Chapter 4 References
This national wind energy resource atlas is a synthesis and update of regional resource assessments that were performed in 1979 and 1980 by the Pacific Northwest Laboratory and other contractors. A list of the regional atlas titles is given in Table A-l; a map of the United States identifying the regional divisions is shown in Map A-l. This appendix summarizes the data sources and analysis techniques used in preparing the wind resource estimates.
A wide variety of data types and analysis techniques were utilized in performing the regional wind energy resource assessments and ultimately producing this wind energy atlas. The techniques developed by the Pacific Northwest Laboratory in the Northwest regional prototype provided the basis for the regional assessment con tractors to follow. However, the contractors were given some flexibility in refining or modifying the techniques where necessary or where a revised technique would improve the resource assessment.
The types of data utilized and the techniques employed varied somewhat from region to region. For example, the data and methods employed in mountainous regions were, by necessity, more varied than those in regions of mostly plains and little vertical relief. In mountainous areas, upper-air wind data were utilized in estimating the wind resource over exposed mountain summits and ridge crests where conventional surface data are scarce or not available. Moreover, more extensive use of qualitative indicators of the wind resource was made in mountainous areas than in nonmountainous areas, since greater variability of the wind resource usually occurs in mountainous areas.
These qualitative indicators included the identification of certain combinations of topographical and meteorological features, areas containing eolian landforms, and areas with flagged trees. Fairly intensive surveys of wind deformed vegetation were conducted in the Northeast and Southwest regions, the former using 30 newspaper surveys to solicit information from the public and the latter conducting aerial and ground surveillance over large areas to identify areas of wind-flagged trees.
Data types and techniques utilized in coastal areas varied from those utilized inland. Information on the coastal wind resource in many areas included data from ship observations within specified coastal marine areas. Techniques employed to estimate the variation of the wind resource inland from the shoreline considered the direction of the prevailing strong winds, alignment of the coastline, topography and vegetation conditions onshore, along with information on the wind resource at coastal and inland stations in the coastal region.
Techniques used in the low-latitude regions (e.g., Hawaii and the Pacific Islands and Puerto Rico and the Virgin Islands) varied from those used in the contiguous United States and Alaska. Revised methods were developed to estimate the wind resource over ridge crests and mountain summits in the regions, as it was recognized that the mean wind profile structure in the trade-wind regimes of Hawaii and Puerto Rico is considerably different than that of the contiguous United States.
Nationwide, the National Climatic Data Center (NCDC) accounted for about 70% of the approximately 3200 surface stations evaluated in the wind resource assessments. However, the types and sources of surface wind data used in the resource assessments varied considerably among the various regions. For example, some regions used considerable amounts of Forest Service data (e.g., the Northwest and Southwest regions), whereas other regions with large areas of forested terrain used little or no Forest Service data. Data from power plant sites made significant contributions to the data base in the Northeast, Southeast, and Great Lakes regions. Canadian data were useful in assessing the wind resource within the five regions bordering Canada. In the Southern Rocky Mountain region, an intensive effort was made to identify and utilize data from various other sources, which accounted for 50% of the data in this region, considerably more than in any other region.
This appendix provides a detailed description of the various data analysis and assessment methodologies employed in the regional wind resource assessments and, in essence, the synthesized national assessment. Tables have been prepared to summarize the various types of information, data, and techniques used in each of the regional assessments. Methods used to evaluate the certainty of the resource estimates and to assess the percentage land area with a given wind resource are also described in this chapter.
The surface wind data on which the resource assessments were based were obtained from a variety of sources: the NCDC, the U.S. Forest Service, university research projects, existing and proposed power plant sites, and various other sources. Table A-2 describes the principal sources of data. Many other wind data sets available from university research projects, U.S. Department of Energy candidate wind turbine sites, wind energy field studies, and various other government and private organizations were also identified and used in the assessments where appropriate.
Table A-3 lists the number of stations with wind data from each source identified in the regional wind resource assessments. Table A-4 lists the number of stations identified in each of the 12 regions, from all sources.
The NCDC accounted for about 60% of all stations with wind data identified from all sources. In all regions except the Northwest and Southwest, the greatest percentage of the stations identified were from the NCDC. Forest Service stations accounted for about 26% of the total from all sources, although most of these data were generally of very limited value. The regions with the greatest number of stations from all sources were the Northwest and Southwest; however, Forest Service stations accounted for 54% (685) of the stations in the Northwest and 46% (443) of the stations in the Southwest. The three western regions, i.e., Northwest, Southwest, and Southern Rocky Mountain, accounted for 78% of the Forest Service stations in the United States.
A substantial number of wind data locations exist that are not reflected in Tables A-3 and A-4. Data from university research projects, private organizations, and government agencies other than the NCDC frequently exist in a format that is not suitable for an assessment of this scope, either as unreduced strip-chart records or as partial compilations of hourly data records collected for very specialized purposes. In most populated areas, adequate summarized data from the NCDC or other sources were usually available. In these areas, little or no effort was made to identify additional data.
Review of Table A-4 indicates the large quantity of wind data identified in the assessments; however, not all of these data need to, or should, be used. Screening procedures were developed to select stations with the most useful data and to eliminate stations that would not significantly contribute information on the distribution of the wind resource.
In general, wind data in summarized or digitized format were chosen in preference to unsummarized data. For selected stations where both summarized and digitized data were available, the digitized data were used to prepare summaries that more fully characterized the wind resource than existing summaries. As previously described, PNL processed the NCDC-digitized data to produce extensive summaries of the wind characteristics. In contrast to many summarizations routinely available from NCDC, PNL's analyses examined the wind record only for periods of constant anemometer height, location, and observation frequency. Thus, the PNL summaries of the NCDC-digitized data were usually chosen in preference to the conventional summaries available from NCDC. However, there were still many stations with the conventional NCDC summaries for which only limited or no PNL summaries of digitized data were available.
