1. Introduction
High wind speeds and wind gusts from nontornadic storms are a significant meteorological hazard in the world that cause considerable economic and social costs as well as damage to properties, infrastructure, agriculture, and trees (Dore 2003; Changnon 2009; Lopes et al. 2009; Pinto et al. 2010). Typically, in Ontario, Canada, these extreme wind events result from intense synoptic storms or convective activity or combinations of both. It is often the wind gusts or the extreme winds that cause the wind-related damages and, hence, are the most crucial measurement for wind hazard analyses (Sanabria and Cechet 2010; Cechet and Sanabria 2010). Wind gusts represent one of the major loads that impact buildings and other infrastructure. Of all structures, it is probably the electrical power distribution and transmission line systems that are among the most affected by extreme winds as a result of downed trees and poles (Changnon 1980). When wind gusts exceed a certain collective wind pressure design or “breaking point” or threshold, damage to structures can increase significantly as a result of wind-induced loads on the structure. For safety and engineering design purposes, climate information of wind gusts is required to determine design wind loads for infrastructure codes and standards. These design wind loads play an important role in preventing or limiting damage and in reducing risks to lives from extreme winds.
To diminish the devastating effects associated with extreme wind gust events, there continues to be a substantial amount of research conducted on determining the relationships that exist between wind speed and wind gust events. It is well known that a gust factor has been applied to estimate the peak wind gust of a specific duration based on the mean wind speed for a period of time (Davis and Newstein 1968; Mitsuta and Tsukamoto 1989; Ahmed 1994; Weggel 1999; Sparks and Huang 2001; Jungo et al. 2002; Cvitan 2003; Paulsen and Schroeder 2005; Graybeal 2006). It can be concluded that the gust factor provides an accurate method for computing wind gust speeds (Jungo et al. 2002) and the method is suitable for the development of a meteorological background to support the standards in designing overhead power-transmission lines (Cvitan 2003).
Potential changes to future wind regimes as a result of a changing climate have also received attention from the climate community. For example, Pryor et al. (2005, 2006) applied an empirical downscaling technique to generate probability distributions of wind speeds at sites in northern Europe for two future periods (2046–65 and 2081–2100). They concluded that projected changes in the downscaled mean and 90th percentile wind speeds from different global climate models (GCMs) are less than ±15% and are comparable to the current variability of historical periods (1961–90 and 1982–2000). Beniston et al. (2007) used a regional climate model (RCM) simulation in Europe to analyze changes in 10-m wind speeds for the two 30-yr time slices (1961–99 and 2071–2100). They found that the numbers of wind speed events per year exceeding the control period 95th and 99th percentiles increase by 10% over Switzerland, especially north of the Alps. More recently, projections of changes in future wind speed or windstorms have been conducted to relate wind energy and windstorm insurance losses. For example, using eight GCM outputs, Ren (2010) showed that due to a reduction in future wind speed projected by GCMs, all eight models indicated that the future accessible wind energy for the period 2071–2100 over China would degrade, with varying reduction rates among the models (decreasing by up to 14% from current conditions). Schwierz et al. (2010) applied RCM and GCM A2-scenario simulations over Europe to drive an operational insurance loss model, with the result that European-wide annual expected losses by 2071–2100 are projected to increase by 44% from their current levels (1961–90), owing to increases in both severity and frequency of wind gusts. Pinto et al. (2010) used a statistical–dynamical downscaling approach to assess windstorm impacts over western Germany under future climate conditions for an A2 scenario. They concluded that wind gusts for the period 2061–2100 are projected to increase by 5%; consequently, windstorm losses are expected to increase substantially by 19%.
However, to date, it appears that very few studies have investigated the possible impacts of climate change on future wind gusts in Canada. Under a global warming scenario, the severity and frequency of future wind gust events could be expected to change late this century. The scientific information on projections of future wind gust events is essential for decision makers to improve the adaptive capacity of the infrastructure at risk of being impacted by extreme wind gust events. In light of this concern, Environment Canada has recently completed a climatological research project to investigate climate change and wind gust events over the province of Ontario. The current paper presents this project, which is made up of a three-step process: 1) regression-based downscaling to derive future station-scale hourly wind speed data, 2) development and validation of wind gust simulation models to identify historical wind gust events, and 3) projections of changes in the frequency of future wind gust events from the historical conditions. We hope this study provides some insight into the future vulnerability of various industries (e.g., transportation, agriculture, energy, commerce) and communities to extreme wind gust events and provides them with scientific information in adapting to and managing the associated risks.
This paper is organized as follows. In section 2, data sources and treatments are described. Section 3 summarizes the statistical downscaling developed by Cheng et al. (2008) since methods from that study are used in the current analysis. Section 4 presents analysis techniques as applied to the development and validation of wind gust simulation models and the projection of future hourly and daily wind gusts. Section 5 includes the results and a discussion on historical hourly and daily wind gust simulation and validation, changes in the frequency of future hourly and daily wind gust events, contrasts with regional climate model simulations, historical evidence to support future projections, and uncertainties of the study. The conclusions from the study are summarized in section 6.
