-chart in Fig. 8. Such procedure is a suitable mix of systematic quantitative controls, based on the comparison between measured gust factors and their reference values, and qualitative judgments, based on the visual examination of the wind velocity records and their skewness and kurtosis values. It is apparent that this algorithm is efficient provided that the number of the qualitative controls is very limited. The procedure involves the following steps: (1) Events for which the following rule applies are classified as depressions (D): G60 G060
r 1:10
ð1Þ
They are strongly stationary and Gaussian not only over 10-min intervals, but also over 1-h intervals. Fig. 9a shows a typical record belonging to this family of events. (2) Events for which the following rule applies are classified as depressions (D): G60 G060
r 1:25 \
G10 G010
r 1:10
ð2Þ
They are strongly stationary and Gaussian over 10-min intervals, while they exhibit some variability over 1-h intervals. Fig. 9b shows a typical record belonging to this family of events. (3) Events for which the following rule applies are classified as thunderstorms (T) or gust fronts (F): G10 G010
4 1:25
ð3Þ
The classification of such events into the category of thunderstorms (T) (see for instance the record in Fig. 9c) or gust fronts (F) (Fig. 9d) is carried out through a qualitative control.
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
139
Fig. 7. Reference gust factors G010, G060 and G01 as functions of the mean wind direction at the anemometers 1, 2 and 3 in the Port of La Spezia.
D Y
Y
D, T, F
1
N
Y
2
Data set Quantitative control Expert judgment
3 N
N
N
4
N Y
T?
N
F?
T? N
Y
Y
F
Y
T
Fig. 8. Flow-chart of the semi-automated algorithm by means of which depressions (D), thunderstorms (T) and gust fronts (F) are separated.
(4) Events for which the following rule applies are classified as depressions (D): G60 G060
o 1:25 \
G1 4 0:80 G10
ð4Þ
Fig. 9e shows a typical record belonging to this family of events. Those events that do not satisfy Eq. (4) are classified as thunderstorms (T) (see for instance the record in Fig. 9f), gust fronts (F) (Fig. 9g) or depressions (D) (Fig. 9h) through a qualitative control. It is worth noting that the event depicted in Fig. 9h involves a sudden growth of the wind velocity after about 20 min from the start of the record; such growth is not classified by the algorithm as a thunderstorm since the corresponding peak wind velocity is lower than the threshold Vnp10 ¼15 m/s. Table 3 summarizes the main statistical parameters of the 8 events shown in Fig. 9. The qualitative controls carried out during the above procedure are based on the visual examination of the wind velocity records and on the interpretation of synthetic statistical parameters with reference to which it is not easy to formulate general quantitative rules. As an example, an indicator
for classifying an event not attributable to a depression (D) as a thunderstorm (T) or a gust front (F) is the ratio G10/G60: when it is less than 0.90, the event is usually a thunderstorm (T); when it is greater than 0.90, the event is usually a gust front (F). A skewness γ10 0 and a kurtosis κ10 3 are parameters typical of a Gaussian depression (D); different values of γ10 and κ10 indicate the occurrence of a non-Gaussian thunderstorm (T) or gust front (F). In the two cases of the data registered by the anemometer 3 of the Port of La Spezia and by the anemometer 4 of the Port of Livorno, the number of the required qualitative control is rather limited: 349 out of 105,264 (0.33%) and 86 out of 131,472 (0.06%), respectively. The higher number of qualitative controls in the Port of La Spezia is mainly associated with the major complexity of its territory. It is worth noting, finally, that there are some anomalous events associated with intense wind velocities, very rare, not clearly attributable to any of the three types of phenomena discussed above (D, T e F). Fig. 10 shows three examples of such phenomena. The first one is characterized by a sudden growth of the wind velocity (Fig. 10a), not followed by an analogous reduction, which occurs in correspondence of a sudden rotation of the wind direction (Fig. 10b). The second one involves an extremely rapid change of the wind velocity (Fig. 10c), not associated to particular changes of the wind direction (Fig. 10d). The third one exhibits a sudden reduction of the wind velocity (Fig. 10e), after a period in which the velocity was intense and stationary, associated with a sudden rotation of the wind direction (Fig. 10f).
6. Extraction of independent events Once depressions (D), thunderstorms (T) and gust fronts (F) have been separated into different sub-datasets by the criterion described in Section 5, it remains to extract, from each of them, a suitable series of independent extreme wind velocities. The maximum values of the depressions associated with different successive storms, both in terms of mean and peak
140
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Fig. 9. Some examples of different events examined by the separation algorithm during subsequent steps: (a) depression; (b) depression; (c) thunderstorm; (d) gust front; (e) depression; (f) thunderstorm; (g) gust front; (h) depression.
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Table 3 Main synthetic statistical parameters of the 8 events shown in Fig. 9. Event
Vp10 (m/s)
G10
G60
G1
γ10
γ60
κ10
κ60
a b c d e f g h
16.49 14.90 18.00 15.90 18.41 15.52 19.07 16.12
1.54 1.42 2.90 1.98 1.81 1.80 1.84 1.28
1.55 1.71 5.76 1.99 1.89 4.53 2.10 2.01
1.36 1.34 1.21 1.52 1.52 1.17 1.40 1.12
0.14 0.06 0.73 0.83 0.29 0.04 0.34 0.09
0.05 0.10 1.33 0.19 0.24 0.93 0.41 0.47
2.62 2.45 2.26 3.39 3.08 1.83 2.92 2.61
2.94 2.41 5.67 3.41 3.22 3.25 3.27 1.58
141
values, are dealt with as independent, provided that they are separated by at least 72 h. The maximum values of the thunderstorms are dealt with as independent provided that they are separated by at least 4 h. In a very preliminary way, not yet sustained by sound motivations, the maximum values of the gust fronts are dealt with as independent provided that they are separated by at least 24 h. It remains an open question, whether the maximum values associated with different events close together or even consequential should be treated as dependent or independent. Fig. 11 shows an example of this situation in which a depression whose
Fig. 10. Anomalous events recorded by the anemometer 4 of the Port of Livorno: wind velocity (a) and direction (b) on 16th October 2010; wind velocity (c) and direction (d) on 31st October 2010; wind velocity (e) and direction (f) on 16th December 2011.
142
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Fig. 11. Intense depression followed by an intense thunderstorm, recorded by the anemometer 1 of the Port of Livorno on 16th December 2011: (a) wind velocity; (b) wind direction.
maximum velocity is classified as an extreme is followed by a thunderstorm that represents, in this family of events, another extreme.