Because many NCDC stations had several different types of summarized wind data covering various time periods, criteria were established to choose the best one or two summaries for those stations with several summaries. Using the NCDC Index to Summarized Wind Data (Changery et al. 1977) to identify the available summaries, the Index - Original Surface Weather Records to determine the frequency of observations, and the National Wind Data Index (Changery 1978) to determine anemometer height histories, summaries were selected that had:
In areas with a high density of stations (such as many of the large metropolitan areas) and a considerable amount of digitized or summarized data, only those stations appearing to have the best exposure and longest periods of unchanged anemometer height and location were usually selected. Conversely, many stations from smaller towns and in more remote locations had only one or two summaries, often in undesirable formats, to choose from. Frequently, anemometer heights were unknown for the summary period and wind observations were limited to daytime hours of operation at these stations.
For those stations with limited or no summarized or digitized wind data, unsummarized data were screened. Unsummarized data may take one of several forms, including WBANs (i.e., Weather Bureau, Air Force and Navy standard format reports), wind records, triple registers, and synoptic records. (Some of the forms, especially in the late 1930s and 1940s, occasionally included monthly average wind speeds.) Information on the type and frequency of observations, anemometer height and exposure, and station location were examined to determine the suitability of the unsummarized data for wind resource assessment. Some stations were eliminated from consideration. For many stations, at least one year of records was identified for evaluating the wind data.
Nationwide, the NCDC accounted for about 70% of the approximately 3200 stations evaluated in the wind resource assessments (see Table A-3). Most of the NCDC stations identified and screened with digitized or summarized data were eventually retained for potential use in the assessments. About 30% of the NCDC stations with unsummarized data were selected for further evaluation.
Forest Service stations were screened largely on the basis of the number of observations, computed seasonal average wind speeds and powers, and location. Only 13% of these stations were ultimately retained to contribute information on the wind resource. Even most of those retained were of quite limited use, because of the once daily observation. Nevertheless, over 225 Forest Service stations provided much of the only quantitative surface wind data for many remote areas of the United States.
A high percentage of the data identified and screened from power plants and other sources were ultimately retained for evaluation.
Table A-4 showed considerable variation among the regions in the percent of stations retained for evaluation in the wind resource analyses. The regions with lowest percentage of stations retained were the Northwest, Southwest, and Southeast. The low percentages for the Northwest and Southwest are largely a result of the relatively large number of Forest Service stations in these two regions and the small percentage of these stations retained. In the Southeast, NCDC digitized data accounted for almost 70% of the stations ultimately retained for the analysis, considerably greater than the percentage for any other region. Because of the wide coverage of the digitized data and the predominantly low wind resource throughout most of this region, there was little need to retain and evaluate much of the additional data identified in order to further characterize the geographical distribution of the wind resource.
Because of the sparseness of wind data stations in the Puerto Rico/Virgin Island region, all wind data obtained were used in the analysis. Over 70% of the stations identified in the the Hawaii and Pacific Islands region were retained for the analysis.
Table A-6 lists the number of stations with wind data evaluated from each source from all 12 regions. Some notable differences among the regions in the types and sources of data utilized are apparent. Except for the Southern Rocky Mountain region, the NCDC was the primary source of most of the stations utilized. In three regions (the East Central, Southeast and South Central) the NCDC accounted for over 90% of the stations.
Forest Service stations accounted for about 30% of the stations utilized in the Southwest and 14% of the stations in the Northwest. These two regions accounted for over 70% of the Forest Service stations in the United States utilized in the analysis.
Wind data from power plant sites utilized in the analysis were most abundant in the Northeast and Great Lakes regions. However, these data accounted for less than 10% of the station data in these regions.
Data from numerous Canadian stations were utilized in the analysis in the five regions bordering Canada. Data from only three Mexican stations were useful in the three regions bordering Mexico.
Wind data from other sources accounted for 50% of the station data in the Southern Rocky Mountain region, 34% in the Hawaii and Pacific Islands region, and 25% in the Great Lakes region. Nationally, about 15% of the station data used in the analysis were obtained from other sources.
Two other types of wind data evaluated, in addition to the surface land-station data described above, were coastal marine area data and upper-air data. Table A-7 presents the extent to which these two types of data were applied in each of the regional atlases. An "X" means that the data were evaluated and applied extensively in estimating the wind resource. An "L" means that some data of this type were evaluated but applied only to a relatively small area or used only in a limited way. For example, coastal marine area data were used extensively in the wind resource assessments of three regions - the Northwest, Hawaii and Pacific Islands, and Puerto Rico and the Virgin Islands. For the Northwest, very few land-surface stations with good exposure were available for estimating the coastal wind resource. Therefore, estimates of the coastal wind resource were based primarily on the coastal marine area data, supplemented where possible by land stations. For Hawaii the coastal wind resource estimates were based primarily on land stations, whereas for the Pacific Islands (e.g., Midway, Wake, Johnston Islands, the Mariannas, the Marshalls, and Guam) open ocean wind power classes were presented along with wind power classes based on land stations for individual islands where available. For Puerto Rico and the Virgin Islands, there were few exposed coastal stations, so extensive use was made of the coastal marine data.
For six other regions, coastal marine data were evaluated but used primarily to supplement the existing station data. For these regions, the coastal wind resource estimates were based largely on the existing station data, although limited use was made of the coastal marine data. In some coastal areas of the United States, wind data were available from offshore "fixed" stations. These data were used instead of the coastal marine (ship observations) data.
Upper-air wind data were used in estimating the wind resource at mountain summit and ridge crest elevations, where existing surface station data are sparse. Extensive use of upper-air wind data was made in eight regions (see Table A- 7). In these regions, mountainous terrain covers a substantial part of the region. In three regions, mountainous terrain represents only a small fraction of the region's area. In one region, the Great Lakes, upper-air data were not needed as no areas of mountainous terrain (local relief >1000 ft) were identified.
Several time scales are encountered in the following discussions of the wind resource: annual, seasonal, monthly, and diurnal. Annual mean values are generally based on an average of the one- or three-hourly observations of wind speed or power in the period of record. However, a complete calendar year's data (covering January 1 to December 31) is used for calculating individual yearly means. At stations with less than 24 hourly observations or 8 three-hourly observations per day, the values are only representative of the times of day for which the data were taken. These values were used only in the absence of other suitable wind data for an area.
