Using GIS to Determine Wind Energy Potential in Minnesota, USA Zihan

Using GIS to Determine Wind Energy Potential in Minnesota, USA

Zihan Yang Department of Resource Analysis, Saint Mary’s University of Minnesota, Winona, MN 55987

Keywords: GIS, Wind Energy, Terrain Elevation, Multi-Criteria, Spatial Analysis

Abstract

Wind power is an alternative to fossil fuels. It is plentiful, renewable, widely distributed, clean, produces no greenhouse gas emissions during operation and uses relatively little land. Recent studies indicate that Minnesota has plentiful wind energy potential including the possibility to use wind to generate 25 times the electricity Minnesota used in 2010 (American Wind Energy Association, 2012a). This study evaluates the potential wind energy resources in Minnesota by analyzing technical, environmental, and political criteria for developing local renewable wind energy resources. The approach assesses suitable locations for exploring wind turbines energy sources with the aid of a geographic information system. The study considers local wind conditions and other restrictions such as terrain, land use, environment, and human activities. The model identifies areas that have an excellent suitability for future wind turbine placement in Minnesota. When it compares to the approximate estimations based solely on analysis of wind speed, this rule-based model may provide a more accurate method to evaluate wind energy in Minnesota.

Introduction contributing cause of global warming is the emission of greenhouse gases (in The issue of global warming is particular carbon dioxide), mainly becoming a potential severe challenge generated by fossil fuel combustion that nations around the world would be (United States Environmental wise to confront. The Protection Agency, 2010). Intergovernmental Panel on Climate The negative effects of fossil Change (IPCC) indicated during the fuel combustion on the environment, in twenty-first century the average global addition to limited fossil fuel supply, surface temperature will likely rise have forced many countries to explore between 1.1° and 2.9° C (2° to 5.2° F) and change to environmentally friendly according to their lowest emissions renewable alternatives in order to meet scenario and between 2.4° and 6.4° C increasing energy demands. Currently, (4.3° to 11.5° F) according to their wind energy is one of the fastest highest emissions scenario (IPCC, developing renewable energy 2007). Scientists have confirmed that a technologies around the world

Zihan Yang. 2013. Using GIS to Determine Wind Energy Potential in Minnesota, USA. Volume 15, Papers in Resource Analysis. 13 pp. Saint Mary’s University of Minnesota University Central Services Press. Winona, MN. Retrieved (date) from http://www.gis.smumn.edu.

(Ramachandra and Shruthi, 2005). (2) Environmental constraints such According to a report by the as wildlife conservation areas American Wind Energy Association and bodies of water. (AWEA, 2012b), by the end of 2012 (3) Human impact factors such as the United States (US) wind industry major roads, noise, state totaled 60,007 MW of cumulative parkland, and electric lines. wind capacity (from more than 45,100 turbines). Over one fifth of the world’s Datasets for the project were installed wind power capacity is obtained or derived from existing located in the US (BP p.1.c., 2013). datasets and analyzed using separate Although the US wind industry has physical, environmental, and human already made significant strides in impact models. This rule-based model developing electric energy from wind integrated individual outputs into one turbines, there is still plenty of final result which identified suitable opportunity for growth (Schrader, sites for wind turbines in Minnesota. 2012). “Minnesota ranked 3rd in the Methods US in 2010 for percentage of electricity derived from wind” (AWEA, Database Development 2012a). This fact sent a signal to wind energy companies that Minnesota is a The rule-based model used in this wind-friendly state. According to study was a spatial multi-criteria model National Wind (2012), Minnesota has a using diverse variables to represent the total wind resource of 489,271 MW, physical, environmental, and human and this capacity of wind power could impact factors determining wind support nearly 25 times the amount of turbine site suitability. By using Esri’s electricity Minnesota currently needs. ArcMap software, data for each of Thus, there are many potential areas these criteria were converted from for wind turbines. vector format to raster format and However, the locations with the assessed using overlay analysis. All highest wind resources are not always tools referenced in this paper are from feasible sites for wind farms. A variety Esri’s ArcMap software. of factors are decisive in the site The study area for this project selection of wind turbine farms. In this included the entire state of Minnesota. study, a rule-based modeling method In 2006, the Minnesota Department of was used to evaluate and target Commerce developed maps of suitable wind power sites. Using a Minnesota wind resources at different similar approach as Rodman and heights, 30, 80, and 100 m above Meentemeyer (2006), this model ground. Since the height of large wind incorporated the following factors: turbines is generally between 50 and 80 m above ground, wind speed data at (1) Physical criteria such as wind 80 m was used in this study (Figure 1). resources, obstacles, and Terrain strongly affects the terrain. performance of wind turbines;

