The Rural Technology Initiative ceased operations in 2011. This site is maintained as an archive of works from RTI collaborators from 2000 to 2011 and is no longer updated.
In 2004, the USDA Forest Service’s Forest Inventory Analysis (FIA) program contracted with the Rural Technology Initiative (RTI), a research and outreach center at the University of Washington’s College of Forest Resources, to investigate the potential of using satellite images as a more cost-effective and repeatable process to determine and calculate land use and its associated change.
The project goals included:
This project focused on non-federal lands, since it was assumed that land use change on federal lands has stayed constant over time, whether or not land cover has changed (for a detailed discussion of land use versus land cover, click here).
The methodology, analysis, and data were completed in September 2006 and were provided back to the FIA program staff, who are currently comparing the data against other sources. Earlier in 2006, we attempted to replicate the methodology for Eastern Washington, but were confronted with a large obstacle to make the methods appropriate for the very different landscape east of the Cascades. This follow-up to the project has been put on hold until additional time and funding allow for more in-depth analysis.
This website will provide interested people and organizations with a summary of our methodology and analysis, as well as a few summary tables and links to the actual digital data. We are currently working with the FIA program and the PNW Research Station to pursue a continuation of land use change analysis in Washington. We encourage you to read through the background and methodology sections before viewing the summary tables and/or using the digital data.
Using Landsat images for land use analysis
Although Landsat images were chosen for this analysis, there are alternative sources of data that may improve future classifications and land use analysis. The difficulty of registering and classifying such a large quantity of images posed problems with data storage, transfer, and repeatability. Future research could focus on using alternative source data and compare to the results using Landsat images.
Using eCognition and object-based classification methods
Object-based classification proved to be a successful method for determining land use (as compared to land cover). eCognition's segmentation capabilities were a key factor in the object-based classification; however, the program was unpredictable and alternative classification software could yield more reliable results.
Additional uses for this data
This data was produced to supplement current FIA efforts of monitoring and tracking forest land resources in the Pacific Northwest. It was a first attempt to use satellite images and remote sensing, rather than aerial photo interpretation, as the primary data source and method of analysis. The authors are aware of the limitations of the data and hope that future work will build off of this first-level foundation.
Some of the goals of this project were to identify a way to more quickly classify land use and classify land use more consistently. Previous photo interpretation techniques and methodologies have been well defined but different analysts have produced different results. Image classification is subject to that same subjective bias when analysts collect samples for a supervised classification. For this Landsat analysis classification training datasets were produced for each image. While this significantly reduced the amount of time and effort necessary to classify the images (since spectral properties did not need to be normalized across time and space) it did introduce analyst bias into the results. Future classifications could use the same spatial locations for training data across multiple datasets. Overlapping areas of imagery and areas that have not changed from year to year would be better locations to take training samples than the unstructured method used for this project.
Western Washington Land Use Change 1988, 1996, 2004 methods, data products and report were authored by:
with much support from Justina Harris, Phil Hurvitz, and Hiroo Imaki!