The four seasons are defined as:
The phrase "seasonal trends" refers to the change in monthly mean values over the course of the four seasons.
Monthly mean values are based on as many hours of data as are available for that month in each year of the period of record.
The daily or diurnal cycles of variation in the hourly mean wind power or speed are referenced to local standard time on a 24-hour clock. Midnight is both 00 and 24.
For the purpose of mapping the geographical variation of the wind resource, wind power density was chosen in preference to wind speed because the power density value combines the effect of the distribution of wind speeds and the dependence of the power density on air density and on wind speed. Quantitative wind data in digitized, summarized, and unsummarized forms were evaluated for mean wind power density, which is calculated as described below for each type of data.
For stations with l-hour and 3-hour digitized data, the average wind power density (Watts/m2) in a vertical plane perpendicular to the wind direction was calculated from
n = the number of observations in the averaging period
i = the air density (kg/m3)
Vi = the wind speed (m/s) at the ith observation time.
The air density was computed from the measured temperature and station pressure or was estimated by correcting standard air density for station elevation. Air density () was calculated from measured temperature (T) and station pressure (P) by
where R is a gas constant. If temperature or station pressure was not available, air density was estimated as a function only of station elevation (Z) by
which approximates the U.S. Standard Atmosphere profile for air density
For stations with wind summaries, was calculated
= the mean air density
The mean air density was usually calculated using Equation (3) above. A few
of the regional atlases incorporated the seasonal variation of air density
in the calculation of wind power.
In those cases for which unsummarized wind data were assessed, the seasonal
and annual average speeds, V, for most stations were estimated from a visual
examination of one year's original weather records. Some station records,
especially Form 1001A during the 1930s and 1940s, frequently reported monthly
mean wind speeds computed from all available hourly observations. In such
cases, the seasonal and annual mean wind speeds were computed based on the
reported monthly mean speeds. The wind power density,
, was then estimated by assuming that the speed
frequency distribution followed a Rayleigh distribution
In certain cases, a more objective estimate of seasonal and annual average
wind power was obtained for selected stations. To circumvent the laborious
task of entering all the hourly (or 3-hourly) observations from a station's
weather records, various techniques were employed. For example, for the East
Central region, average seasonal wind speeds were computed based on every
third observation over a one-month period from each season. This was done
only for those stations that indicated high wind energy potential based on
visual examination and that recorded at least eight observations per day
during the diurnal cycle. Wind power was computed by Equation (5). For the
North Central region, the average seasonal and annual cubed speed,
,was estimated by manually averaging all
V3every third day for one year for selected stations with unsummarized
data. Wind power density for these stations was then estimated by:
For Alaska and Puerto Rico and the Virgin Islands, the wind power density
was estimated from the mean wind speeds by assuming Weibull speed frequency
distributions other than the Rayleigh (see the respective regional atlases
For the Northeast and Southeast regions, seasonal and annual wind speed frequency
distributions were constructed manually from the original surface weather
records of selected stations by scanning 1 year's data and entering every
fourth hour into the distribution. The wind power density
was then computed by equation (4).
Table A-7 showed those regions of the United States
where the standard methods developed by PNL to calculate or estimate wind
power density were supplemented or refined by other methods. In practically
all cases, these other methods were used only to calculate or estimate wind
power density for stations with unsummarized data.
The anemometer height above the surface rarely was at either the 10-m (33-ft)
or 50-m (164-ft) reference levels chosen for the presentation of the wind
resource. A power law was used to adjust the long-term mean wind speed or
power density to the reference level:
c = the number of wind speed classes
fj = frequency of occurrence of winds in the jth class
Vj = the median wind speed of the jth class.
The visual examination of the data provided a crude but fast and inexpensive means of making a rough estimate of a station's seasonal and annual mean wind speeds. Generally, the best that could be achieved by this subjective technique was to estimate the mean wind speeds as light (<10 mph), moderate (10 to 12 mph), or strong (>12 mph). Wind speeds on record forms were usually in either mph or knots. Nevertheless, this subjective estimate of the mean wind speeds often provided the only information on the wind resource in many areas of the United States.
For Hawaii and the Pacific Islands, the NCDC unsummarized original weather records were digitized and summarized for selected periods (periods were normally 12 months and ranged from a few months up to 2 years).
which approximates the U.S. Standard Atmosphere profile for air density (NOAA 1976).
For stations with wind summaries, was calculated from
= the mean air density
The mean air density was usually calculated using Equation (3) above. A few of the regional atlases incorporated the seasonal variation of air density in the calculation of wind power.
In those cases for which unsummarized wind data were assessed, the seasonal and annual average speeds, V, for most stations were estimated from a visual examination of one year's original weather records. Some station records, especially Form 1001A during the 1930s and 1940s, frequently reported monthly mean wind speeds computed from all available hourly observations. In such cases, the seasonal and annual mean wind speeds were computed based on the reported monthly mean speeds. The wind power density, , was then estimated by assuming that the speed frequency distribution followed a Rayleigh distribution (Cliff 1977):
In certain cases, a more objective estimate of seasonal and annual average wind power was obtained for selected stations. To circumvent the laborious task of entering all the hourly (or 3-hourly) observations from a station's weather records, various techniques were employed. For example, for the East Central region, average seasonal wind speeds were computed based on every third observation over a one-month period from each season. This was done only for those stations that indicated high wind energy potential based on visual examination and that recorded at least eight observations per day during the diurnal cycle. Wind power was computed by Equation (5). For the North Central region, the average seasonal and annual cubed speed, ,was estimated by manually averaging all V3every third day for one year for selected stations with unsummarized data. Wind power density for these stations was then estimated by:
For Alaska and Puerto Rico and the Virgin Islands, the wind power density was estimated from the mean wind speeds by assuming Weibull speed frequency distributions other than the Rayleigh (see the respective regional atlases for details).