2 therefore, a slope dataset was derived components: average annual wind from a 30 m digital elevation model speed at 80 m height, obstacles, and (DEM) developed by the United States terrain. Geological Survey (USGS, 2009). The annual average wind speed The environmental data for is useful as a predominant indicator of Minnesota, such as wildlife wind resources. Ten meters is the conservation areas and open water standard height for a typical areas were obtained by the Minnesota meteorological station to measure wind Department of Natural Resources speed. The wind speed suitability for (MNDNR, 2013). Human impact developing wind turbines was factors such as populated areas, major classified into four categories based on roads, and state parkland were also wind velocity at 10 m: poor, marginal, obtained from the MNDNR (2013). good, and excellent (Table 1) The energy grid dataset for Minnesota (Ramachandra and Shruthi, 2005). was obtained from the Minnesota Land Management Information Center Table 1. Annual mean wind speed and (2007). The data above were converted potential value of the wind energy resource at to 30 m resolution raster datasets for 10 m height (Ramachandra and Shruthi, 2005.) the raster overlay analysis. Annual mean wind Indicated value speed at 10 m of wind resource <4.5 m/s Poor 4.5-5.4 m/s Marginal 5.4-6.7 m/s Good >6.7 m/s Excellent

Since the hub height of a modern wind turbine is greater than 50 m and the available wind speed data was measured at 80 m above ground, the wind speed classification at 10 m was converted to the wind speed at 80 m in order to provide a uniform standard for analysis. The following formula developed by Davenport (1960) was used to extrapolate wind speed at different heights: Figure 1. Minnesota average annual wind speed at 80 m height. V Z Z (1) Model Development VG Z G

Physical Model VZ = the known mean wind speed at height Z in the study area (m/s) The physical model incorporated three VG = the mean wind speed at height ZG, 3

which needs to be extrapolated wind flow and decrease the efficiency (m/s) Table 3. Annual mean wind speed and Z = the height for which the wind potential value of the wind energy resource at speed VZ is computed (m) 80 m above ground in Minnesota.

ZG = the height at which VG is first Annual mean wind Indicated value observed in the same terrain speed at 80 m of wind resource <6.03 m/s Poor (1) α = an empirical exponent, which 6.03-7.24 m/s Good (2) depends on the roughness of the >7.24 m/s Excellent (3) surface. of wind power generation. Since wind According to the available turbines sites are best located away wind data in this research, formula (1) from highly populated areas (described was adapted as the following equation later), obstructions such as buildings (2), which estimated wind speed in the are unlikely obstacles. The presence of same terrain for 80 m height. trees is recognized as an obstacle in open areas. Wind turbine suitability 80 0.14 decreases as the density of trees = V 10 (2) 10 increase (Rodman and Meentemeyer, 2006). For this model, a forest layer was derived from the National Gap By using equation (2), the new Analysis Program (2001) Land Cover wind classification at 80 m height was Data. Forest densities were determined developed (Table 2). from vegetation types at a given location. If the attribute table of the Table 2. Annual mean wind speed and vegetation dataset showed the land was potential value of the wind energy resource at occupied by trees in the study area, it 80 m height. was defined as a high density forest. If Annual mean wind Indicated value the attribute table showed there were speed at 80 m of wind resource no trees in the study area, it was <6.03 m/s Poor defined as no forest. 6.03-7.24 m/s Marginal The “no forest” class was 7.24-8.98 m/s Good further divided into two subclasses. >8.98 m/s Excellent These were “Shrub Land and Grassland” and “Agricultural However, the range of wind Vegetation.” The Ames Laboratory speed values published by the (2010) reported “wind turbines help Minnesota Department of Commerce channel beneficial breezes over nearby (4.92 m/s to 8.86 m/s) was narrower plants.” This can keep the crops cooler than the resulting suitability and drier, which helps them fend off classification. Therefore, the wind fungal infestations. Meanwhile, the speed criterion at 80 m above ground ability of crops to absorb carbon was modified as displayed in Table 3. dioxide (CO2) from the air and soil can Tall obstacles can obstruct be improved by wind turbines (Ames