7. Conclusions and prospects This paper deals with the management of large wind velocity datasets, in order to separate and classify independent extreme wind events through a semi-automated procedure involving a suitable mix of systematic quantitative controls of synthetic statistical parameters and specific qualitative judgments based on visual inspections of continuous records. Two aspects are worth noting. First, though most of the papers published on the present topic tend to separate thunderstorms from synoptic non-thunderstorm events, this classification cannot be easily pursued. As Kasperski (2002) first pointed out, there is at least a third class of wind events, with intermediate properties, which greatly complicate the “binary” approach that prevails in the literature. However, the meteorological knowledge of these events is definitely poor and their understanding deserves further research. Second, in order to apply the classification criterion discussed in the present paper, several statistical parameters not commonly available should be used and recourse should be made to qualitative judgments based upon the visual inspection of continuous velocity records. This throws quite a few shadows on some extraction and classification criteria reported in literature, involving the use of only synthetic parameters. Since the separation of wind events in different classes is functional to carry out refined analyses, their meaning becomes somewhat questionable if it is not equally accurate the preliminary separation process. Both these aspects have profound impact on the dynamic wind-excited response of structures and on the mixed statistics of extreme wind velocities and extreme wind-induced effects (Solari, 2013). In mid-2013, the European Community approved a new project, “Wind, Ports and Sea”, carried out by the same partners already involved in the project “Wind and Ports” that provided the data for these analyses. Among other issues, this project is aimed at strengthening and extending the wind monitoring network described in Section 2. Taking into account the focal role of huge and high quality data to carry out studies such as those reported in the present paper, the authors rely on the possibility of improving the results of this research when the new data will be available.
References Burlando, M., De Gaetano, P., Pizzo, M., Repetto, M.P., Solari, G., Tizzi, M., 2013. Wind short-term forecast in port areas. In: Proceedings of the 6th European and African Conference on Wind Engineering. Cambridge, U.K. Charba, J., 1974. Application of gravity current model to analysis of squall-line gust front. Mon. Weather Rev. 102, 140–156. Chay, M., Albermani, F., 2005. Dynamic response of a SDOF system subjected to simulated downburst winds. In: Proceedings of the 6th Asia-Pacific Conference on Wind Engineering. Seoul, Korea. pp. 1562–1584. Chen, L., Letchford, C.W., 2004. Parametric study on the alongwind response of the CAARC building to downbursts in the time domain. J. Wind Eng. Ind. Aerodyn. 92, 703–724. Choi, E.C.C., 1999. Extreme wind characteristics over Singapore – an area in the equatorial belt. J. Wind Eng. Ind. Aerodyn. 83, 61–69. Choi, E.C.C., Hidayat, F.A., 2002a. Gust factors for thunderstorm and nonthunderstorm winds. J. Wind Eng. Ind. Aerodyn. 90, 1683–1696. Choi, E.C.C., Hidayat, F.A., 2002b. Dynamic response of structures to thunderstorm winds. Prog. Struct. Eng. Mater. 4, 408–416. Choi, E.C.C., Tanurdjaja, A., 2002. Extreme wind studies in Singapore. An area with mixed weather system. J. Wind Eng. Ind. Aerodyn. 90, 1611–1630. Cook, N.J., 1982. Towards better estimation of extreme winds. J. Wind Eng. Ind. Aerodyn. 9, 295–323. Cook, N.J., Harris, R.I., Whiting, R., 2003. Extreme wind speeds in mixed climates revisited. J. Wind Eng. Ind. Aerodyn. 91, 403–422. Duranona, V., Sterling, M., Baker, C.J., 2006. An analysis of extreme non-synoptic winds. J. Wind Eng. Ind. Aerodyn. 95, 1007–1027. Engineering Sciences Data Unit, 1993. Computer Program for Wind Speeds and Turbulence Properties: Flat or Hill Sites in Terrain with Roughness Changes, ESDU Item 92032. London, UK. Fujita, T.T., 1990. Downburst: meteorological features and wind field characteristics. J. Wind Eng. Ind. Aerodyn. 36, 75–86. Goff, R.G., 1976. Vertical structure of thunderstorm outflows. Mon. Weather Rev. 104, 1429–1440. Gast, K.D., Schroeder, J.L., 2003. Supercell rear-flank downdraft as sampled in the 2002 thunderstorm outflow experiment. In: Proceedings of the 11th International Conference on Wind Engineering. Lubbock, Texas. Gomes, L., Vickery, B.J., 1976. On thunderstorm wind gusts in Australia. Civil Eng. Trans. Inst. Eng. Aust. 18, 33–39. Gomes, L., Vickery, B.J., 1977/1978. Extreme wind speeds in mixed climates. J. Ind. Aerodyn. 2, 331–344. Gunter, W.S., Schroeder, J.L., 2013. High-resolution full-scale measurements of thunderstorm outflow winds. In: Proceedings of the 12th Americas Conference on Wind Engineering. Seattle, Washington. Hjelmfelt, M.R., 1988. Structure and life cycle of microburst outflows observed in Colorado. J. Appl. Meteorol. 27, 900–927. Holmes, J.D., 1999. Modelling of extreme thunderstorm winds for wind loading of structures and risk assessment. In: Larsen, A., Larose, G. (Eds.), Wind Engineering into the 21st Century. Balkema, Rotterdam Holmes, J.D., Hangan, H.M., Schroeder, J.L., Letchford, C.W., Orwig, K.D., 2008. A forensic study of the Lubbock-Reese downdraft of 2002. Wind Struct. 11 (2), 19–39. Kasperski, M., 2002. A new wind zone map of Germany. J. Wind Eng. Ind. Aerodyn. 90, 1271–1287. Kasperski, M., 2009. Specification of the design wind load-A critical review of code concepts. J.Wind Eng.Ind. Aerodyn. 97, 335–357. Kwon, D.K., Kareem, A., 2009. Gust-front factor: new framework for wind load effects on structures. J. Struct. Eng. ASCE 135 (6), 717–732.
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Kwon, D.K., Kareem, A., 2013. Generalized gust-front factor: a computational framework for wind load effects. Eng. Struct. 48, 635–644. Lombardo, F.T., 2012. Improved extreme wind speed estimation for wind engineering applications. J. Wind Eng. Ind. Aerodyn. 104–106, 278–284. Lombardo, F.T., Main, J.A., Simiu, E., 2009. Automated extraction and classification of thunderstorm and non-thunderstorm wind data for extreme-value analysis. J. Wind Eng. Ind. Aerodyn. 97, 120–131. Riera, J.D., Nanni, L.F., 1989. Pilot study of extreme wind velocities in a mixed climate considering wind orientation. J. Wind Eng. Ind. Aerodyn. 32, 11–20. Riera, J.D., Viollaz, A.J., Reimundin, J.C., 1977. Some recent results on probabilistic models of extreme wind speeds. J. Wind Eng. Ind. Aerodyn. 2, 271–287. Rowcroft, J., 2011. Vertical wind shear profiles in downburst events and the insufficiency of wind turbine design codes. In: Proceedings of the 13th International Conference on Wind Engineering. Amsterdam, The Netherlands. Sherman, D.J., 1987. The passage of a weak thunderstorm downburst over an instrumented tower. Mon. Weather Rev. 115, 1193–1205. Solari, G., 1993. Gust buffeting. I: peak wind velocity and equivalent pressure. J. Struct. Eng. ASCE 119, 365–382. Solari, G., 2013. Emerging issues and new scenarios for wind loading on structures. In: Proceedings of the 6th International Symposium on Wind Effects on
143
Buildings and Structures: Current State-of-the-Art in Wind Engineering and Outlook for the Future. Tokyo, Japan. pp. 153–176. Solari, G., De Gaetano, P., Repetto, M.P., 2013. Thunderstorm response spectrum, In: Proceedings of 12th Americas Conference on Wind Engineering. Seattle, Washington. Solari, G., Repetto, M.P., Burlando, M., De Gaetano, P., Pizzo, M., Tizzi, M., Parodi, M., 2012. The wind forecast for safety management of port areas. J. Wind Eng. Ind. Aerodyn. 104–106, 266–277. Thom, H.C.S., 1967. Toward a universal climatological extreme wind distribution. In: Proceedings of the 2nd International Conference on Wind Effects on Buildings and Structures. Ottawa, Canada. pp. 669–683. Thom, H.C.S., 1968. New distributions of extreme wind speeds in the United States. J. Struct. Div. ASCE, 94; , pp. 1787–1801. Twisdale, L.A., Vickery, P.J., 1992. Research on thunderstorm wind design parameters. J. Wind Eng. Ind. Aerodyn. 41, 545–556. Wakimoto, R.M., 1982. The life cycle of thunderstorm gust fronts as viewed with Doppler radar and rawinsonde data. Mon. Weather Rev. 110, 1060–1082. Yeo, D.H., 2011. Database assisted design for high-rise structures in mixed extreme wind climates. In: Proceedings of the 13th International Conference on Wind Engineering. Amsterdam, The Netherlands.