For the Northeast and Southeast regions, seasonal and annual wind speed frequency distributions were constructed manually from the original surface weather records of selected stations by scanning 1 year's data and entering every fourth hour into the distribution. The wind power density was then computed by equation (4).
Table A-7 showed those regions of the United States where the standard methods developed by PNL to calculate or estimate wind power density were supplemented or refined by other methods. In practically all cases, these other methods were used only to calculate or estimate wind power density for stations with unsummarized data.
The anemometer height above the surface rarely was at either the 10-m (33-ft) or 50-m (164-ft) reference levels chosen for the presentation of the wind resource. A power law was used to adjust the long-term mean wind speed or power density to the reference level:
and = the mean wind speed or wind power density at heights Za,r (the anemometer and reference level, respectively) = the power law exponent.
An examination of long-term mean wind speeds at airport locations at which the anemometer height was changed and at tower sites with multiple levels of anemometry indicated an ~ 1/7 to be widely applicable to low surface roughness and well exposed sites from which conventional NCDC data are available (Elliott 1979a). Thus, an of 1/7 was used to adjust mean wind speed and wind power density to 10-m (33 ft) and 50-m (164 ft) reference levels at most stations with only one level of data. For a few stations for which the anemometer height was unknown, a height of 10 m (33 ft) was assumed.
A few sites had two (or more) levels of data appropriate for extrapolating to the 10-m (33 ft) and 50-m (164 ft) levels. At these sites, the value calculated from the mean wind speed and/or power densities at the measurement heights was used to adjust the wind speed and power density to the 10-m (33 ft) and 50-m (164 ft) levels.
For the Northeast and Southeast regions, an adjustment factor based on appropriate roughness values was used for a few sites located in towns or surrounded by wooded areas having surface roughness greater than that at airport locations. The adjustment of wind data of stations having this characteristic, to the 10-m (33-ft) and 50-m (164-ft) reference levels, was accomplished using the following relationship:
where and are the same as in Equation (7) and the Ba,r values are adjustment factors determined from a comparison of log-law wind speeds using appropriate roughness values for various surface environments (Wegley et al. 1980).
Since existing surface data from mountain summits and ridge crests were very sparse, upper-air wind data were identified for potential use in estimating the free-air wind speeds at mountain summit and ridge crest elevations in mountainous areas. An earlier study by Wahl (1966) had shown a strong correlation between mountain-top and free-air wind speeds.
A method was developed to delineate mountainous areas in the United States and to determine representative elevations of mountain summits and ridge crests in the mountainous areas. Mountainous areas, consisting of prominent mountain summits and/or ridge crests where the local relief (over a unit square 6 mi across) exceeds 300 m (1000 ft), were delineated using maps of classes of land-surface form (Hammond 1964) and topographic maps.
Tablelands with canyons or valleys over 300 m (1000 ft) deep, hilly terrain areas with relief less than 300 m (1000 ft), and individual isolated mountains (e.g., an individual mountain located on a plain and less than 3 k (2 mi) in extent) were generally not designated as mountainous areas.
Sectional aeronautical charts were used in most of the regions to determine representative mean elevations of mountain summits and ridge crests in the mountainous areas. Other types of topographic contour maps were used in a few of the regions.
For the nine regions in the contiguous United States, a mean mountain-summit or ridge-crest elevation was determined for each 1/4° latitude by l/3° longitude cell over the mountainous areas. For Alaska, the cells were l/2° latitude by 1° longitude. For Hawaii and the Pacific Islands and Puerto Rico and the Virgin Islands, the cells were l/8° latitude by l/8° longitude.
The procedures used to estimate the mean free-air wind speed and wind power density at mountain-summit and ridge-crest elevations varied according to the sources of upper-air data and other climatological information used. A summary of upper-air data and procedures used in the regional wind energy assessments is provided below. For a more detailed and accurate description on a given region, refer to the wind energy resource atlas for that region. From the results in the regional assessments, one major revelation became apparent on estimating the mean wind speeds and wind power densities for mountain summits and ridge crests; that is, there is no universal procedure for reliably estimating the wind energy potential over mountainous areas. A procedure that appears to work well in one area of the country may give totally unrealistic estimates in another part of the country. Moreover, a procedure may not apply to all seasons of the year. For example, in the Southwest where the most abundant surface data existed from mountain summits and ridge crests, use of the conventional upper-air wind data to estimate ridge crest wind speeds gave fairly reliable estimates for winter but unrealistically low estimates for summer.
The procedure developed by PNL and applied in the Northwest prototype assessment made use of northern hemisphere upper-air climatologies for the 850-, 700-, and 500-mb levels (about 5,000, 10,000, and 18,000 ft elevation) by Crutcher (1959). The free-air wind speed at mountain-summit or ridge-crest elevation was interpolated from the mean scalar speeds on the constant-pressure surfaces. Estimates of the mountain-top wind speeds were made on a grid 1/4° latitude by 1/3° longitude. In each such cell of the grid over mountainous areas, the mean ridge-crest or mountain summit elevation and appropriate constant pressure surface mean wind speeds were estimated. Linear extrapolation provided the mean free-air wind speed at the terrain elevation and the application of a Rayleigh speed distribution gave the mean free-air wind power. The mean wind power at the 10-m (33 ft) and 50-m (164 ft) reference heights was taken to be one-third and two-thirds of the free-air value, respectively, to account for the frictional slowing of the wind near the surface.
These estimates were considered lower limits for exposed ridge crests and mountain summits, since local terrain features in these mountainous areas can enhance the wind power considerably. Also, a major uncertainty in mountainous areas is the representativeness of some of the upper-air wind data from the rawinsonde stations upon which the free-air estimates are based. At some of the upper-air stations, the 850 mb level (about 5,000 ft elevation) is near or below the surface. Even the 700 mb level (about 10,000 ft elevation) at some stations (e.g., at Lander, Wyoming) is below the average mountain-summit elevations of nearby mountains.
Wherever possible, the estimates based on upper-air data were compared with surface data from ridge crests and mountain summits. Over some of the mountainous areas, the estimates were adjusted based on the location of the upper-air recording station and/or the surface data from ridge crests.