Laboratory, 2010). Also, agricultural Environmental Model producers can generate additional revenue by using their land to generate “Two general types of local impacts to wind energy (Rodman and birds have been demonstrated at Meentemeyer, 2006). Therefore, existing wind facilities: (1) direct agricultural land was considered more mortality from collisions and (2) suitable than general shrub land and indirect impacts from avoidance of an grassland (Table 4). area, habitat disruption and behavioral effects” (National Wind Coordinating Table 4. Land obstacles and corresponding Collaborative (NWCC), 2010). Only suitability score. direct mortality has been shown in bat Obstacles Suitability populations (NWCC, 2010). Strategic High Density Forest Poor (1) placement of turbines away from Shrub Land & Good (2) important breeding grounds and high Grassland population areas can reduce their Agricultural Vegetation Excellent (3) environmental impact (van Haaren and Fthenakis, 2011). Therefore, to prevent Construction of turbines on birds or bats from being killed by wind slopes greater than 10% is difficult turbines, wind facilities need to be because the turbine components are established as least 500 m from generally large and heavy. If the slope wildlife conservation areas or areas of is too steep, it limits the accessibility known high use (Yue and Wang, 2006). of the cranes needed to lift such heavy The wildlife conservation area dataset equipment (van Haaren and Fthenakis, was obtained from the MNDNR (2013). 2011). Rodman and Meentemeyer After converting the data format from (2006) classified the degree of slope vector to raster, a raster dataset further by dividing slope into several depicting distance from wildlife suitability classes (Table 5). This conservation areas was developed. classification was adapted for the Then, the areas less than 500 m from model used in this study. wildlife conservation areas were designated as “Unsuitable Area” and Table 5. The slope degree intervals and the areas farther than 500 m from corresponding suitability score (Rodman and wildlife conservation areas as Meentemeyer, 2006). “Suitable Area”. Slope Suitability Although there are no strict Unsuitable (NA) regulations that dictate the appropriate >30° distance to water bodies, Baban and Poor (1) Parry (2001) noted guidelines from 60 16-30° local authorities in the United Good (2) Kingdom by means of a survey and 7-16° determined a representative distance of Excellent (3) 400 m to water bodies was 0-7° recommended (van Haaren and Fthenakis, 2011). The water body 5 dataset was derived from the National existing grid connections and roads is a Gap Analysis Program (2001) Land critical element that needs to be Cover Data. This raster dataset considered in the human impact model. included all open water in Minnesota, Baban and Parry (2001) included a including wetlands, lakes, and rivers. maximum distance of 10 km to the Areas within 400 m from water bodies electric grid as a criterion in their study. were defined as “Unsuitable Area” and The electric transmission lines dataset the areas over 400 m from water was obtained from the Minnesota Land bodies were defined as “Suitable Management Information Center. Area”. Areas within 10 km of the electric grid were regarded as “Suitable Area”, and Human Impact Model the remaining areas were regarded as “Unsuitable Area”. Additionally, since In consideration of safety, visual heavy cranes and trucks carrying pollution, and noise impact for humans, components need to be navigated to wind turbines should be placed away the turbine sites, wind turbines should from populated areas. HGC not be located farther than 10 km from Engineering (2007) conducted a roads (Baban and Parry, 2001). A summary of guidelines for the dataset including all the major roads in maximum sound pressure level in Minnesota was obtained from the provinces of Canada in 2007. The level MNDNR. The areas less than 10 km between 40 and 55 dB is considered an from major roads were regarded as ideal range for human settlement “Suitable Area”, and the remaining within a distance of 500 m from wind areas were regarded as “Unsuitable turbines (HGC Engineering, 2007). Area”. Public opposition to sight of wind turbines in recreational areas was Model Integration taken into consideration in the human impact model by Rodman and The performance of wind turbines can Meentemeyer (2006). Therefore, a be influenced by every factor discussed criterion that wind turbines be at least above; however, each factor presents 1,000 m from public parkland was an entirely different element of included. The parkland dataset was importance to the overall suitability converted from vector format to raster measurement. format. Then, parkland and the areas In the physical model, there within 1,000 m of parkland were were three components: average defined as “Unsuitable Area”; the areas annual wind speed, obstacles, and 1,000 m from parkland were defined as terrain. A weighting scheme similar to “Suitable Area”. the one used by Rodman and A goal of wind energy Meentemeyer (2006) was adopted. developers is to maximize profits for Since the availability of adequate wind themselves and for investors (van resources is the most important Haaren and Fthenakis, 2011). criterion for the physical model, it was Therefore, the utilization of the assigned the value of three as its