2. Data sources and treatment
To evaluate possible impacts of climate change on future wind gust events, a number of weather variables, including surface meteorological observations, reanalysis data, and GCM simulations, are needed. The information on data descriptions, variables, and sources is briefly presented in Table 1. Surface hourly meteorological observations for 14 stations in the province of Ontario, as shown in Fig. 1, were used in the study. With the exception of wind gusts, when the observed hourly data were missing for three consecutive hours or less, missing data were interpolated using a temporal linear method; days with data missing for four or more consecutive hours were excluded from the analysis. Across the study area, 1.07% of the total days required missing data interpolation; after interpolation, the dataset was 99.84% complete. In addition to the meteorological observations, for statistical downscaling, the 6-hourly reanalysis data at 0600, 1200, 1800, and 0000 UTC for the period 1958–2007 from the National Centers for Environmental Prediction (NCEP) web site were included in the analysis. To combine the gridded reanalysis data with the surface observed data, the reanalysis data from the surrounding four-grid domain fields were interpolated to the selected weather stations using the inverse-distance method (Shen et al. 2001). Cheng et al. (2010) have tested two other domain sizes (i.e., 16 and 36 grids) and concluded that the four-grid interpolation is best based on its correlation with radiosonde data.
Data used in the study.
Study area and location of the selected cities in Ontario.
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
Of all measured and archived meteorological fields, wind data are probably one of the most variable in Canada for protocols, instrument siting, anemometer heights and types, measurement averaging periods, instrument maintenance, and uncertainties. Typically, the averaging period of the wind measurements observed at hourly meteorological stations can vary from an hourly average value to a 10-min average taken before the hour to 1- and 2-min spot wind averages prior to the top of the hour down to gusts for a few seconds. In Canada, prior to 1980, hourly wind speeds were reported by averaging speeds over a period of 1 min just prior to the hour, which was changed to a 2-min averaging period after 1980. There are two kinds of observed wind gusts in Environment Canada: hourly and daily wind gusts. An hourly wind gust is defined as a sudden increase in wind speed occurring during the 10-min period prior to the observation with a speed of at least 28 km h−1 (15 kt) and measured at 9 km h−1 (5 kt) greater than the 2-min-average wind speed prior to the observation (Environment Canada 1977). The daily wind gust is defined as a daily peak wind that is the highest instantaneous wind greater than 28 km h−1 (15 kt) measured during an entire 24-h period. Wind gust speed recordings can be some of the most difficult to interpret for a number of reasons. Anemometers are typically maintained and calibrated for mean or averaged wind speeds, while the averaging period used in wind gust measurements is more dependent on the transient response of the anemometer and its chart recorder during a short wind gusts (e.g., 1–3-s duration), if present. Variations in the actual averaging periods represented by the measurements pose a serious limitation on the consistency and accuracy of the wind gust data (Cechet and Sanabria 2010). On the other hand, wind gust measurements are less sensitive to anemometer and site exposures than are winds obtained using longer averaging periods and may be more spatially consistent and representative of the wind climatology of the larger area.
To more effectively develop wind gust simulation models, the consistency of observed wind data in terms of instrument seting and anemometer heights over the time period is very important. Regarding variations in the observed wind speed and wind gust data, a certain time period of the wind data record with a certain consistency in the wind data observations needs to be identified. To achieve this, the wind metadata and annual total frequencies of wind speed/wind gust greater than the certain value were analyzed for each of 14 selected stations in Ontario. The wind metadata describe changes in instrument seting, anemometer heights and types, measurement averaging periods, and instrument maintenance, for the observation period. Annual total frequencies of the events with wind speeds >43 km h−1 and daily wind gusts ≥60 km h−1 are shown in Fig. 2 for Windsor as an example. Note that, however, the patterns of the event frequencies are similar when the events are defined by different wind speed values. To select a specific time period of the data record, the time series of the frequency of wind speed/wind gust events and wind metadata shown in Fig. 2 were visually examined. To maintain consistency in the frequency of wind speed/wind gust events and in instrument location and anemometer heights, for Windsor, the daily wind gust and hourly wind speed data observed for the time period 1964–2007 were selected for the analysis. Similarly, the time period of daily wind gust and hourly wind speed data records was selected for each of the other stations; for example, the time periods 1963–2007, 1968–2007, 1971–2007, and 1974–2007 for Trenton, London, Ottawa, and Toronto, respectively, were chosen. Hourly wind gust data observed for the time period 1994–2007 were used for all stations studied since the data are available only back to 1994.
Wind metadata and annual total occurrence frequency of the events with (top) hourly wind speed >43 km h−1 and (bottom) daily wind gust ≥60 km h−1 in Windsor.
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
In addition to observed meteorological data, daily future climate data projected from eight GCMs with two Intergovernmental Panel on Climate Change (IPCC) climate change scenarios [Special Report on Emissions Scenarios (SRES) A2 and B1], as described in Table 1, were used in the analysis. The GCM simulations were retrieved from the web site of the Program for Climate Model Diagnosis and Intercomparison (PCMDI 2008). These eight GCMs were selected simply because the simulations of surface and upper-air temperature, pressure, and u and υ wind are available online, used in our climate change impact studies. Scenarios A2 and B1 consider different assumptions of future greenhouse gas (GHG) emissions derived from a distinctly different direction for future population growth, economic development, and technological change. As described in the IPCC Climate Change 2007: Synthesis Report (Bernstein et al. 2007), the global GHG emissions for scenario A2 almost linearly increase in this century, with emissions by the year 2100 being 2.5 times greater than in 2000. Compared to scenario A2, scenario B1 produces much lower emissions: slightly increasing in the first half of this century with emissions by the year of 2040 being about 45% greater than in 2000; then decreasing in the second half of this century with emissions by the year 2100 being about 40% lower than in 2000. Consequently, scenario B1 projects less future warming than does scenario A2: the global average temperature increase in 2090–2100 relative to 1980–99 is projected to be 3.4° and 1.8°C as the best estimates for scenarios A2 and B1, respectively. Both scenarios were used in the study to generate a range of projections of possible climate change impacts on future hourly and daily wind gust events.