Table A-7 shows the regions that used the "standard" (PNL-developed) procedure described above to arrive at the ridge-crest estimates and those regions that used other data and procedures or some refinements of the standard method developed in the Northwest prototype assessment. Five regions used the standard method and six regions used other methods or made refinements to the standard method. One region, the Great Lakes, did not have any mountainous areas. Below are the refinements made to the standard method.
Southwest. In the Southwest region, considerable surface wind data from mountain summits and ridge crests were available for verifying the estimates based on upper-air wind data. Particularly over some of the mountainous areas of the Southwest, applying the free-air winds often results in gross errors, especially during the warm season. Upper-air winds over the Southwest are extremely weak during the warm season, yet surface winds on many of the mountains are frequently strong due to the presence of thermally produced circulations of a mesoscale (monsoonal sea-breeze) and/or toposcale (surface slope heating) nature. Many unusual wind regimes unconnected to free-air flow have been documented in mountainous areas of California. The free-air technique would, for example, predict a mean power density of 15 to 25 W/m2 over a mountain range with a true representative power density of 300+ W/m2. The discrepancy is due to the presence of a modified sea-breeze flow into the desert. Mountain summit estimates were adjusted accordingly during the spring, summer, and fall when substantial fire weather data were available. Winter estimates closely followed the free-air technique, as thermal effects are greatly reduced in the cold season and synoptic effects are increased.
Northeast and Southeast. In the Northeast and Southeast regions, the estimation of wind power over mountains made use of Winds Aloft Summaries (NCDC 1970) for existing rawinsonde stations. Monthly and annual wind speed frequency distributions for 150 m (492 ft) and 300 m (1,000 ft) above the surface as well as for 500 m (1,640 ft) or 1,000 m (3,280 ft) above sea level were used to estimate wind power values for the mean ridge-crest or mountain- summit elevations. Wind power values were first computed for each level that was available for each station. The wind powers at 300 m (1,000 ft) above surface and the upper levels were then used to compute values of a power law exponent for each station for each month and on an annual basis. For the stations nearest each grid cell, the appropriate power law exponents were utilized to compute the wind power in the free atmosphere at the stations, using the cell's mean ridge crest or mountain summit elevation. Horizontal interpolation between stations to each grid cell was performed using an inverse distance squared weighting scheme. This procedure provided an estimate of the free-air wind power at the terrain elevation at the location of each grid cell. As in the standard method, the wind power at the 10-m (33 ft) and 50-m (164 ft) reference heights was taken to be one-third and two-thirds of the free-air value.
Alaska. Over Alaska, the estimation of wind power over mountains made use of Meridional Cross Sections, Upper Winds Over the Northern Hemisphere (Crutcher 1961). The procedure used was very similar to the standard procedure, except that the free-air wind speed at mountain-summit or ridge-crest height was extrapolated (or interpolated) from the mean scalar speeds on the meridional cross sections, and estimates of the mountaintop wind speeds were made on a grid 1/2° latitude by 1° longitude.
Hawaii and Pacific Islands. Over Hawaii and the Pacific Islands, the estimation of wind power over mountains made use of tropical upper-air wind climatologies (Wiederanders 1961), satellite-derived winds, and NCDC upper-air summaries. Major differences from previous atlases occur in Hawaii. The dominant weather system is the trade wind. Peak trade wind speeds occur at 600 m (2,000 ft) and a strong temperature inversion overlies the trades at about 2,200 m (7,000 ft). For mountains above 1,500 m (5,000 ft), the trades are diverted around the mountain barrier. However, elongated low ranges such as Oahu's Koolau Range (900 m or 3,000 ft) allow significant enhancement of open ocean winds as they pass the crest. For the Oahu ranges, estimates from representative wind stations were applied to the mountain crests. For taller Hawaiian mountains and for those in the Pacific Islands (where the trade inversion is not as significant a factor), the procedures developed by PNL were used.
Puerto Rico and the Virgin Islands. Over Puerto Rico and the Virgin Islands, the estimation of wind power over mountains made use of the northern hemisphere upper-air wind climatology for the 850-mb constant-pressure level by Crutcher (1959) and the 1,500-m (7,000 ft) winds-aloft summaries (NCDC 1970). The free-air wind speeds at mountain-peak or ridge-crest heights were estimated using seasonal mean scalar speeds on the 850-mb pressure surfaces, San Juan pibal summaries (Stone 1942), and the San Juan and St. Croix 1,500 m winds-aloft summaries. Extrapolation down to mountain-peak or ridge-crest height was done using a mean trade-wind profile observed by Riehl (1954) and modified by seasonal variations noted by Stone (1942). Estimates of average wind speeds were made for the mean mountain-top/ridge-crest elevations within each 1/8° latitude-longitude cell. The Rayleigh distribution was first used to convert mean free-air wind speeds to free-air wind power. However, the Rayleigh distribution was found to significantly overestimate free-air wind power in the region due to the steadiness of the trade winds. Consequently, the annual and seasonal Rayleigh free-air wind power estimates were corrected by fitting a Weibull distribution to the data (using a least-squares fit) and applying only a percentage of the Rayleigh wind power based upon the value of the Weibull shape parameter. The mean wind power at the 10-m (33 ft) and 50-m (164 ft) reference heights was assumed to be one-third and two-thirds of the free-air values, respectively.
Although approximately 3,000 stations provided the wind resource assessment of the United States with quantitative data, these stations were not uniformly distributed. Most of the stations are located in populated areas and along transportation corridors. Large areas in the United States are devoid of any form of quantitative wind data suitable for this assessment. Furthermore, in areas of complex terrain (including the various mountainous areas), most observation sites (except for some Forest Service fire lookout sites) are confined to valley locations. To evaluate the distribution of the wind resource in data-sparse areas, three qualitative indicators of the wind speed or power were developed and employed, where applicable, in various regions of the United States.
The most widely used technique depended on certain combinations of topographical and meteorological features (Elliott 1979a) that were associated with high or low wind speeds. Those features indicative of high mean wind speeds are:
Features that signal rather low mean wind speeds are:
Areas in which the appropriate features occur were determined by examining topographic contour and shaded relief maps and synoptic and climatological maps of pressure patterns and air flow.