6 weight. Because the sufficient wind between two and three (3> S ≥2), the resource is the most critical criterion suitability was considered “good”; if compared to the presence of obstacles, the calculated value was less than two, the obstacle layer was assigned the the suitability was regarded as “poor”. lesser weight of two (Rodman and For the environmental model Meentemeyer, 2006). A weight of one and human impact model, a binary was given to the terrain dataset. Terrain scoring system was applied rather than was considered only in terms of initial a continuous system. The constraints, turbine construction suitability and not such as wildlife conservation areas, in regards to its impact on long-term water bodies, populated areas, roads, operation; therefore, it was considered grid lines, and state parkland, were less critical than the wind resource and used to further identify the locations presence of obstacles. where wind turbines are either suitable The datasets were reclassified or not suitable. Therefore, there was no into several suitability categories or need to assign any score or weight for values, as described in Tables 3-5. To these layers. For the locations where evaluate the physical model, not only the unsuitable factors were found, the the suitability score of each criterion cell in the resulting raster dataset was was needed, but also the weight of set to “0”; for the suitable areas, the each criterion’s importance to the cell was assigned the value “1”. overall model. Thus, the equation from A composite of all three Rodman and Meentemeyer (2006) was suitability models was also generated. used to evaluate the physical model as If any suitable area defined by the follows: physical model corresponded with an area that had been recognized as n unsuitable according to either the S W ij i environmental or human impact model, S i , n then the overall suitability rating was Wi i considered to be “unsuitable” at that location, regardless of the score from where n is the number of input layers, the physical model and the area was

Sij is the score for the jth class of the ith removed from the resulting map. input layer, Wi is the weights of the ith input layer, and S is the calculated Result suitability factor for each grid cell A summary of the criteria for wind location in the model output. Each cell turbine placement included in the in the resulting raster dataset generated physical model, environmental model, using the above equation had a value and human impact model is shown in between one and three. If the Table 6. The physical model calculated value equaled three, the encompassed the locations where there corresponding suitability of the cell is sufficient wind, a flat gradient, and was classified as “excellent”, since it few obstacles. According to the result has the best performance in every of the physical model (Figure 2), the category. If the calculated value fell most suitable areas are located in 7 southwestern Minnesota. Abundant areas within 10 km of major roads and wind resources and the lack of energy lines were selected as suitable. vegetation obstacles in southwestern Populated areas and state parkland Minnesota were decisive were considered unsuitable due to characteristics. Areas of excellent noise and visual impact concerns. physical suitability (score = 3) for large wind turbines contained 8,473,152.42 ha, good suitability (score = 2) contained 3,429,473.22 ha, and poor suitability (score = 1) contained 6,870,818.97 ha. Figure 3 indicates the result of the environmental model. As explained earlier, wildlife conservation areas and open water areas were recognized as unsuitable areas for wind facilities due to environmental concerns and physical constraints. These layers were merged together to exclude any areas that were within a certain distance of wildlife conservation areas and open water (Figure 4). Considering the aspects of Figure 2. The suitability result of the physical accessibility and economy, only the model.

Table 6. Summary of wind farm location criteria. Objective Criteria Physical Model: 1. Wind Speed Greater than 7.24 m/s at 80 m height 2. Obstacle Should be located in low vegetation area 3. Terrain Slope Should not be located in areas with steep slopes

Environmental Model: 1. Wildlife Conservation At least 500 m away from wildlife conservation Areas areas 2. Open Water At least 400 m away from open water

Human Impact Model: 1. Populated Area 500 m away from nearest populated area 2. Road Should not be located greater than 10 km from road 3. Grid Line Should not be located greater than 10 km from grid 4. State Parkland Should be located 1,000 m away from parkland

wind turbines. If areas occupied by existing wind turbines are subtracted, the remaining potential area is 6,149,487.33 ha.