Contents lists available at ScienceDirect
Journal of Wind Engineering and Industrial Aerodynamics journal homepage: www.elsevier.com/locate/jweia
Separation and classification of extreme wind events from anemometric records Patrizia De Gaetano, Maria Pia Repetto n, Teresa Repetto, Giovanni Solari Department of Civil, Chemical and Environmental Engineering, University of Genova, Italy
art ic l e i nf o
a b s t r a c t
Article history: Received 12 July 2013 Received in revised form 9 December 2013 Accepted 19 January 2014 Available online 14 February 2014
The separation and classification of intense wind events into homogeneous families is a key topic to study the wind-excited response of structures and to determine the distribution of extreme wind velocities and extreme wind-induced effects. This paper deals with the management of large sets of wind velocity data, in order to separate and classify independent extreme wind events through a semiautomated procedure involving a suitable mix of systematic quantitative controls and specific qualitative judgments. The proposed method is applied to an extensive dataset of continuous wind measurements provided by a wide and high quality monitoring network realized in the five main ports of the Northern Tyrrhenian Sea – namely Genova, Savona, La Spezia, Livorno and Bastia – in the framework of the European Project “Wind and Ports”. & 2014 Elsevier Ltd. All rights reserved.
Keywords: Dataset Depression Extreme wind events Gust factor Gust front Monitoring network Thunderstorm
1. Introduction In mixed wind climates, the separation and classification of intense wind events into homogeneous families is a crucial aspect to determine sound distributions of the extreme wind velocity and to carry out homogenous analyses of the wind-excited response of structures. With reference to statistical analysis, Thom (1967) first proposed to deal with the mixed populations of extra-tropical and tropical cyclones by two combined distributions, then showed that one-third of the yearly peak wind velocities in the United States occur during thunderstorms (Thom, 1968). Gomes and Vickery (1976) carried out a study of the extreme wind velocities in Australia, in which they separated thunderstorm from nonthunderstorm winds, determined the distributions of these two phenomena and derived a so-called mixed distribution. Also Riera et al. (1977) first represented thunderstorms and cyclones by different distributions, then they combined such distributions into a unique mixed distribution. Gomes and Vickery (1977/1978) extended the above formulations to mixed climates including several phenomena of different nature, namely extra-tropical pressure systems, thunderstorms, hurricanes and tornadoes; in addition to increasing the accuracy, they noted that the separate analysis of different phenomena enables the use of statistical
n
Corresponding author. Tel.: +39 0103532121.
0167-6105/$ - see front matter & 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jweia.2014.01.006
models especially suited to each phenomenon. Twisdale and Vickery (1992), Choi (1999), Choi and Hidayat (2002a) and Lombardo et al. (2009) separated thunderstorm from nonthunderstorm winds, determined the distributions of the maximum wind velocity of each phenomenon, and combined them into a mixed distribution. Choi and Tanurdjaja (2002) performed the same operation by separating large scale phenomena, identified with monsoons, from small scale phenomena, including squall lines and thunderstorms. Kasperski (2002) first noted that thunderstorms cannot be clearly separated from frontal depressions, since a third family of phenomena exist, called as gust fronts by author, with intermediate properties; therefore he applied mixed statistics to these three families of events. Cook et al. (2003) extended the theory of Gomes and Vickery (1977/1978) in a wider and more general analytical and operative framework. As far as concerns the wind-excited response of structures, the separation of the analyses related to different phenomena is mainly associated with their stationary or non-stationary and Gaussian or non-Gaussian properties. Choi and Hidayat (2002b) first discussed the different behaviour of single-degree-of-freedom (SDOF) systems subjected to non-stationary thunderstorms and stationary synoptic winds. Analogous studies were carried out later by Chen and Letchford (2004) and Chay and Albermani (2005). Kwon and Kareem (2009, 2013) proposed a method through which the dynamic response and the equivalent static force due to non-stationary thunderstorm winds are evaluated as a correction of the corresponding quantities due to stationary
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
synoptic phenomena. Solari et al. (2013) determined the dynamic response of SDOF systems to different families of wind events, focusing on the main differences among the behaviour of structures. Yeo (2011) and Lombardo (2012) separated thunderstorm from non-thunderstorm winds in order to carry out a dataset assisted design of structures in mixed climates. The methods for the separation and classification of intense wind events may be subdivided into two families mainly associated with the meteorological and wind engineering sectors. The first family of methods is aimed at examining specific phenomena of relevant interest through detailed inspections and reconstructions of the meteorological conditions that occurred in such events. For this reason, it relies on the surface measurement of the main meteorological parameters, radar and satellite images, soundings, and any other suitable data to clarify the atmospheric conditions. Charba (1974) studied an intense gust front that occurred in Oklahoma on 31st May 1969 by anemometers, thermometers, barometers and hygrometers mounted above or around a 444 m high transmission tower; he also used radar images. Goff (1976) carried out the analysis of the outflows of 20 thunderstorms that occurred in Oklahoma between 1971 and 1973; he used the meteorological instruments put on a 461 m high tower and the radar images acquired by the National Severe Storm Laboratory. Wakimoto (1982) examined the life cycle of thunderstorms in the framework of the Project NIMROD; the data is provided by 3 doppler radars and 27 surface stations collecting temperature, pressure and wind velocity; rawinsondes were also launched in serial ascents. Sherman (1987) provided a detailed description of a downburst that struck Bald Hills in Queensland on 5th November 1977; for this, he used the wind velocity records acquired at 4 levels of a transmission tower, surface measurements of temperature, pressure and humidity, images provided by the meteorological radar in the airport of Brisbane, 13.5 km far from the tower. Hjelmfelt (1988) examined the morphology of the downbursts acquired for the project JAWS, adopting doppler radar images, surface measurements of the wind velocity and soundings. Fujita (1990) described the results of the experiments carried out for the projects NIMROD, JAWS and MIST, respectively in 1978, 1982 and 1986; he used between 27 and 81 surface anemometers and 5 doppler radars. Gast and Schroeder (2003) studied the wind records caused by a super-cell that produced a rear-flank downdraft that passed over 7 monitored towers in Lubbock, Texas, on 4th June 2002; this phenomenon was also examined through doppler radars data and meteorological soundings. A detailed analysis of the same phenomenon was carried out by Holmes et al. (2008). Gunter and Schroeder (2013) recently started a research project aimed at furnishing a detailed description of thunderstorms based on surface meteorological measurements and images produced by couples of mobile doppler radars with high space and time resolution. The second family of methods consists in the systematic separation and classification of the measurements belonging to large wind datasets, with the aim of developing statistical analyses of the extreme wind velocities and their effects on structures. In the light of the huge amount of the examined data, this family of methods rejects the prohibitive idea of providing a detailed meteorological representation of all the wind phenomena that take place. Therefore, synthetic information is adopted, based on the available data, to make the separation and classification process as automatic as possible. Gomes and Vickery (1976) identified the peak values of the thunderstorm winds with the daily peak values of the wind velocity during thunderstorm days. Riera and Nanni (1989) separated thunderstorm from synoptic wind events depending on the duration of the intense event, the occurrence of thunder and lightning, rainfalls and abrupt temperature drops. Twisdale and Vickery (1992) assumed that the