Table A-7 showed the extent to which topographic/meteorological indicators of the wind resource were used in each of the 12 regions of the United States. Except in the Southeast region, considerable use was made of topographic/meteorological indicators to subjectively estimate the wind resource potential in data-sparse areas. The most extensive use of these indicators was in the assessments of the mountainous areas of the United States, where these indicators were often applied to infer considerable variability in the spatial distribution of the wind resource.
Evidence of strong persistent winds can also be found in wind-deformed vegetation, as first described by Putnam (1948). Mean wind speeds can be deduced from the extent of such deformation on trees and shrubs, as discussed by Hewson et al. (1979) and Wade and Hewson (1980). However, there are a number of practical limitations to the use of trees as indicators of mean wind speed. Although wind-flagged trees may indicate that the mean wind speeds are stronger than 4 m/s, trees that are unflagged do not indicate that the winds are light. There may be locations where strong winds come from several directions, and persistence from any one direction is insufficient to cause wind flagging. Nevertheless, in spite of the possible errors that are inherent in use of trees as an indicator of mean annual wind speed, they are useful in identifying potential areas with moderate-to-high wind resource.
Areas of wind-deformed vegetation were identified and evaluated to deduce estimates of the wind resource in four of the regional wind energy assessments: the Northeast, Northwest, Southwest, and Alaska. Methods employed to identify areas of wind-deformed vegetation included aerial, ground, and newspaper surveys.
In the Northeast region, a public survey was conducted through 30 newspapers in the Northeast to solicit information from the public concerning areas of wind-deformed trees as well as other qualitative information on particularly windy areas. As a result of the newspaper surveys, wind resource estimates for 61 sites with wind-deformed trees were included in the Northeast assessment.
In the Southwest region, an aerial survey of Nevada mountains and a ground survey of the California coastal mountains were made to identify areas of wind-deformed vegetation.
In the Northwest region, results of aerial and ground surveys reported by Hewson et al. (1978) provided information on areas of wind-deformed trees and estimates of the wind resource in western Washington and Oregon.
In Alaska, wind-deformed vegetation in the Portage Pass area was evaluated to deduce mean wind speeds and estimates of the wind resource.
The removal and deposition of surface materials by the wind to form playas, sand dunes, and other types of eolian landforms indicate strong winds from a nearly constant direction. Correlating characteristics of eolian features to long-term mean wind speeds has proven difficult (Marrs and Gaylord 1979). However, the distribution of eolian features was used to delineate locations with strong winds and potentially high wind resource in some data-sparse areas in three of the regional assessments: Northwest, Southwest, and Hawaii and the Pacific Islands.
Areas in the Northwest containing eolian landforms were identified by Marrs and Kopriva (1978). The most extensive areas of eolian landforms identified in the Northwest were in central and southern Wyoming.
In Hawaii, eolian features aided in delineating the high wind energy area of northwestern Molokai.
The production of mean wind power density maps depended on the coherent synthesis of several pieces of information. The goal of the synthesis process was to present wind power density values representative of sites that are well exposed to the wind. Hilltops, ridge crests, mountain summits, large clearings, and other locations free of local obstructions to the wind are expected to have good exposure to the wind. In contrast, locations in narrow valleys and canyons, downwind of hills and obstructions, or in forested or urban areas are likely to have poor exposure. The wind power density shown on the maps in this atlas is not representative of poorly exposed locations. Estimates for areas of ridge crests and summits depicted on the maps are lower limits to the wind power expected at well-exposed sites. In such areas, local terrain features can enhance the wind power considerably (e.g., by a factor of 2 or 3). By specifying the type of wind exposure to which the map values of wind power pertain, we avoid the ambiguity that typical-location or average-for-the-terrain values might introduce.
To represent the wind resource at well-exposed sites, it was necessary to evaluate the general land-surface form and topography in the vicinity of every data site. Maps were prepared showing the location of stations, the mean wind power density, the character of anemometer exposure (if known), and the topography and land-surface form.
Preliminary analyses of the wind power density were drawn, noting peculiar anomalies in the site values and the analyzed patterns. The wind characteristics at the anomalous sites (e.g., sites with wind power density significantly different from nearby sites in the area) and topography in the vicinity of the sites were then evaluated in greater detail to determine the possible causes of the anomalous values.
For example, if a site's wind power appeared anomalously high, the evaluator might take a closer look at the site exposure to determine if the site is located on a hilltop, ridge crest, or other elevated feature. In some areas, combinations of topographical and meteorological features that could be indicative of the high resource were considered, where appropriate. If no feasible explanation for the anomalously high value could be found, then the evaluator might consider the value to be unrepresentative and choose to ignore it or adjust it accordingly. If a site's value appeared anomalously low, the evaluator might look for indicators of poor anemometer exposure, poor site exposure, or where appropriate, meteorological and topographical features typically associated with low wind resource. Estimates of the wind power over mountain summits and ridge crests based on upper-air data were evaluated and compared with surface data from ridge crests and summits, where available. In some areas, this comparison resulted in an adjustment of the ridge-crest estimates (e.g., in southern California). Qualitative indicators of the wind resource contributed information in some data-sparse areas.
Only after all these data and information were completely evaluated were the final analyses constructed. The maps for each state were merged into a regional mosaic. The regional mosaics were eventually synthesized into a national assessment.
The analysis of wind power maps departs from conventional isopleth analyses by showing the boundaries of wind power density classes. Each wind power class represents the range of wind power densities likely to be encountered at exposed sites within an area designated as having that wind power class. Table A-8 gives the power density limits for the wind power classes used in the regional atlas for the 10-m (33-ft) and 50-m (164-ft) reference levels.