3. Summary of statistical downscaling
To project future hourly and daily wind gusts, future hourly wind speed data at the station scale are necessary for use in the wind gust simulation models, which will be described and developed in section 4. To achieve this, the regression-based downscaling method developed by Cheng et al. (2008) was adapted for this study to downscale future hourly wind information. The downscaling method comprises a two-step process: 1) spatially downscaling daily u and υ winds from the GCM domain field to the selected weather stations shown in Fig. 1 and 2) temporally downscaling daily u and υ wind speeds to hourly time steps. The downscaling transfer functions for winds were constructed using three regression techniques: multiple stepwise regression, orthogonal regression, and autocorrelation correction regression. The predictors selected in development of the hourly/daily wind downscaling transfer functions include hourly/daily u and υ wind, west–east and south–north sea level pressure gradients across the weather station, and u and υ winds from the previous hour. As discussed in Cheng et al. (2008), these regression techniques are suitable for developing wind downscaling transfer functions since the u and υ wind data and selected predictors are normally distributed.
Performance of the wind downscaling transfer functions was evaluated by 1) analyzing amodel coefficient of determination R2 of downscaling transfer functions, 2) validating downscaling transfer functions using the leave-one-year-out cross-validation scheme, and 3) comparing extreme characteristics and data distributions between downscaled GCM historical runs and observations over a comparative time period (1961–2000). The evaluation results showed that the regression-based downscaling method performed very well in deriving future hourly and daily station-scale wind data. Most of the daily downscaling transfer functions for surface winds possess R2 greater than 0.8; the corresponding R2 of the hourly downscaling transfer functions range from 0.69 to 0.92 with half of them greater than 0.89. One major reason for the performance of the hourly downscaling transfer functions being slightly better than the daily ones is that hourly downscaling transfer functions were developed based on relationships between observed hourly and daily winds. However, daily downscaling transfer functions were constructed based on relationships between station-scale observed and regional-scale reanalysis winds. In addition, the distributions of the daily wind speed data derived from downscaled GCM historical runs and raw GCM simulations have been evaluated to show the performance of downscaling transfer functions. The evaluation results indicate that downscaling transfer functions can remove most GCM biases, as shown in Fig. 3, as an example. It can be seen that, from Fig. 3, compared to the observed winds, raw CGCM3 and GFDL-CM2.0 simulations clearly show overestimated and underestimated results, respectively. However, all downscaled GCM historical runs have data distributions similar to the observations over a comparable time period (1961–2000). Details of the hourly and daily downscaling methods and their results are not presented in this current paper due to the limitations of space; refer to Cheng et al. (2008) for details.
The distribution of wind speeds as shown by (top, column 1) observations in Toronto, (top, columns 2–4) downscaled GCM historical runs, and (bottom) interpolated from raw GCM historical runs, over a comparable time period (1961–2000).
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
4. Analysis techniques
a. Development and validation of wind gust simulation models
Two kinds of wind gust factors are used in this study: 1) the hourly wind gust factor defined by the ratio of the hourly wind gust to hourly wind speed and 2) the daily wind gust factor defined by the ratio of the daily wind gust to the 1-h maximum wind speed (the maximum wind speed of hourly observations during a day). The wind gust factors defined in the study are not a constant value but vary with the wind speed V. The relationships between kg(V) and V were examined for each of the selected stations. As an example, the results of relationships between hourly wind gust factors and hourly wind speed in each of the four seasons for Windsor are summarized in Fig. 4; the corresponding relationship between daily wind gust factors and 1-h maximum wind speed are illustrated in Fig. 5. Similar results were found for other stations as well (not shown). The wind gust simulation models were developed separately for four seasons in order to assess climate change impacts on two different scale windstorm systems (synoptic and localized convective windstorms) that typically occurred in the study area. Synoptic windstorms occur during the winter season, while localized convective windstorms generally occur during summer. The solid lines shown in Fig. 4 represent the average hourly wind gust factors at each of the observed wind speed values in kilometers per hour. These lines were used to multiply with hourly wind speed for each individual hour to simulate hourly wind gusts. As shown in Fig. 5, daily wind gusts were simulated by the corresponding lines representing the relationships between the daily wind gust factor and 1-h maximum wind speed of hourly observations during a day. These wind gust simulation models were validated using a leave-one-year-out cross-validation procedure in which the procedure was repeatedly run to develop a wind gust simulation model that would validate one year of independent data for each year in the dataset. The calibrated/validated hourly/daily gust speeds were then compared with observations to evaluate model performance. These simulation models were used to project future hourly/daily wind gusts based on downscaled future hourly wind speeds.
Relationship between hourly wind gust factors and hourly wind speed during 1994–2007 in Windsor.
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
Relationship between daily wind gust factors and 1-h maximum wind speed during 1964–2007 in Windsor.
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
b. Projection of future hourly and daily wind gusts
Following downscaling of the future hourly wind speed described above, future hourly and daily wind gusts are able to be projected using wind gust simulation models. Although the wind gust simulation models were validated based on the independent dataset of historical observations, it is necessary to ascertain whether the models are suitable for future projection. This can be achieved through comparing hourly/daily wind gusts projected by downscaled GCM historical runs with observations over a comparable time period (1961–2000). Figure 6 shows quantile–quantile (Q–Q) plots of sorted hourly/daily wind gusts derived from downscaled GCM historical runs and observations in Windsor over the time period 1961–2000. If both datasets come from populations with a common distribution, the points should fall approximately along the diagonal reference line. From Fig. 6, it is clear that the data distributions of both datasets are similar, so it can be concluded that the methods used in the study are suitable for projecting or downscaling future hourly and daily wind gust information at station scale. The similar results for hourly/daily wind gust distributions between the downscaled GCM historical runs and observations were found for the rest of the selected stations, not shown.