133
maximum daily velocity has thunderstorm origins during thunderstorm days. Holmes (1999) noted that this assumption is often wrong. Choi (1999) and Choi and Hidayat (2002a) distinguished wind phenomena into thunderstorm and non-thunderstorm winds by defining thunderstorms as those events that occur with thunders and rainfalls; they noted also that the gust factor of thunderstorms is always much greater than the gust factor of nonthunderstorm winds. Choi and Tanurdjaja (2002) separated wind phenomena into large- and small-scale events through visual examinations of long wind velocity records. Kasperski (2002) subdivided the data belonging to frontal depressions, thunderstorms and intermediate events, or gust fronts, based on three parameters: the mean wind velocity, the peak wind velocity and the gust factor. Cook et al. (2003) classified as thunderstorm events the daily maximum wind velocities that occurred in the course of days tagged as “thunderstorms seen or thunderstorms heard”. Duranona et al. (2006) defined as non-synoptic those events that satisfy the following four criteria: the peak velocity is greater than 15 m/s; the ratio between the peak and the mean wind velocity is greater than 1.5; the velocity increases in less than 3 min; the velocity diminishes in less than 10 min. Lombardo et al. (2009) developed an automated procedure, applied to the US datasets acquired through the ASOS network, in order to separate thunderstorms from non-thunderstorm events; after observing that the manual separation of the data is prohibitive, they formulated a method framed into the following steps: (1) the peak wind velocities are extracted from each dataset; (2) the beginning and the end of each thunderstorm are registered; (3) the peak wind velocities that occur between the beginning and the end of a thunderstorm are associated to thunderstorms, all the other peaks being associated to non-thunderstorm events; (4) two datasets including independent thunderstorm and nonthunderstorm events are created, by defining as independent thunderstorm events those separated by at least 4 h, as independent non-thunderstorm events those separated by at least 4 days. Rowcroft (2011) extracted thunderstorm events from measurements carried out on monitored Australian towers for over 20 years; first, he selected the peak wind velocities greater than 40 m/s; then, he defined as thunderstorms those events that satisfy the following four conditions: (1) they last between 5 and 30 min; (2) the temperature has a drop of 1.5 1C or more; (3) there is an increase in wind speed at more than one height of the towers; (4) the wind speed differential is greater than 10 m/s. This paper belongs to this second family of methods, i.e. those methods that systematically separate and classify the anemometric records of large datasets without providing any specific meteorological survey of the examined data. It is also part of a wide research activity carried out in the framework of the European Project “Wind and Ports”, aimed at investigating extreme wind events in port areas (Solari et al., 2012). Section 2 describes the monitoring network that has been realized in the main ports of the High Tyrrhenian Sea for the project “Wind and Ports”; thanks to this network the raw wind velocity data is continuously acquired with a sampling rate of 10 Hz; this data is processed in order to extract the main statistical parameters referred to subsequent 10-min periods; such parameters are ordered into a dataset that represents the starting point for the separation and classification procedure described in the following sections. Section 3 examines the above wind velocity records and the related statistical parameters, classifying them into three families corresponding to stationary and Gaussian, nonstationary and non-Gaussian, and stationary and non-Gaussian events associated, respectively, with extra-tropical depressions, thunderstorms and intermediate phenomena called gust fronts by Kasperski (2002). Coherently with this classification, Section 4 illustrates the properties of the sub-datasets into which the wind
134
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Fig. 1. Anemometric stations installed in the port areas involved in the project “Wind and Ports”.
velocity records and the related statistical parameters are separated. Section 5 describes the algorithm used to carry out such separation; it involves a suitable mix of systematic quantitative controls of synthetic statistical parameters and specific qualitative judgments based on visual inspections of continuous records. Section 6 illustrates the criteria by means of which independent extreme wind velocities are extracted for depressions, thunderstorms and intermediate events, respectively. Section 7 summarizes the main conclusions and points out some perspectives.
Table 1 Main properties of the anemometers of the monitoring network of the project “Wind and Ports”. Port
Anemometer no.
h (m)
Type
Sampling rate (Hz)
Savona
0 1 2 3 4 5 1 2 1 2 3 4 1 2 3 4 5 1 2 3 4 5
84 33 18 27 32 42 61.4 13.3 15.5 13 10 11 20 20 20 20 75 10 10 13 10 10
Tri-axial
10
Bi-axial
10
Bi-axial
10
Tri-axial
10
Bi-axial
2
Genova
2. Monitoring network and main datasets La Spezia
Fig. 1 shows the anemometric monitoring network realized for the project “Wind and Ports – The forecast of the wind for the safety and the management of port areas” (Solari et al., 2012). The project, financed by the European Territorial Cooperation Objective Crossborder program “Italy–France Maritime 2007–2013” and carried out in the period 2010–2012, involved the port authorities of the five main ports in the High Tyrrhenian Sea, namely Genova, La Spezia, Livorno, Savona (Italy) and Bastia (France). The Department of Civil, Chemical and Environmental Engineering (DICCA) of the University of Genova was the scientific actuator. The network is constituted by 22 ultrasonic anemometers (circles) whose most important properties are listed in Table 1; h is the height of the anemometers above ground. Nine new anemometers co-financed by the Port Authority of Genova have been recently installed in the Port of Genova (squares in Fig. 1). In mid-2013, the European Territorial Cooperation Objective Cross-border program “Italy–France Maritime 2007–2013” has financed a new project, “Wind, Ports and Sea – The monitoring and forecast of weather and sea conditions for safe access to the port areas” – which will be developed by the same partners in the period 2013–2015. In the framework of this new project, the monitoring network will be enhanced by the addition of 6 more anemometers, a set of thermometers, barometers and hygrometers and, above all, 3 lidars installed in the Port of Genova, Savona and Livorno. The position of the anemometric instruments has been chosen in order to cover homogeneously the port areas and to register undisturbed wind velocity histories. The instruments are mounted
Livorno
Bastia
on high rise towers or at the top of buildings, at least at 10 m height above ground level, with particular attention to avoid local effects contaminating the measures. Wind measurements are collected with a precision of 0.01 m/s and 11 for intensity and direction, respectively. A set of local servers, placed in each port authority headquarter, receives the measures acquired by the anemometers in their own port area, and elaborates the basic statistics on 10-min periods, namely the mean and peak wind velocities and the mean wind direction. Each server automatically sends this information to the central server located in DICCA. Two files are sent every 10 min containing, for each anemometer, the raw data and the statistical values of the previous 10-min periods. The operational centre of
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
135
Fig. 2. Stationary Gaussian depression recorded on 22th November 2011 by the anemometer 3 of the Port of Savona: (a) 1 h wind speed time-history; (b) distribution of wind speed: histogram and Gaussian density function obtained by data parameters; (c) wind direction time-history; (d) polar histogram of the wind direction.
