Visualize Complexity, Discover Solutions, Shatter Limits
|
|
Visualize Complexity, Discover Solutions, Shatter Limits |
|
Case Study: Oregon |
|
|
Client: |
Oregon’s Department of Environmental Quality and Watershed Enhancement |
| Project: |
Stream Temperature Impairment
Western Oregon: streams, gauge locations, and cluster assignments |
|
Issues:
|
Oregon’s Department of Environmental Quality and
Watershed Enhancement Board relies on computer models to evaluate the impact
of growth and development on the state’s famous natural areas. These
agencies collaborated with the USGS (Portland, OR) in an extensive field
data acquisition program to collect long-term time series of stream
temperatures from 160 sites throughout the western third of the state, an
area of some 60,000 square miles. The gauging sites were carefully chosen to
insure that they were unimpaired by man-made causes. They were also selected
to represent a wide range of conditions, e.g., shaded/unshaded,
mountains/plains, etc. Augmenting the stream temperature data were ambient
conditions from several weather stations, snow pack depths, and gauging site
characteristics (elevation, orientation, streambed morphology, etc).
|
| ADMi's Role: |
Using signal processing, time series clustering, and
machine learning, USGS and
ADMi researchers created a model comprised of optimally
configured sub-models to predict what the unimpaired stream temperature
would be for any location within the study area.
For a given site, the model computes temperatures to which
field measurements can be compared to assess the degree of thermal
impairment. The model is now being converted into a user-friendly
application for final delivery.
|
|
Results:
|
An interesting finding from the work is that the USGS’s formal protocol for characterizing gauging sites is flawed for purposes of Data Mining. It requires that multiple readings be taken, which have some spatial separation. However, the readings are then averaged, hindering the potential for better characterizing a site from the un-averaged data. In general, making assumptions about data before analyzing it can diminish the chances of attaining the best possible result. |
|
Home | Products | Services | Training | Technologies | News | Links | Contact |
Send mail to
info@advdataming.com with questions or comments about this web site.
Copyright © 2003 Advanced Data Mining, LLC