Wind power density is proportional to the third moment of the wind speed distribution and to air density; therefore, a unique correspondence between power density and mean wind speed (the first moment of the speed distribution) does not exist. However, by specifying a Rayleigh wind speed distribution and a standard sea level air density (1.22 kg/ m3), a mean wind speed can be determined for each wind power class limit. The decrease of air density with elevation requires the mean Rayleigh speed to increase by about 3%/1,000 m elevation (1%/1,000 ft) to maintain the same wind power density. If the wind speed distribution is more sharply peaked than the Rayleigh distribution, the equivalent mean speed will be slightly higher than the value in Table A-8. Conversely, a broader distribution of wind speeds will slightly reduce the equivalent mean speed.
The physical characteristics of the land-surface form affect the number of wind turbines that can be sited in exposed places. For example, over 90% of the land area in a flat plain may be favorably exposed to the wind. However, in mountainous terrain only the ridge crests and passes, which may be only a small percentage (<5%) of the land area, may represent exposed sites. The map of classes of land-surface form by Hammond (1964) provided information on the distribution of plains, tablelands, hills and mountains in the United States.
For each class of land-surface form, the percentage of land area that is representative of well exposed, moderately exposed, and poorly exposed sites has been estimated. These percentages were determined subjectively as a function of the slope, local relief, and profile type specified by Hammond. Table A-9 gives the average percentage of land area that is designated as exposed terrain for the different classes of land-surface form found in the Northwest region. There are slight variations in these average percentages from region to region.
The analyses of wind power density at exposed sites shown on the wind power maps depend on the subjective integration of several factors: quantitative wind data, qualitative indicators of wind speed or power, the characteristics of exposed sites in various terrains, and familiarity with the meteorology, climatology and topography of the region. As a result, the degree of certainty with which the wind power class can be specified depends on:
A certainty rating, from 1 (low) to 4 (high), of the wind energy resource estimate has been made for each grid cell of a 1/4° latitude by l/3° longitude grid over the contiguous U.S. by considering the influence of the above three factors on the certainty of the estimate of the wind power class. Different sized grid cells were used for the other regions. The certainty ratings have been digitized for each grid cell in the United States.
The definitions for the certainty ratings are adopted from those used by Voelker et al. (1979) in a resource assessment of U.S. Forest service tracts. The certainty ratings for the wind resource assessment are defined as follows:
Rating 1. The lowest degree of certainty. A combination of the following conditions exists:
Rating 2. A low-intermediate degree of certainty. One of the following conditions exists:
Rating 3. A high-intermediate degree of certainty. One of the following conditions exists:
Rating 4. The highest degree of certainty. Quantitative data exist at exposed sites in the vicinity of the cell and can be confidently applied to exposed areas in the cell because of the low complexity of terrain and low spatial variability of the resource.
The assignment of a certainty rating requires subjective evaluation of the interaction of the factors involved.
Areal Distribution of the Wind Resource
As noted above, the wind power density class values shown on the maps apply only to sites well exposed to the wind. Therefore, the map area designated as having a particular wind power class does not indicate the true land area experiencing this wind power. Instead, there is a complicated and difficult- to-quantify relationship among the class of land-surface form, the land-surface area and the map value of wind power density. For each land-surface form, the fraction of land area that would be favorably exposed to the winds (i.e., have the wind power density indicated on the map) was estimated. Table A-9 shows the averages for various land-surface forms. Furthermore, to be able to establish a wind power density for the remaining area, it was also necessary to specify a factor by which the wind power shown on the map is reduced in the less exposed areas. As an additional complication, some land-surface forms, isolated hills, and ridges that rise above a nearly flat landscape may even experience a higher power density than the map indicates.
To accommodate these various situations, the land area represented by a given land-surface form was divided into four exposure categories: better exposure than typical for the terrain, exposure typical for that land-surface form, partially sheltered exposure, and very sheltered exposure. The partitioning of the land-surface forms into the four categories was based primarily on the parameters used to classify the land-surface forms.
In order to adjust the wind power density from the map value to the various exposure categories, the power density was scaled to be greater than, equal to, slightly less than, and much less than the map value power density. The factor by which the map value was adjusted to represent the wind power density in each category was determined by the magnitude of elevation relief given in Hammond's maps of the land-surface form. The minimum power density allowed for a category was the median value of wind power density class 1. The scaling factors for the wind power density were based on a conservative application of a power-law type vertical adjustment with the height change specified by the terrain relief code.
For each grid cell, the land-surface form was specified, and the wind power class associated with a typically exposed site in that land-surface form was determined. By partitioning the area of the cell into the four exposure categories, and by scaling the wind power class to each category, the contribution of that cell to the areal distribution was determined.
A cell-by-cell representation of the areal distribution is given in a map that indicates the percentage land area in a cell over which the wind power class equals or exceeds a threshold value. In the regional atlases, areal distribution maps on a state-by-state basis are shown for threshold values of wind power classes 2, 3, 4, and 5. In this atlas, national areal distribution maps are shown for threshold values of wind power class 3 and 4.
In each of the 12 regional atlases, a summary table of the areal distribution that combines the contributions by each cell is provided for the region and for each state in the region. For each wind power class, the sum of the area contributed by each exposure category was also determined for each state and the region. Summing the area associated with each wind power class in each cell gives the area of the state or region over which the power class exceeds a given value. The table gives the estimated land area (km2) and the percentage of land area associated with each power class.
Both of these presentations of the areal distribution of the wind resource are highly dependent on the estimate used to partition the land area into the four exposure categories and on the scaling of the power density for each category of exposure. Therefore, the areal distribution derived from the wind power and land-surface form maps must be considered only an approximation. The quantity and quality of wind data and topographic information required to make a highly accurate cell-by-cell appraisal of the wind resource goes far beyond the scope of these assessments. However, as wind information becomes available through new measurement programs or through the discovery and processing of existing data sets, the evaluation of the areal distribution of the wind resource can be improved on a cell-by-cell basis.
Changery, M. J., W. T. Hodge and J. V. Ramsdell. 197 Index-Summarized Wind Data. BNWL-2220 WIND-I I NOAA/Pacific Northwest Laboratory, National Climatic Data Center, Asheville, North Carolina.