DICCA stores this data into a central dataset after having systematically checked and validated the data received. The real time transfer is crucial for the short term forecasting (Burlando et al., 2013). Starting from the set of the raw data, a new dataset is created, denoted by 1, which collects the main statistical parameters of the anemometric measures. One record for each subsequent 10-min period is stored, which gathers such parameters into three groups: (1) peak velocity Vp10 averaged on τ ¼ 1 s, mean velocity Vm10, mean direction αm10, gust factor G10 ¼Vp10/Vm10, turbulence intensity I10, skewness γ10 and kurtosis κ10 in T ¼10-min time intervals; (2) mean velocity Vm60, gust factor G60 ¼Vp10/Vm60, turbulence intensity I60, skewness γ60 and kurtosis κ60 in the 1-h interval centred around T; (3) maximum value in T of the mean wind velocity averaged over 1-min Vm1 and gust factor G1 ¼Vp10/Vm1. The dataset 1 represents the starting point for the separation and classification procedure described in the next sections.
3. Classification of intense wind events An examination of the huge amount of the collected data shows that intense wind events can be classified into three families characterized by different properties: (1) stationary (S) Gaussian (G) events, with relatively large mean velocities and small gust factors; such events usually
corresponds to neutral synoptic atmospheric conditions; they are referred to as extended pressure systems by Gomes and Vickery (1976, 1977/1978) and strong frontal depressions by Kasperski (2002, 2009); they are called herein extra-tropical depressions (D); (2) non-stationary (NS) non-Gaussian (NG) events, with large peak velocities and gust factors but relatively small mean velocities; coherently with current literature, they are referred herein to as thunderstorms (T); (3) stationary (S) non-Gaussian (NG) events, with relatively small mean velocities but large peaks and gust factors; they are referred to as intermediate events or gust fronts (F) by Kasperski (2002); in spite of an almost total lack of meteorological interpretations, it seems reasonable to assume that they are associated to strongly unstable atmospheric conditions. A record is referred herein to as stationary when it exhibits statistical regularity for a time period of 10 min. Coherently with the separation and classification procedure described in the next sections, all records reported in the present paper refer to the 1 h interval centred on the considered 10 min record, with the aim of providing a more general overview of the examined wind event. Figs. 2–4 show three typical events of the aforementioned type. For each event, the scheme (a) shows the time-history of the wind speed raw data, the mean value over 1-h periods (horizontal line), the mean values over 10-min subsequent periods (dotted line),
136
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Fig. 3. Non-stationary non-Gaussian thunderstorm recorded on 25th October 2011 by the anemometer 3 of the Port of La Spezia: (a) 1 h wind speed time-history; (b) distribution of wind speed: histogram and Gaussian density function obtained by data parameters; (c) wind direction time-history; (d) polar histogram of the wind direction.
and the 1-s peak (circle) (obviously smaller than the instantaneous peak); the scheme (b) shows the histogram of the wind speed measures, compared with the ideal Gaussian density function matching the mean wind speed and the standard deviation; the scheme (c) shows the time-history of the wind direction raw data, the mean value over 1-h periods (horizontal line) and the mean values over 10-min subsequent periods (dotted line); the scheme (d) shows the polar histogram of the wind direction. Fig. 2 shows a typical 1-h record of a S G event corresponding to a depression (D). The record is registered by the anemometer 3 of the Port of Savona (Fig. 1 and Table 1). The measured event is characterized by a relatively high mean wind velocity (Vm10 ¼ 14.61 m/s, Vm60 ¼15.03 m/s) and by a 1-s gust peak Vp10 ¼22.46 m/s; the turbulence intensity (I10 ¼0.18, I60 ¼ 0.19) and the gust factor (G10 ¼ 1.53, G60 ¼ 1.49) are typical of neutral atmospheric conditions. The skewness (γ10 ¼0.37, γ60 ¼0.02) and the kurtosis (κ10 ¼ 3.42, κ60 ¼3.08) denote a rather typical Gaussian process, especially with reference to a 1-h period; the Gaussian density function fV(v) is close to the histogram of the recorded velocity v. The wind direction has stationary features. Fig. 3 shows a typical 1-h record of a NS NG event corresponding to a thunderstorm (T). The record is registered by the anemometer 3 of the Port of La Spezia (Fig. 1 and Table 1). The measured event is characterized by a relatively low mean velocity (Vm10 ¼9.09 m/s, Vm60 ¼ 7.33 m/s), a very high 1-s gust peak Vp10 ¼33.36 m/s and a very high gust factor (G10 ¼3.67, G60 ¼4.55). The skewness (γ10 ¼1.55, γ60 ¼1.20) and the kurtosis
(κ10 ¼4.47, κ60 ¼5.60) denote a typical non-Gaussian distribution. The time-history of the wind direction shows a sudden rotation of about 901 just in correspondence of the gust peak. The polar histogram of the wind direction points out this aspect through a bi-modal distribution. Fig. 4 shows a typical 1-h record of a S NG event corresponding to a gust front (F). The record is registered by the anemometer 3 of the Port of La Spezia (Fig. 1 and Table 1). The measured event is characterized by a low mean velocity (Vm10 ¼8.12 m/s, Vm60 ¼ 8.77 m/s) and by a relatively intense 1-s gust peak Vp10 ¼ 18.66 m/s; the turbulence intensity (I10 ¼ 0.38, I60 ¼ 0.31) and the gust factor (G10 ¼2.30, G60 ¼2.13) are much higher than those which are typical of neutral atmospheric conditions. The skewness (γ10 ¼0.42, γ60 ¼0.33) and the kurtosis (κ10 ¼2.60, κ60 ¼ 2.89) denote a moderately non-Gaussian distribution. Likewise the wind intensity, also the time-history of the wind direction is characterized by a relatively large variability, clearly pointed out by the spread in the polar histogram of the wind direction.