Quantile–quantile plots of (left) hourly and (right) daily wind gust quantities derived from downscaled GCM historical runs vs observations over a comparatable time period (1961–2000) in Windsor. (The diagonal reference line suggests that both datasets come from populations with the same distribution.)
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
5. Results and discussion
a. Historical hourly and daily wind gust simulation and validation
To effectively evaluate the performance of the hourly and daily wind gust simulation models, four correctness levels—excellent, good, fair, and poor—were subjectively defined based on the absolute difference between observed and simulated hourly/daily wind gusts. An excellent simulation is defined when the absolute difference is less than or equal to 15% of the observation. A good simulation is defined when the simulation does not meet the excellent criterion but the absolute difference is less than or equal to 30% of the observation. A fair simulation is defined when the simulation does not meet either the excellent or the good criteria but the absolute difference is less than or equal to 45% of the observation. A poor simulation is defined when the absolute difference is greater than 45% of the observation. The results of the model performance evaluation for Windsor are summarized in Fig. 7. Similar results were also found for the rest of the selected cities. The model evaluation was undertaken for seven categories based on observed wind gusts greater than or equal to certain thresholds: 1) ≥28, 2) ≥40, 3) ≥50, 4) ≥60, 5) ≥70, 6) ≥80, and 7) ≥90 km h−1. Some of these thresholds (e.g., 70 and 90 km h−1) were selected to correspond with the criteria currently used by Environment Canada for issuing wind gust warnings (Environment Canada 2010). The threshold of 40 km h−1 was selected since a wind turbine delivers most of its power only once the wind blows faster than this threshold value (Moyer 2009). In addition, the threshold of 28 km h−1 was selected since it is the minimum value to record wind gust observations. It was observed that the proportion of the simulations that fell into the excellent and good correction categories was greater than the proportion that fell into the fair and poor levels. From Fig. 7, the percentage of excellent and good simulations among all seven wind gust categories ranges from 94% to 100% for hourly wind gusts and from 69% to 95% for daily wind gusts, for both model development and validation. There is not much difference between the development and validation of hourly/daily wind gust simulation models for all four correctness levels except for the excellent level. The percentage at the excellent level for the model validation is slightly less than for the model development, especially for severe wind gust events (e.g., ≥90 and ≥80 km h−1), partially due to rare cases of these events. For example, for the category with wind gusts ≥90 km h−1, the percentage at the excellent level for the hourly wind gust model development and validation is 65% and 55%, respectively; the corresponding percentage for the daily wind gust model development and validation is 50% and 45%. The results from wind gust simulation model development and validation suggest that the models can capture historical hourly/daily wind gust events.
Evaluation results of hourly and daily wind gust simulation model performance for model development and validation in Windsor.
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
As shown in Fig. 7, the hourly wind gust simulation models usually perform better than the daily simulation models. The major reason for this is the different measurement averaging periods for recording the hourly wind speeds and hourly/daily wind gusts. As described above, hourly wind speeds were reported by averaging speeds over a period of 2 min just prior to the hour. Hourly wind gusts represent the peak wind speeds during the 10-min period prior to the hour, while daily wind gusts represent the peak wind speeds during an entire 24-h period. As a result, the daily wind gust is usually greater than the hourly wind gust during a day. When pooling all 14 stations for the period 1994–2007, including all days with wind gust observations, about 39% of the daily wind gusts are greater than the 1-h maximum wind gust (the maximum wind gust of hourly observations during a day); about 57% of the days possess the same values for both gusts. This is reflected in relationships between hourly/daily wind gusts and wind speeds, as shown in Figs. 4 and 5. By comparing Figs. 4 and 5, it can be seen that daily wind gusts are distributed with more variation around the simulation lines than for hourly wind gusts; ultimately, the model simulations of daily wind gusts are generally worse than the hourly wind gust simulations.
b. Changes in frequency of future hourly and daily wind gust events
To more effectively present the results on changes in frequency of future hourly and daily wind gust events, the wind gust stations were divided into four regions, based on the magnitude of observed frequencies of hourly and daily extreme wind gust events. The regions defined here, as shown in Fig. 1, are not standard for Environment Canada’s forecast regions. When moving from Region I through to Region IV (i.e., from south to north) across the province of Ontario, the annual mean number of cases with hourly and daily extreme wind gust events is observed to become smaller and smaller (Table 2). For example, the annual mean number of days with wind gusts ≥80 km h−1 observed in the historical period is 6.3, 5.1, 3.2, and 1.4, respectively, for Regions I to IV.
Region-averaged annual mean number of hours and days observed with wind gust events ≥ 60, 70, 80, and 90 km h−1.
Projections of the future annual mean number of hours and days with wind gust event (greater than or equal to the thresholds) were analyzed for each of the four regions, as shown in Figs. 8 and 9. In addition, the region-averaged annual mean frequencies of historical observed wind gust events and the modeled events derived from downscaled GCM historical runs are presented in Figs. 8 and 9 as the baselines for future projections. It is noteworthy that for all regions and wind gust events studied, the number of hourly and daily wind gust events derived from downscaled GCM historical runs is substantially similar to those from historical observations. This implies that the wind gust simulation models and regression-based downscaling methods used in the study are suitable to project the occurrence frequency of future hourly and daily wind gust events.