Changery, M. J. 1978. National Wind Data Index. HCO/T1041-01, DOE/NOAA, F(49-26)-1041, National Climatic Data Center, Asheville, North Carolina.
Cliff, W. C. 1977. The Effect of Generalized Wind Characteristics on Annual Power Production Estimates From Wind Turbine Generators. PNL-2436, Pacific Northwest Laboratory, Richland, Washington.
Crutcher, H. L. 1959. Upper Wind Statistics of the Northern Hemisphere, Vol. 1. NAVAER 50-IC-535, Washington, DC.
Crutcher, H. L. 1961. Meridional Cross-Sections. Upper Winds Over the Northern Hemisphere. Technical Paper No. 41, National Weather Records Center, U.S. Weather Bureau, Asheville, North Carolina.
Elliott, D. L. 1977. Synthesis of National Wind Energy Assessments. BNWL-2220 WIND-5, Pacific Northwest Laboratory, Richland, Washington.
Elliott, D. L. 1979a. "Adjustment and Analysis of Data for Regional Wind Energy Assessments." Paper presented at the Workshop on Wind Climate, Asheville, North Carolina, November 12-13, 1979.
Elliott, D. L. 1979b. "Meteorological and Topographical Indicators of Wind Energy for Regional Assessments." Proceedings of the Conference on Wind Characteristics and Wind Energy Siting. American Meteorological Society, Boston, Massachusetts.
Furman, R. W., and G. E. Brink. 1975. The National Fire Weather Data Library. USDA For. Ser. Gen. Tech. Rep. RM-19, Rocky Mt. Fo. and Range Exp. Stn., Fort Collins, Colorado.
Hammond, E. H. 1964. "Analysis of Properties in Landform Geography: An Application to Broadscale Landform Mapping," Annals, Association of American Geographers, 54:11-19, Map Supplement Number 4.
Hewson, E. W., et al. 1978. Vegetation as an Indicator of High Wind Velocity. RLO/2227-T24-78-2. Available from National Technical Information Service, Springfield, Virginia.
Hewson, E. W., et al. 1979. A Handbook on the Use of Trees as an Indicator of Wind Power Potential. RLO/2227- 79/3. Available from National Technical Information Service, Springfield, Virginia.
Marlatt, W. E., P. Tierney, P. Meikle, M. Baer, and J. Childs. 1979. Assessment of the Applicability of the National Fire Weather Library to Wind Energy Analyses. PNL-2538, Pacific Northwest Laboratory, Richland, Washington.
Marrs, R. W., and D. R. Gaylord. 1979. A Guide to Interpretation of Windflow Characteristics From Eolian Landforms. RLO/2343-79/2. Available from National Technical Information Service Springfield, Virginia.
Marrs, R. W., and S. Kopriva. 1978. Regions of the Continental United States Susceptible to Eolian Action. RLO-2343-78/2, University of Wyoming, Laramie, Wyoming.
National Climatic Data Center (NCDC). 1970. Winds Aloft Summaries by Month. Available from National Weather Records Center, Federal Building, Asheville, North Carolina.
National Climatic Data Center (NCDC). 1972. Environmental Guide for the U.S. Gulf Coast. Available from the National Climatic Data Center, Asheville, North Carolina.
National Climatic Data Center (NCDC), and University of Alaska. 1977. Climatic Atlas of the Outer Continental Shelf and Coastal Regions of Alaska. Available from Arctic Environmental Information and Data Center, University of Alaska, 707 A Street, Anchorage, Alaska, and from the National Technical Information Service, Springfield, Virginia.
National Oceanic and Atmospheric Administration (NOAA). 1976. U. S. Standard Atmosphere. NOAA- S/T76-1562, Washington, DC.
Naval Weather Service Detachment. 1974. U.S. Navy Marine Climatic Atlas of the World: Volume I - North Atlantic Ocean. Naval Oceanography and Meteorology, Asheville, North Carolina.
Naval Weather Service Detachment. 1976. Climatic Study of the Near Coastal Zone: West Coast of the United States. Naval Oceanography and Meteorology, Asheville, North Carolina.
Putnam, P. C. 1948. Power From the Wind. Van Nostrand, New York, New York.
Riehl, H. 1954. Tropical Meteorology. McGraw-Hill Book Company, Inc., New York, New York.
Stone, R. G. 1942. "On the Mean Circulation of the Atmosphere Over the Caribbean." Bulletin of the American Meteorological Society, 23:4.
U.S. Naval Oceanography Command. 1975. Summary of Synoptic Meteorological Observations: North American Coastal Areas-Revised. U.S. Naval Oceanography Command, Washington, DC.
Verholek, M. G. 1977. Summary of Wind Data From Nuclear Power Plant Sites. BNWL-2220 WIND-4, Pacific Northwest Laboratory, Richland, Washington.
Voelker, A. H., et al. 1979. A Systematic Method for Resource Rating With Two Applications to Potential Wilderness Areas. ORNL/TM-6759, Oak Ridge National Laboratory, Oak Ridge, Tennessee.
Wade, J. E. and E. W. Hewson. 1980. A Guide to Biological Wind Prospecting. DOE/ET/20316-80-2. Available from National Technical Information Service, Springfield, Virginia.
Wahl, E. W. 1966. Windspeed on Mountains. No. AF19(628)-3873, USAF-CAL, Bedford, Massachusetts.
Wegley, H. L., et al. 1980. A Siting Handbook for Small Wind Energy Conversion Systems. PN L-252 I Rev. 1, Pacific Northwest Laboratory, Richland, Washington.
Wiederanders, C. J. 1961. Analysis of Monthly Mean Resultant Winds for Standard Pressure Levels Over the Pacific. Rep. No. 3, Hawaii Institute of Geophysics, University of Hawaii, Honolulu, Hawaii.
Wyrtki, K., and G. Meyers. 1975. The Trade Wind Field Over the Pacific Ocean. Part 1. The Mean Field and the Mean Annual Variation. HIG 75-1, Hawaii Institute of Geophysics, University of Hawaii, Honolulu, Hawaii.
Table of Contents
Return to RReDC Homepage ( http://rredc.nrel.gov )