Brendan Loughran and Kelly Henderson
Colgate University, Hamilton, NY
May 2008

  1. Introduction
    As part of Colgate University’s Environmental Studies Seminar entitled “Interdisciplinary Investigation of Environmental Issues: Alternative Energies,” a group of student geographers, economists and physicists formed a wind energy group with the overarching goal of investigating Colgate’s potential to commit to a program devoted to constructing a wind turbine on or off Colgate’s landholdings. Within the wind energy group, our “Data Analysis and Acquisition” subgroup was responsible for (1) acquiring a professional-grade anemometer, (2) erecting the anemometer at a test site, (3) taking wind speed and direction data from the established site and (4) modeling the data set to a third party’s data for the same logging period to facilitate further analysis.
  2. Anemometer Specifications
    In our search for the correct anemometer, we wanted a professional-grade device that could take both directional and speed data while also having data logging capabilities.

    anemometer crop

    Figure 1: Our HOBO Wind Speed/Direction Smart Sensor.

    After researching various options, we decided to purchase Onset Computer Corporation’s HOBO Wind Speed / Direction Smart Sensor (Fig 1). The monitoring station also accepts up to four sensors, which allows for the possibility of purchasing additional Smart Sensors for measuring other weather parameters or additional identical Wind Sensors for wind profiling at multiple heights. The anemometer calculates instant wind velocity based on cup revolutions accumulated every three seconds and logs wind speed by calculating the average over the specified logging interval. Directional data is given using the vector wind direction accumulated every three seconds. The value given in the data set is the average calculated from the sum of the vector components over the logging interval.
  3. Site Specific Details (Bonney Hill, Hamilton, NY)
    For our test site, we chose to set up the anemometer on the roof of Professor Ian Helfant’s house at coordinates N 42◦50’130” W 075◦31’004” and an altitude of 1520 feet. These measurements were obtained using a portable GPS device. We ordered 5 meters (16.4 feet) of welded steel tubing along with the device itself to use as a mounting mast. The base of the mast was secured vertically (using metal clamps) along the length of the television antenna, which was already firmly attached to the chimney of the Helfant’s house. The completed assembly also required a grounding kit consisting of a length of copper wiring attached to the bottom of the mast that ran down the side of the house and terminated in a copper rod driven into the ground. Wiring also ran down the side of the house, from the sensor to the data logger, and into the house where the sensor was permanently attached to an indoor computer. Our site selection was ideal for a test site since our data could be constantly checked and retrieved due to the proximity of the computer to the sensor.


    Figure 2: Our Anemometer set-up. In this picture you can see the obstructive tree that we took into account in our data analysis.
  4. Logging Program Information
    We were logging on five second intervals and after the gap we switched to ten second intervals. The data is handled through the Onset HOBOware software for data logger management, data graphing and data analysis. The HOBOware software presents the data in graphical and tabular form, recording gust speed, wind speed, and wind direction for each logged time interval. The data can either be manipulated in this program to display averages and maximums over a specified interval, or it can be exported to an Excel document where the data can be analyzed and errors can be corrected beyond the capabilities of the software (or at least, beyond our understanding of the capabilities of the software). Our month’s worth of data logging yielded approximately 450,000 data points, which we exported to 11 total Excel documents.

    Given the restrictions of needing data to analyze within the time constraints of a semester-long class, we are trying to see if anemometer data taken over a period of time under a year can be reliably compared to a third party data source to extrapolate an estimate for the average yearly wind speed in the anemometer site. In order to do this, we had to search for a reliable source as close to Hamilton, NY as possible that had records of at least daily wind speed averages over the past year. Preferably the site would also have similar terrain features to those at the site of our anemometer.
  5. Data and Analysis
    Once the anemometer data run was complete, we compiled the data to observe wind velocities and compare to known third party data sets. Because of the small time intervals selected (5 sec per point for first logging run, 10 sec per point for second run), we had an extremely large set of data points. In order to reduce the data into a manageable number, we averaged the intervals to five minute increments. The data was then averaged for daily averages (to compare with the intervals available in the third party data) and overall averages. The anemometer total average came to 4.491867079 mph.

    Next, we considered the third party data from both the Colgate University heating plant anemometer and the Hamilton weather station sponsored by Weather Underground. The data from the heating plant anemometer was recorded as a daily average. Records extend from the present day (4/29/08) to roughly a year prior (5/1/07). Over this time, we found an average of 4.5301676 mph. For the Weather Underground information, we took daily averages of the data available online for the time our anemometer was operating. Here we find an average (over our program period) of 0.946153846 mph.

    From a comparison of these averages, a trend can be seen between the three data sets First of all, the averages from the heating plant and the Bonney Hill anemometer are relatively close. We can largely consider these two sites as similar. However, when comparing the Weather Underground averages to heating plant and Bonney Hill anemometer averages, we notice a large discrepancy. The slower velocities recorded by the Weather Underground averages are definitively due to a difference in elevation, as this station is at a lower elevation (1134 ft) compared to the heating plant and the Bonney Hill anemometer. In addition, we suspect a large obstacle interfering with the Weather Underground station, due to the large difference in wind velocity values and the overall slow wind velocities recorded.

    In this vein, we next consider any obstacles that would have affected the Bonney Hill anemometer data points. The first known obstacle that we observe is a large tract of data where the logger froze at zero for the wind velocity and gust velocity. This is first observed from 3/27/08 19:51:44 - 3/28/08 10:33:09 and was caused by the anemometer cups freezing in place.
    After comparing these findings to the GIS analysis of the wind potential of Colgate landholdings done by the geographers in our working group, we would like to erect our anemometer on the Bewkes property, located in the Township of Lebanon next to Seymour Pond. In order to set up our anemometer at this new site, we need to construct a guyed tower that will extend at least the standard 33 feet off the ground on its own (that is, without the added height offered by the Helfants’ house at our first location). Also, ideally, data will be taken at this location for a full year. We will construct the anemometer in the middle of an open field, so obstructions will no longer be a contributing factor in our data analysis. We are currently hesitant to conclude anything regarding the accuracy of extrapolating an annual average wind speed from a smaller period of data collection due to the large number of error factors that contributed to our collected data. Pending permission to construct on the proposed site, this project will begin in late May to early June 2008. Only after annual data is collected at the Bewkes site and after we have done a similar analysis to a subset of the data set will we be able to draw conclusions regarding the accuracy of intrannual extrapolation.


Application Story