Projected annual mean frequency of hourly wind gust events derived from the downscaled eight-GCM ensemble, summarized by region and city (six bars in each of the panels are in the order of O: historical observations over the period 1994–2007, H: downscaled GCM historical runs over the period 1961–2000, scenarios A2 and B1 over the period 2046–65, and scenarios A2 and B1 over the period 2081–2100).
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
As in Fig. 8, but for daily wind gust events.
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
It is immediately apparent from Figs. 8 and 9 that the province of Ontario could experience more wind gust events late this century. To more clearly present changes in the frequency of future wind gust events, region-averaged percentage increases for the events ≥28, ≥40, and ≥70 km h−1 from the current and past conditions are shown in Table 3. For wind gust events ≥80 and ≥90 km h−1, it is not appropriate to calculate the percentage increases due to the rare cases of severe wind gust events, especially in the northern areas (Regions III and IV) for hourly wind gust events. From Figs. 8 and 9 and Table 3, it can be seen that the magnitude of these increases for more severe wind gust events was generally projected to be greater. For example, in Region I, the annual mean frequency of future hourly wind gust events ≥90 km h−1 derived from the ensemble of eight downscaled A2 simulations is projected to be 2.4 h for the period 2046–65, which is about 70% higher than the observed average during the period 1994–2007. The corresponding frequency for future hourly wind gust events ≥70 km h−1 is projected to be 24 h, about 17% higher than the current and past historical conditions. Similarly, the annual mean frequency of future hourly wind gust events ≥40 km h−1 is projected to be 610 h, about 13% higher than for the current conditions. In addition, the projected annual mean frequency of future wind gust events derived from the A2 scenario differs from for the B1 scenario. The projected increases derived from a downscaled A2 scenario are usually slightly greater than those derived from a downscaled B1 scenario. The A2 scenario projects higher emissions and more future warming than does the B1 scenario. For more detailed differences, regarding interscenario and inter-GCM-model uncertainties, refer to section 5e.
Region-averaged percentage increases in the frequency of future daily and hourly wind gust events (≥28, ≥40, and ≥70 km h−1) from the current conditions, presented by the eight-GCM A2 and eight-GCM B1 ensembles.
As discussed above, the wind gust events that occurred in the study area were usually triggered by storm systems on two different scales: synoptic and local convective windstorms. Synoptic windstorms typically occur during the winter season, while local convective windstorms generally occur during the summer. For future projections, the question is: are there any differences in the change patterns of future hourly and daily wind gust events between the seasons? To answer this question, the projections of future hourly and daily wind gust events ≥80 km h−1 were analyzed by seasons (Figs. 10 and 11). From Figs. 10 and 11, it can be seen that the frequencies of observed hourly and daily wind gust events for all seasons except summer significantly decrease when moving from Region I through to Region IV. However, during the summer season, the regional differences in observed hourly and daily wind gust events are much smaller, especially for daily wind gust events. The frequencies of future hourly and daily wind gust events for all seasons are projected to increase from current and past conditions. For example, in Region I, the seasonal mean frequency of future hourly wind gust events ≥80 km h−1 derived from the ensemble of eight downscaled A2 simulations is projected to be 2.9, 0.5, 2.5, and 2.6 h for the period 2046–65, respectively, for spring, summer, autumn, and winter (the historical observations are 2.5, 0.2, 1.7, and 1.7 h). The corresponding frequency for future daily wind gust events is projected to be 2.3, 1.1, 1.3, and 2.6 days versus the historically observed conditions of 2.2, 0.9, 1.1, and 2.1 days. For individual cities, as shown in Figs. 10 and 11, the increase in values should be more or less around these regional averages. For example, in Windsor, the seasonal mean frequency of future hourly wind gust events (≥80 km h−1) is projected to be 4.5, 0.7, 2.8, and 3.8 h versus the historical observations of 3.1, 0.4, 2.2, and 2.6 h, respectively, for spring, summer, autumn, and winter. The corresponding frequency for future daily wind gust events (≥80 km h−1) is projected to be 2.8, 1.4, 1.5, and 2.4 days versus the historical observations of 2.6, 1.3, 1.2, and 1.9 days.
Projected seasonal mean frequency of hourly wind gust events ≥80 km h−1 derived from the downscaled eight-GCM ensemble, summarized by region and city (six bars in each of the panels are in the order of O, historical observations over the period 1994–2007; H, downscaled GCM historical runs over the period 1961–2000; scenarios A2 and B1 over the period 2046–65; and scenarios A2 and B1 over the period 2081–2100).
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
As in Fig. 10, but for daily wind gust events.