4. Separation flowchart into sub-datasets Fig. 5 shows the conceptual flowchart of the procedure by means of which the original set of the raw data and the related statistical parameters are transformed into more sub-datasets associated with the different wind events described in Section 3.
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
137
Fig. 4. Stationary non-Gaussian gust front recorded on 16th December 2011 by the anemometer 3 of the Port of La Spezia: (a) 1 h wind speed time-history; (b) distribution of wind speed: histogram and Gaussian density function obtained by data parameters; (c) wind direction time-history; (d) polar histogram of the wind direction.
α
γ
κ
Fig. 5. Conceptual flowchart of the procedure by means of which the original base of the raw data is transformed into more sub-datasets associated with different wind events.
(a) The series of the mean wind velocities Vm10 stored in the dataset 1 (Section 3) is processed by the method of storms proposed by Cook (1982). The mean wind velocities over subsequent 10 h periods is first calculated; such velocities are searched sequentially from the beginning, in order to find the first occurrence on which the wind speed drops below a given threshold Vn, defined as the start of a lull; between the start of two subsequent lulls there is a wind storm; the threshold Vn is chosen in such a way as to identify, for each anemometer, about 100 storms per year.
Table 2 Number of records in each dataset, with reference to the anemometer 3 of the Port of La Spezia and the anemometer 4 of the Port of Livorno. Dataset
Anemometer 3 Port of La Spezia
Anemometer 4 Port of Savona
1 2 3 4 5 6 7 8 9 10 11 12 13
105,264 695 244 24 17 450 17 228 200 35 12 32 15
131,472 3275 4559 71 83 3230 24 21 4506 94 21 17 81
(b) Starting from the dataset 1, two datasets denoted by 2 and 3 are created. The dataset 2 contains the records for which the peak wind velocities satisfy the condition Vp10 ZVnp10; the dataset 3 contains the records for which the mean wind velocities satisfy the condition Vm10 ZVnm10; in this case, Vnp10 ¼15 m/s and Vnm10 ¼10 m/s. (c) Starting from the datasets 2 and 3, two datasets denoted by 4 and 5 are created. They contain, respectively, the sub-set of the records associated with the maximum values of Vp10 and Vm10 for each storm, irrespective of any separation into families of homogeneous events.
138
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Fig. 6. (a) Roughness length z0 of the area of the Port of Livorno; reference turbulence intensity I0 at 20 m height above ground, when the wind blows from: (b) α ¼ 1251; (c) α ¼ 3051.
(d) Starting from the dataset 2, three datasets denoted by 6, 7 and 8 are extracted. The dataset 6 contains the sub-set of the records associated with all the Vp10 values corresponding to S G depressions (D). The dataset 7 contains the sub-set of the records associated with all the Vp10 values corresponding to NS NG thunderstorms (T). The dataset 8 contains the sub-set of the records associated with all the Vp10 values corresponding to S NG gust fronts (F). The classification of each event into one of the three categories D, T or F is carried out by means of the algorithm illustrated in Section 5. (e) Starting from the dataset 3, one dataset denoted by 9 is extracted. It contains the sub-set of the records associated with all the Vm10 values due to S G depressions (D) and corresponds, in terms of mean wind velocities, to the dataset 6 in terms of peak wind velocities. (f) Starting from the datasets 6, 7, 8 and 9, four datasets denoted by 10, 11, 12 and 13 are extracted. The datasets 10 and 13 contain, for each subsequent storm, the sub-set of the records associated with the maximum independent Vp10 and Vm10 values corresponding, respectively, to S G depressions (D). The datasets 11 and 12 contain the sub-set of the records associated with the maximum independent Vp10 and Vm10 values corresponding, respectively, to NS NG thunderstorms (T) and S NG gust fronts (F); the concept of independence is discussed in Section 6. Table 2 shows the number of records belonging to each of the above sub-datasets, with reference to the anemometer 3 of the Port of La Spezia and to the anemometer 4 of the Port of Livorno (Fig. 1 and Table 1). The smaller percentage of the records belonging to the datasets 2 and 3 in the Port of La Spezia with respect to the Port of Livorno depends on the less intense velocities measured, due to the lower height of the anemometer and to the major complexity of its surrounding.
5. Separation and classification algorithm Each anemometer is qualified via a card that reports its position and the main technical characteristics. This card also provides, at height h over ground and as a function of the direction αm10 of the oncoming wind, the turbulence intensity I010 and the three reference gust factors G010, G060 and G01; the apex 0 indicates that such values are evaluated numerically assuming, as it is classical for synoptic events, that intense wind velocities occur in neutral atmospheric conditions and are stationary Gaussian processes. Fig. 6 shows the turbulence intensity I010 in the Port of Livorno, at 10 m height above ground, when the wind blows from
αm10 ¼1251 (a) and 3051 (b). These results are obtained by the procedure formulated by ESDU (1993), in an advanced version developed at DICCA (Solari et al., 2012). Fig. 7 shows the diagrams of the reference gust factors G010, G060 and G01 as functions of the mean wind direction αm10 at the anemometers 1, 2 and 3 in the Port of La Spezia (Fig. 1 and Table 1). These quantities are evaluated through the method proposed by Solari (1993). It is apparent that their values decrease when the wind blows from the sea. The separation among depressions (D), thunderstorms (T) and gust fronts (F), indicated in Fig. 5 as the extraction of the datasets 6, 7 and 8 from the dataset 2, and of the dataset 9 from the dataset 3, is carried out through a semi-automated procedure shown by the flow-chart in Fig. 8. Such procedure is a suitable mix of systematic quantitative controls, based on the comparison between measured gust factors and their reference values, and qualitative judgments, based on the visual examination of the wind velocity records and their skewness and kurtosis values. It is apparent that this algorithm is efficient provided that the number of the qualitative controls is very limited. The procedure involves the following steps: (1) Events for which the following rule applies are classified as depressions (D): G60 G060
r 1:10
ð1Þ
They are strongly stationary and Gaussian not only over 10-min intervals, but also over 1-h intervals. Fig. 9a shows a typical record belonging to this family of events. (2) Events for which the following rule applies are classified as depressions (D): G60 G060
r 1:25 \
G10 G010
r 1:10
ð2Þ
They are strongly stationary and Gaussian over 10-min intervals, while they exhibit some variability over 1-h intervals. Fig. 9b shows a typical record belonging to this family of events. (3) Events for which the following rule applies are classified as thunderstorms (T) or gust fronts (F): G10 G010
4 1:25
ð3Þ
The classification of such events into the category of thunderstorms (T) (see for instance the record in Fig. 9c) or gust fronts (F) (Fig. 9d) is carried out through a qualitative control.
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
139
Fig. 7. Reference gust factors G010, G060 and G01 as functions of the mean wind direction at the anemometers 1, 2 and 3 in the Port of La Spezia.
D Y
Y
D, T, F
1
N
Y
2
Data set Quantitative control Expert judgment
3 N
N
N
4
N Y
T?
N
F?