Figure 3. The suitable areas defined by the environmental model.

Combining the three models resulted in areas of suitable areas amassing 6,176,487.33 ha. This area Figure 4. The suitable areas defined by the only includes locations deemed to have human impact model. excellent suitability (Appendix A). According to a study Discussion announced by the United States Department of Energy (2013), at the The amount of remaining wind energy end of 2012 the installed capacity for resources can be estimated using the wind power in Minnesota was 2,986 model and mathematical deduction. MW, with wind power accounting for Kimballa (2010) indicated a typical 14.3% of the electricity generated in wind farm consists of approximately the state. Therefore, this model 15 wind turbines per 1,000 acres. By predicts that Minnesota has a potential converting the acres to hectares, a large capacity of 680,088 MW from wind turbine usually occupies 27 ha. The power. If this amount of energy is Pipestone, Minnesota Chamber of developed, the fossil fuel power Commerce (2013) has compiled data generation industry in Minnesota could suggesting there are approximately be drastically reduced. 1,000 existing wind turbines in There are four large clustered Minnesota. wind farms in Minnesota. By If the area occupied by one comparing the results of this research wind turbine is 27 ha, 1,000 wind with the existing wind farm locations turbines will cover 27,000 ha. The shown in Appendix A, all existing suitability model concluded there are wind farm locations fall within areas 6,176,487.33 ha areas suitable to build the model deemed suitable. In addition,

9 these four wind farms (Buffalo Ridge Dakota and the southern portion that Wind Farm, Fenton Wind Farm, borders Iowa also has a high suitability Nobles Wind Farm, and Bent Tree for building wind turbines. The final Wind Farm) are generally distributed model was verified by comparing the in the south and southwest portions of results to existing wind farm locations Minnesota. The western and in Minnesota. mid-western areas of Minnesota also have vast potential for building wind Acknowledgements turbines, especially the areas bordering North Dakota. Moreover, the I am indebted to the many people who overwhelming land use in these areas have assisted me in the writing of this is farmland. Therefore, placing wind paper. First, I would like to express my turbines there would neither jeopardize heartfelt gratitude to my instructor, Ms. the ecological environment nor Greta Bernatz, for her warm-hearted severely impact human life. encouragement and valuable advice, especially for her patient instruction Conclusion and guidance every time I came across a problem. Without her help, I could The development of wind energy is not have completed this graduate one of the most important tasks for the project. I would like to express my future, not only because greenhouse gratitude to Ms. Nancy K. Rader who gas emissions will likely continue to works for the Minnesota Geospatial impact global warming, but also Information Office. As a GIS data because fossil fuels, the most widely coordinator, she provided access to used energy source are limited in much of the necessary data. I also want abundance. to thank Ms. Melissa Gordon who Wind, an alternative energy helped me review my graduate paper source, is clean and widely distributed and correct grammar. Finally, I would around the world. This research like to express thanks to my family and evaluated the suitability of building my friend Yabing Gao for their wind turbines in Minnesota. By valuable encouragement and spiritual integrating the corresponding criteria, support. three models related to physical, environmental, and human impact References constraints were created. By using spatial analysis in a geographic American Wind Energy Association information system, the models were (AWEA). 2012a. Fourth Quarter combined to produce a final result of 2012 Market Report. Retrieved potential wind farm locations spring 2013 from http://www.awea. (Appendix A). The suitable areas for org/learnabout/publications/upload/4 building wind turbines were generally Q-11-Minnesota.pdf. aggregated in the southwest portion of American Wind Energy Association Minnesota. Additionally, the part of (AWEA). 2012b. AWEA U.S. Wind western Minnesota that borders North Industry Annual Market Report.