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
c. Contrast with regional climate model simulations
Following the projection of the frequency increases in future hourly and daily wind gust events using statistically downscaled GCM simulations, one question could be asked: can the similar changes be projected using RCM simulations? To answer this question, 3-hourly wind speed simulations under the IPCC SRES A2 emission scenario from five RCMs for two time periods (1968–2000 and 2038–70) were applied to daily wind gust simulation models to project frequency changes of future daily wind gust events. These five RCMs are the Canadian Regional Climate Model–Canadian Centre for Climate Modelling and Analysis (CCCma) Coupled General Circulation Model version 3 (CRCM–CGCM3), the Hadley Regional Model 3–third climate configuration of the Met Office Unified Model (HRM3–HadCM3), version 3 of the Regional Climate Model–CGCM3 (RCM3–CGCM3), RCM3–Geophysical Fluid Dynamics Laboratory Climate Model version 2.1 (RCM3–GFDL CM2.1, and the Weather Research and Forecasting model–Community Climate System Model version 3 (WRFG-CCSM3). The RCM wind speed simulations at a grid point nearest to the stations shown in Fig. 1 were used in the analysis. For each of the days within RCM simulation periods, the maximum wind speed selected from the eight RCM simulations per day was applied to daily wind gust simulation models for projecting future daily wind gusts. The relative changes for the wind gust events from current and past conditions are calculated by the difference in annual mean frequency between the RCM future projection (2038–70) and the historical run (1968–2000) divided by the observed annual mean frequency. Across-study-area mean relative changes and spatial variations in frequency of future daily wind gust events (≥40, ≥60, ≥70, ≥80, and ≥90 km h−1) derived from the five-RCM A2 ensemble and statistically downscaled eight-GCM A2 ensemble are shown in Table 4. The results clearly indicate that the frequency of future daily wind gust events projected from the downscaled eight-GCM ensemble increases consistently across the study area. However, the corresponding frequency projected from a five-RCM ensemble increases at some stations and decreases at others. For example, for daily wind gust events ≥80 km h−1, the relative changes projected by the five-RCM ensemble spatially vary from −35% to +15%; the corresponding changes for daily wind gust events ≥90 km h−1 vary from −40% to +45%. This implies that RCM simulations may not be suitable to be directly used for climate change impact analyses at local or station scales, while statistical downscaling techniques may be necessarily employed to further remove the RCM bias.
Across-study-area mean percentage changes and spatial variations in frequency of future daily wind gust events derived from the downscaled eight-GCM A2 ensemble and the raw five-RCM A2 ensemble.
d. Historical evidence to support future projections
Following projections of the frequency increases in future wind gust events, another question should be asked: is there any evidence in the historical period to support future projections? To answer this question, the wind gust trend analyses were evaluated using daily observed wind gust records and other variables (temperature and sea level air pressure). Two kinds of wind gust trend analyses were evaluated: the speed of daily wind gust events ≥50 km h−1 versus the climatological 1) daily temperature anomaly and 2) daily pressure anomaly. In addition, the frequency of daily wind gust events ≥90 km h−1 versus both daily temperature and pressure anomalies was analyzed. If these indicators increase as the daily temperature anomaly increases and/or as the daily pressure anomaly decreases in the historical period, then we can conclude that these are evidence to support projections of frequency increases in future wind gust events since temperature increases and pressure decreases under future changing climate (Meehl et al. 2007).
The results of the wind gust trend analyses on the speed of gust events ≥50 km h−1 versus the climatological daily temperature and pressure anomalies over the historical period in Windsor are illustrated in Fig. 12. It can be seen that the speed of the wind gusts significantly increased as the daily temperature anomaly increased and the daily pressure anomaly decreased over the past 44 years in Windsor. For example, in the summer season, for each daily temperature anomaly increase of 1°C, the speed of wind gust events ≥50 km h−1 increases by 0.64 km h−1; the corresponding wind gust speed increases by 0.33 km h−1 for each daily pressure anomaly decrease of 1 hPa. Similar results are discovered for other stations studied as well; however, they are not displayed owing to space limitations. To demonstrate more detailed information for these trend analyses on daily bases, the scatterplots were analyzed to show the relationships between the daily wind gust speed versus daily temperature and pressure anomalies. These relationships are statistically significant, most of which possess a p value < 0.01. For example, the scatterplot for the relationships between daily wind gust speed and daily pressure anomalies in spring at Windsor is illustrated in Fig. 13.
Relationships between the speed of observed daily wind gust events ≥50 km h−1 vs (left) the observed climatological daily temperature anomaly and (right) climatological daily pressure anomaly in Windsor over the period 1964–2007.
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
Scatterplot of relationships between the speed of observed daily wind gust events ≥50 km h−1 vs climatological daily pressure anomaly in Windsor over the period 1964–2007.
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
The relationships between frequency of daily wind gust events ≥90 km h−1 and daily temperature/pressure anomalies are illustrated in Fig. 14 as an ensemble of all 14 stations. Most of the daily wind gust events (66%–74% varying across seasons) occurred with positive daily temperature anomalies and negative daily pressure anomalies over the historical period. Some of the events (20%–28% across seasons) are associated with negative daily temperature anomalies and negative daily pressure anomalies. Very rare events (1%–4%) are related to negative daily temperature anomalies and positive daily pressure anomalies. These results imply that there is evidence to support the projections of increases in the frequency of future wind gust events. From raw GCM outputs, as indicated in the 2007 IPCC report (Meehl et al. 2007) and downscaled future climate data derived from this study, future temperatures could be higher than, and future pressure could be lower than, the past and current conditions. In addition, the relationships between temperature/pressure and wind gust events are likely to be similar under a future, warmer, climate since the weather conditions that currently favor wind gust events will be similar in the future. Consequently, a possible reason for the model-projected increases in frequency of future hourly and daily wind gust events over the province of Ontario could be that, as temperature increases and pressure decreases under a future changing climate, both synoptic and local convective windstorm systems could occur more frequently in the study area late this century. This conclusion could be supported by a previous study (i.e., Lambert and Fyfe 2006) that suggests that the number of intense midlatitude cyclones, projected by the GCMs participating in the IPCC Fourth Assessment Report (AR4) diagnostic exercise under the A2 and B1 scenarios, would increase by the periods 2046–65 and 2081–2100.
Relationship between observed daily wind gust events ≥90 km h−1 vs climatological daily temperature and daily pressure anomalies (ensemble of all 14 cities over the province of Ontario studied).