T? N
Y
Y
F
Y
T
Fig. 8. Flow-chart of the semi-automated algorithm by means of which depressions (D), thunderstorms (T) and gust fronts (F) are separated.
(4) Events for which the following rule applies are classified as depressions (D): G60 G060
o 1:25 \
G1 4 0:80 G10
ð4Þ
Fig. 9e shows a typical record belonging to this family of events. Those events that do not satisfy Eq. (4) are classified as thunderstorms (T) (see for instance the record in Fig. 9f), gust fronts (F) (Fig. 9g) or depressions (D) (Fig. 9h) through a qualitative control. It is worth noting that the event depicted in Fig. 9h involves a sudden growth of the wind velocity after about 20 min from the start of the record; such growth is not classified by the algorithm as a thunderstorm since the corresponding peak wind velocity is lower than the threshold Vnp10 ¼15 m/s. Table 3 summarizes the main statistical parameters of the 8 events shown in Fig. 9. The qualitative controls carried out during the above procedure are based on the visual examination of the wind velocity records and on the interpretation of synthetic statistical parameters with reference to which it is not easy to formulate general quantitative rules. As an example, an indicator
for classifying an event not attributable to a depression (D) as a thunderstorm (T) or a gust front (F) is the ratio G10/G60: when it is less than 0.90, the event is usually a thunderstorm (T); when it is greater than 0.90, the event is usually a gust front (F). A skewness γ10 0 and a kurtosis κ10 3 are parameters typical of a Gaussian depression (D); different values of γ10 and κ10 indicate the occurrence of a non-Gaussian thunderstorm (T) or gust front (F). In the two cases of the data registered by the anemometer 3 of the Port of La Spezia and by the anemometer 4 of the Port of Livorno, the number of the required qualitative control is rather limited: 349 out of 105,264 (0.33%) and 86 out of 131,472 (0.06%), respectively. The higher number of qualitative controls in the Port of La Spezia is mainly associated with the major complexity of its territory. It is worth noting, finally, that there are some anomalous events associated with intense wind velocities, very rare, not clearly attributable to any of the three types of phenomena discussed above (D, T e F). Fig. 10 shows three examples of such phenomena. The first one is characterized by a sudden growth of the wind velocity (Fig. 10a), not followed by an analogous reduction, which occurs in correspondence of a sudden rotation of the wind direction (Fig. 10b). The second one involves an extremely rapid change of the wind velocity (Fig. 10c), not associated to particular changes of the wind direction (Fig. 10d). The third one exhibits a sudden reduction of the wind velocity (Fig. 10e), after a period in which the velocity was intense and stationary, associated with a sudden rotation of the wind direction (Fig. 10f).
6. Extraction of independent events Once depressions (D), thunderstorms (T) and gust fronts (F) have been separated into different sub-datasets by the criterion described in Section 5, it remains to extract, from each of them, a suitable series of independent extreme wind velocities. The maximum values of the depressions associated with different successive storms, both in terms of mean and peak
140
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Fig. 9. Some examples of different events examined by the separation algorithm during subsequent steps: (a) depression; (b) depression; (c) thunderstorm; (d) gust front; (e) depression; (f) thunderstorm; (g) gust front; (h) depression.
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Table 3 Main synthetic statistical parameters of the 8 events shown in Fig. 9. Event
Vp10 (m/s)
G10
G60
G1
γ10
γ60
κ10
κ60
a b c d e f g h
16.49 14.90 18.00 15.90 18.41 15.52 19.07 16.12
1.54 1.42 2.90 1.98 1.81 1.80 1.84 1.28
1.55 1.71 5.76 1.99 1.89 4.53 2.10 2.01
1.36 1.34 1.21 1.52 1.52 1.17 1.40 1.12
0.14 0.06 0.73 0.83 0.29 0.04 0.34 0.09
0.05 0.10 1.33 0.19 0.24 0.93 0.41 0.47
2.62 2.45 2.26 3.39 3.08 1.83 2.92 2.61
2.94 2.41 5.67 3.41 3.22 3.25 3.27 1.58
141
values, are dealt with as independent, provided that they are separated by at least 72 h. The maximum values of the thunderstorms are dealt with as independent provided that they are separated by at least 4 h. In a very preliminary way, not yet sustained by sound motivations, the maximum values of the gust fronts are dealt with as independent provided that they are separated by at least 24 h. It remains an open question, whether the maximum values associated with different events close together or even consequential should be treated as dependent or independent. Fig. 11 shows an example of this situation in which a depression whose
Fig. 10. Anomalous events recorded by the anemometer 4 of the Port of Livorno: wind velocity (a) and direction (b) on 16th October 2010; wind velocity (c) and direction (d) on 31st October 2010; wind velocity (e) and direction (f) on 16th December 2011.
142
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Fig. 11. Intense depression followed by an intense thunderstorm, recorded by the anemometer 1 of the Port of Livorno on 16th December 2011: (a) wind velocity; (b) wind direction.
maximum velocity is classified as an extreme is followed by a thunderstorm that represents, in this family of events, another extreme.
7. Conclusions and prospects This paper deals with the management of large wind velocity datasets, in order to separate and classify independent extreme wind events through a semi-automated procedure involving a suitable mix of systematic quantitative controls of synthetic statistical parameters and specific qualitative judgments based on visual inspections of continuous records. Two aspects are worth noting. First, though most of the papers published on the present topic tend to separate thunderstorms from synoptic non-thunderstorm events, this classification cannot be easily pursued. As Kasperski (2002) first pointed out, there is at least a third class of wind events, with intermediate properties, which greatly complicate the “binary” approach that prevails in the literature. However, the meteorological knowledge of these events is definitely poor and their understanding deserves further research. Second, in order to apply the classification criterion discussed in the present paper, several statistical parameters not commonly available should be used and recourse should be made to qualitative judgments based upon the visual inspection of continuous velocity records. This throws quite a few shadows on some extraction and classification criteria reported in literature, involving the use of only synthetic parameters. Since the separation of wind events in different classes is functional to carry out refined analyses, their meaning becomes somewhat questionable if it is not equally accurate the preliminary separation process. Both these aspects have profound impact on the dynamic wind-excited response of structures and on the mixed statistics of extreme wind velocities and extreme wind-induced effects (Solari, 2013). In mid-2013, the European Community approved a new project, “Wind, Ports and Sea”, carried out by the same partners already involved in the project “Wind and Ports” that provided the data for these analyses. Among other issues, this project is aimed at strengthening and extending the wind monitoring network described in Section 2. Taking into account the focal role of huge and high quality data to carry out studies such as those reported in the present paper, the authors rely on the possibility of improving the results of this research when the new data will be available.