Retrieved spring 2013 from: ftp://ftp. lmic.state.mn.us/outgoing/ http://www.awea.org/suite/upload/A archive/transmission_lines. WEA_USWindIndustryAnnualMarke National Gap Analysis Program. 2001. tReport2012_ExecutiveSummary.pdf. Land Cover Dataset. Retrieved spring Ames Laboratory. 2010. Wind 2013 from http:// gapanalysis.usgs. Turbines on Farmland May Benefit gov/gaplandcover/. Crops. Retrieved spring 2013 from National Wind. 2012. Minnesota Wind https://www.ameslab.gov/news/news- Statistics. Retrieved spring 2013 from releases/wind-turbines. http://www.nationalwind.com/minnes Baban, S., and Parry, T. 2001. ota-wind-data/. Developing and Applying a National Wind Coordinating GIS-assisted Approach to Locating Collaborative (NWCC). 2010. Wind Wind Farms in the UK. Renewable Turbine Interactions with Birds, Bats, Energy, 24, 59-71. and their Habitats: A Summary of BP p.l.c. 2013. Wind Power. Retrieved Research Results and Priority May, 2013 from http://www.bp.com/ Questions. Retrieved spring 2013 sectiongenericarticle.do?categoryId= from http://www1.eere.energy.gov/ 9024940&contentId=7046497. wind/pdfs/birds_and_bats_fact_sheet. Davenport, A.G. 1960. A Rationale for pdf. the Determination of Design Wind Pipestone, Minnesota Chamber of Velocities. Proceedings ASCE Commerce. 2013. Wind Power in Structural Division, 86, 39-66. Pipestone County, Minnesota. Howe Gastmeier Chapnik Limited Retrieved spring 2013 from (HGC Engineering). 2007. Wind http://www. Pipestoneminnesota. Turbines and Sound: Review and com/visitors/windpower/. Best Practice Guidelines. Retrieved Ramachandra, T. V., and Shruthi, B. V. spring 2013 from http://www.canwea. 2005. Wind Energy Potential ca/images/uploads/File/CanWEA_Wi Mapping in Karanataka, India, Using nd_Turbine_Sound_Study_F_Final.p GIS. Energy Conversion and df. Management, 46, 1561- 1578. Kimballa, J. 2010. Farming Wind Rodman, L.C., and Meentemeyer, R.L. Versus Farming Corn for Energy. 2006. A Geographic Analysis of Retrieved spring 2013 from Wind Turbine Placement in Northern http://8020vision.com/2010/09/02/far California. Energy Policy, 34, ming-wind-versus-farming-corn-for-e 2137-2149. nergy/. Schrader, S. 2012. North American Minnesota Department of Natural Wind Energy Growth. Green Chip Resources (MNDNR). 2013. DNR Stocks. Retrieved spring 2013 from Date Deli. Retrieved spring 2013 http://www.greenchips from http://deli.dnr.state.mn.us. tocks.com/articles/north-american-wi Minnesota Land Management nd-energy-growth/1591. Information Center. 2007. Electric The Intergovernmental Panel on Transmission Lines and Substations Climate Change (IPCC). 2007. IPCC Dataset. Retrieved spring 2013 from Fourth Assessment Report: Climate

Change 2007. Retrieved spring 2013 from http://www.ipcc.ch/publications _and_data/ar4/wg1/en/ch10s10-es.ht ml. United States Department of Energy. 2013. Wind Powering America: Installed U.S. Wind Capacity and Wind Project Locations. Retrieved April 3 2013 from http://www. windpowering america.gov/. United States Environmental Protection Agency (EPA). 2010. Overview of Greenhouse Gases. Retrieved spring 2013 from http:// www.epa.gov/climatechange/ghgemi ssions/gases/co2.html. United States Geological Survey (USGS). 2009. National Elevation Dataset. Retrieved spring 2013 from http://nationalmap.gov. van Haaren, R., and Fthenakis, V. 2011. GIS-based Wind Farm Site Selection Using Spatial Multi-criteria Analysis (SMCA): Evaluating the Case for New York State. Renewable and Sustainable Energy Reviews, 15, 3332-3340. Yue, C., and Wang, S. 2006. GIS-based Evaluation of Multifarious Local Renewable Energy Sources: A Case Study of the Chigu Area of Southwestern Taiwan. Energy Policy, 34, 730-42.

Appendix A. The Potential Suitable Areas for Building Wind Turbines in Minnesota.