Citation: Journal of Climate 25, 9; 10.1175/JCLI-D-11-00198.1
e. Uncertainty of the study
Considerable effort was made in this study to transfer GCM-scale wind data to station-scale climate information using regression-based downscaling transfer functions. Through the downscaling process, the GCM model bias was removed using approximately 40-yr historical relationships between regional-scale predictors and station-scale observed daily/hourly wind data (Katz 2002). As a result, the quality of future GCM climate projections, following downscaling, was much improved. For example, the data distribution (including extreme events) of downscaled GCM historical runs was similar to that of observations over a comparable time period (1961–2000).
However, conclusions made in this study about the impacts of climate change on future daily/hourly wind gust events still rely on GCM scenarios/projections and, as a result, there is corresponding uncertainty about the study findings. One of the most important sources of uncertainty in climate change impact studies comes from GCM modeling (Katz 2002). Because of model resolution and complexity, GCM models must inevitably omit some factors that affect climate; in turn, GCM models are unable to resolve subgrid-scale processes and generate uncertainty through model parameterizations. To quantitatively assess inter-GCM-model and interscenario uncertainties of future wind gust projections, we have analyzed the region-averaged absolute difference among eight selected GCM models for both A2 and B1 scenarios as well as the region-averaged absolute difference between both scenarios. For inter-GCM-model uncertainty calculations, there are 28 pairs in total obtained from the eight GCM models included in the analysis. The absolute difference used in the analysis is to avoid negative values canceling out positive values. As shown in Table 5, overall, the inter-GCM-model uncertainty of the percentage increases in frequency of future daily/hourly wind gust projections are similar to or greater than the interscenario uncertainties. Comparing Table 5 with Table 3, it can be seen that for wind gust events <70 km h−1, both inter-GCM-model and interscenario uncertainties, overall, are less than the projected percentage increases in the frequency of future daily/hourly wind gust events. Specifically, on average over the four regions for both future time periods and two scenarios together, the projected percentage increases in the frequencies of future daily/hourly wind gust events ≥28 and ≥40 km h−1 are about 90%–100% and 60%–80% greater than the inter-GCM-model and interscenario uncertainties, respectively. For wind gust events ≥70 km h−1, the corresponding projected percentage increases are about 25%–35% greater than interscenario uncertainties and are generally similar to inter-GCM-model uncertainties. However, for wind gust events ≥80 and ≥90 km h−1, the uncertainties generally are greater than the projected percentage increases, not shown since it is not appropriate to calculate percentage increases due to rare cases in the study area, especially in northern Ontario.
Region-averaged inter-GCM-model and interscenario uncertainties of percentage increases in the frequency of future daily and hourly wind gust events (≥28, ≥40, and ≥70 km h−1) from the current conditions.
6. Conclusions
The overarching purpose of this study is to assess possible impacts of climate change on future hourly and daily wind gust events over the province of Ontario, Canada, based on historical wind gust simulation models as well as downscaled future hourly wind climate data. The hourly and daily wind gusts were simulated using their wind gust factors multiplied with hourly wind speed and 1-h maximum wind speed during a day, respectively. The study results on the development and validation of wind gust simulation models show that the methods used in the study are useful for verifying historical wind gust events, as well for projecting changes in the frequency of occurrence of future hourly and daily wind gust events.
The projected results clearly show that the province of Ontario could possibly experience more wind gust events later this century than in the past. The magnitude of these increases is generally projected to be greater with more severe wind gust events, but also with greater uncertainty between selected GCM models and climate change scenarios. For wind gust events <70 km h−1, both inter-GCM-model and interscenario uncertainties are less than the projected percentage increases in the frequency of future daily/hourly wind gust events. For wind gust events ≥70 km h−1, the interscenario uncertainties are less than and the inter-GCM-model uncertainties and are generally similar to the projected percentage increases. However, for wind gust events ≥80 and ≥90 km h−1, the uncertainties are greater than the projected percentage increases due to rare cases in the study area, which are not appropriate for calculating percentage increases.
Across the province of Ontario, as shown in Table 3, the annual mean frequency of future hourly wind gust events ≥28, ≥40, and ≥70 km h−1 for the period 2081–2100 derived from the ensemble of eight downscaled IPCC SRES A2 simulations is projected to be about 10%–15%, 10%–20%, and 20%–40% greater than the observed conditions during the period 1994–2007, respectively. The corresponding relative increase for future daily wind gust events is projected to be less than 10%, about 10%, and 15%–25%. In addition, the projected increases in future hourly–daily wind gust events derived from downscaled A2 simulations are usually slightly greater than those derived from downscaled B1 simulations. The implications of these increases should be taken into consideration and integrated into policies and planning for adaptation strategies, including measures to incorporate climate change into engineering infrastructure design standards and disaster risk reduction measures.
Acknowledgments
The authors would like to thank two anonymous reviewers for providing detailed comments that significantly improved the original manuscript.
REFERENCES
Ahmed, S., 1994: Gust factors within frontal depressions at Rio, Greece. Meteor. Appl., 1, 205–208.
Beniston, M., and Coauthors, 2007: Future extreme events in European climate: An exploration of regional climate model projections. Climatic Change, 81, 71–95, doi:10.1007/s10584-006-9226-z.
Bernstein, L., and Coauthors, Eds., 2007: Climate Change 2007: Synthesis Report. Cambridge University Press, 104 pp.
Cechet, R. P., and L. A. Sanabria, 2010: Extreme-value time-series analysis of Australian Region A gust wind speeds to examine instrument bias. IOP Conf. Ser.: Earth Environ. Sci., 11, 012016, doi:10.1088/1755-1315/11/1/012016.