References Burlando, M., De Gaetano, P., Pizzo, M., Repetto, M.P., Solari, G., Tizzi, M., 2013. Wind short-term forecast in port areas. In: Proceedings of the 6th European and African Conference on Wind Engineering. Cambridge, U.K. Charba, J., 1974. Application of gravity current model to analysis of squall-line gust front. Mon. Weather Rev. 102, 140–156. Chay, M., Albermani, F., 2005. Dynamic response of a SDOF system subjected to simulated downburst winds. In: Proceedings of the 6th Asia-Pacific Conference on Wind Engineering. Seoul, Korea. pp. 1562–1584. Chen, L., Letchford, C.W., 2004. Parametric study on the alongwind response of the CAARC building to downbursts in the time domain. J. Wind Eng. Ind. Aerodyn. 92, 703–724. Choi, E.C.C., 1999. Extreme wind characteristics over Singapore – an area in the equatorial belt. J. Wind Eng. Ind. Aerodyn. 83, 61–69. Choi, E.C.C., Hidayat, F.A., 2002a. Gust factors for thunderstorm and nonthunderstorm winds. J. Wind Eng. Ind. Aerodyn. 90, 1683–1696. Choi, E.C.C., Hidayat, F.A., 2002b. Dynamic response of structures to thunderstorm winds. Prog. Struct. Eng. Mater. 4, 408–416. Choi, E.C.C., Tanurdjaja, A., 2002. Extreme wind studies in Singapore. An area with mixed weather system. J. Wind Eng. Ind. Aerodyn. 90, 1611–1630. Cook, N.J., 1982. Towards better estimation of extreme winds. J. Wind Eng. Ind. Aerodyn. 9, 295–323. Cook, N.J., Harris, R.I., Whiting, R., 2003. Extreme wind speeds in mixed climates revisited. J. Wind Eng. Ind. Aerodyn. 91, 403–422. Duranona, V., Sterling, M., Baker, C.J., 2006. An analysis of extreme non-synoptic winds. J. Wind Eng. Ind. Aerodyn. 95, 1007–1027. Engineering Sciences Data Unit, 1993. Computer Program for Wind Speeds and Turbulence Properties: Flat or Hill Sites in Terrain with Roughness Changes, ESDU Item 92032. London, UK. Fujita, T.T., 1990. Downburst: meteorological features and wind field characteristics. J. Wind Eng. Ind. Aerodyn. 36, 75–86. Goff, R.G., 1976. Vertical structure of thunderstorm outflows. Mon. Weather Rev. 104, 1429–1440. Gast, K.D., Schroeder, J.L., 2003. Supercell rear-flank downdraft as sampled in the 2002 thunderstorm outflow experiment. In: Proceedings of the 11th International Conference on Wind Engineering. Lubbock, Texas. Gomes, L., Vickery, B.J., 1976. On thunderstorm wind gusts in Australia. Civil Eng. Trans. Inst. Eng. Aust. 18, 33–39. Gomes, L., Vickery, B.J., 1977/1978. Extreme wind speeds in mixed climates. J. Ind. Aerodyn. 2, 331–344. Gunter, W.S., Schroeder, J.L., 2013. High-resolution full-scale measurements of thunderstorm outflow winds. In: Proceedings of the 12th Americas Conference on Wind Engineering. Seattle, Washington. Hjelmfelt, M.R., 1988. Structure and life cycle of microburst outflows observed in Colorado. J. Appl. Meteorol. 27, 900–927. Holmes, J.D., 1999. Modelling of extreme thunderstorm winds for wind loading of structures and risk assessment. In: Larsen, A., Larose, G. (Eds.), Wind Engineering into the 21st Century. Balkema, Rotterdam Holmes, J.D., Hangan, H.M., Schroeder, J.L., Letchford, C.W., Orwig, K.D., 2008. A forensic study of the Lubbock-Reese downdraft of 2002. Wind Struct. 11 (2), 19–39. Kasperski, M., 2002. A new wind zone map of Germany. J. Wind Eng. Ind. Aerodyn. 90, 1271–1287. Kasperski, M., 2009. Specification of the design wind load-A critical review of code concepts. J.Wind Eng.Ind. Aerodyn. 97, 335–357. Kwon, D.K., Kareem, A., 2009. Gust-front factor: new framework for wind load effects on structures. J. Struct. Eng. ASCE 135 (6), 717–732.
P. De Gaetano et al. / J. Wind Eng. Ind. Aerodyn. 126 (2014) 132–143
Kwon, D.K., Kareem, A., 2013. Generalized gust-front factor: a computational framework for wind load effects. Eng. Struct. 48, 635–644. Lombardo, F.T., 2012. Improved extreme wind speed estimation for wind engineering applications. J. Wind Eng. Ind. Aerodyn. 104–106, 278–284. Lombardo, F.T., Main, J.A., Simiu, E., 2009. Automated extraction and classification of thunderstorm and non-thunderstorm wind data for extreme-value analysis. J. Wind Eng. Ind. Aerodyn. 97, 120–131. Riera, J.D., Nanni, L.F., 1989. Pilot study of extreme wind velocities in a mixed climate considering wind orientation. J. Wind Eng. Ind. Aerodyn. 32, 11–20. Riera, J.D., Viollaz, A.J., Reimundin, J.C., 1977. Some recent results on probabilistic models of extreme wind speeds. J. Wind Eng. Ind. Aerodyn. 2, 271–287. Rowcroft, J., 2011. Vertical wind shear profiles in downburst events and the insufficiency of wind turbine design codes. In: Proceedings of the 13th International Conference on Wind Engineering. Amsterdam, The Netherlands. Sherman, D.J., 1987. The passage of a weak thunderstorm downburst over an instrumented tower. Mon. Weather Rev. 115, 1193–1205. Solari, G., 1993. Gust buffeting. I: peak wind velocity and equivalent pressure. J. Struct. Eng. ASCE 119, 365–382. Solari, G., 2013. Emerging issues and new scenarios for wind loading on structures. In: Proceedings of the 6th International Symposium on Wind Effects on
143
Buildings and Structures: Current State-of-the-Art in Wind Engineering and Outlook for the Future. Tokyo, Japan. pp. 153–176. Solari, G., De Gaetano, P., Repetto, M.P., 2013. Thunderstorm response spectrum, In: Proceedings of 12th Americas Conference on Wind Engineering. Seattle, Washington. Solari, G., Repetto, M.P., Burlando, M., De Gaetano, P., Pizzo, M., Tizzi, M., Parodi, M., 2012. The wind forecast for safety management of port areas. J. Wind Eng. Ind. Aerodyn. 104–106, 266–277. Thom, H.C.S., 1967. Toward a universal climatological extreme wind distribution. In: Proceedings of the 2nd International Conference on Wind Effects on Buildings and Structures. Ottawa, Canada. pp. 669–683. Thom, H.C.S., 1968. New distributions of extreme wind speeds in the United States. J. Struct. Div. ASCE, 94; , pp. 1787–1801. Twisdale, L.A., Vickery, P.J., 1992. Research on thunderstorm wind design parameters. J. Wind Eng. Ind. Aerodyn. 41, 545–556. Wakimoto, R.M., 1982. The life cycle of thunderstorm gust fronts as viewed with Doppler radar and rawinsonde data. Mon. Weather Rev. 110, 1060–1082. Yeo, D.H., 2011. Database assisted design for high-rise structures in mixed extreme wind climates. In: Proceedings of the 13th International Conference on Wind Engineering. Amsterdam, The Netherlands.