Changnon, S. A., 1980: Climatology of high damaging winds in the Midwest. Investigation Rep. 95, Illinois State Water Survey, Champaign, IL, 54 pp.
Changnon, S. A., 2009: Temporal and spatial distributions of wind storm damages in the United States. Climatic Change, 94, 473–482, doi:10.1007/s10584-008-9518-6.
Cheng, C. S., G. Li, Q. Li, and H. Auld, 2008: Statistical downscaling of hourly and daily climate scenarios for various meteorological variables in south-central Canada. Theor. Appl. Climatol., 91, 129–147, doi:10.1007/s00704-007-0302-8.
Cheng, C. S., G. Li, Q. Li, and H. Auld, 2010: A synoptic weather typing approach to simulate daily rainfall and extremes in Ontario, Canada: Potential for climate change projections. J. Appl. Meteor. Climatol., 49, 845–866.
Cvitan, L., 2003: Determining wind gusts using mean hourly speed. Geofizika, 20, 63–74.
Davis, F. K., and H. Newstein, 1968: The variation of gust factors with mean wind speed and with height. J. Appl. Meteor., 7, 372–378.
Dore, M. H. I., 2003: Forecasting the conditional probabilities of natural disasters in Canada as a guide for disaster preparedness. Nat. Hazards, 28, 249–269.
Environment Canada, 1977: MAMOBS: Manual of surface weather observations. 7th ed. Meteorological Service of Canada, Environment Canada, Toronto, ON, Canada, 386 pp.
Environment Canada, cited 2010: Public alerting criteria. [Available online at http://www.ec.gc.ca/meteo-weather/default.asp?lang=En&n=D9553AB5-1.]
Graybeal, D. Y., 2006: Relationships among daily mean and maximum wind speeds, with application to data quality assurance. Int. J. Climatol., 26, 29–43.
Jungo, P., S. Goyette, and M. Beniston, 2002: Daily wind gust speed probabilities over Switzerland according to three types of synoptic circulation. Int. J. Climatol., 22, 485–499.
Katz, R. W., 2002: Techniques for estimating uncertainty in climate change scenarios and impact studies. Climate Res., 20, 167–185.
Lambert, S. J., and J. C. Fyfe, 2006: Changes in winter cyclone frequencies and strengths simulated in enhanced greenhouse warming experiments: Results from the models participating in the IPCC diagnostic exercise. Climate Dyn., 26, 713–728.
Lopes, A., S. Oliveira, M. Fragoso, J. A. Andrade, and P. Pedro, 2009: Wind risk assessment in urban environments: The case of falling trees during windstorm events in Lisbon. Bioclimatology and Natural Hazards, K. Strelcová et al., Eds., Springer, 55–74.
Meehl, G. A., and Coauthors, 2007: Global climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 747–845.
Mitsuta, Y., and O. Tsukamoto, 1989: Studies on spatial structure of wind gust. J. Appl. Meteor., 28, 1155–1160.
Moyer, M., 2009: The way the wind blows. Sci. Amer., 301 (October), 27–28.
Paulsen, B. M., and J. L. Schroeder, 2005: An examination of tropical and extratropical gust factors and the associated wind speed histograms. J. Appl. Meteor., 44, 270–280.
PCMDI, cited 2008: WCRP CMIP3 multi-model dataset. [Available online at http://www-pcmdi.llnl.gov/ipcc/about_ipcc.php.]
Pinto, J. G., C. P. Neuhaus, G. C. Leckebusch, M. Reyers, and M. Kerschgens, 2010: Estimation of wind storm impacts over western Germany under future climate conditions using a statistical–dynamical downscaling approach. Tellus, 62A, 188–201.
Pryor, S. C., J. T. Schoof, and R. J. Barthelmie, 2005: Climate change impacts on wind speeds and wind energy density in northern Europe: empirical downscaling of multiple AOGCMs. Climate Res., 29, 183–198.
Pryor, S. C., J. T. Schoof, and R. J. Barthelmie, 2006: Winds of change?: Projections of near-surface winds under climate change scenarios. Geophys. Res. Lett., 33, L11702, doi:10.1029/2006GL026000.
Ren, D., 2010: Effects of global warming on wind energy availability. J. Renewable Sustainable Energy, 2, 052301, doi:10.1063/1.3486072.
Sanabria, L. A., and R. P. Cechet, 2010: Severe wind hazard using dynamically downscaled climate simulations. IOP Conf. Ser.: Earth Environ. Sci., 11, 012021, doi:10.1088/1755-1315/11/1/012021.
Schwierz, C., P. Köllner-Heck, E. Z. Mutter, D. N. Bresch, P.-L. Vidale, M. Wild, and C. Schär, 2010: Modelling European winter wind storm losses in current and future climate. Climatic Change, 101, 485–514, doi:10.1007/s10584-009-9712-1.
Shen, S. S. P., P. Dzikowski, G. Li, and D. Griffith, 2001: Interpolation of 1961–97 daily temperature and precipitation data onto Alberta polygons of ecodistrict and soil landscapes of Canada. J. Appl. Meteor., 40, 2162–2177.
Sparks, P. R., and Z. Huang, 2001: Gust factors and surface-to-gradient wind-speed ratios in tropical cyclones. J. Wind Eng. Ind. Aerodyn., 89, 1047–1058.
Weggel, J. R., 1999: Maximum daily wind gusts related to mean daily wind speed. J. Struct. Eng., 125, 465–468.
