Location: Rangeland Resources Research
Title: Comparing two ground-cover measurement methodologies for semiarid rangelands Authors
|Weber, Keith -|
|Chen, Fang -|
|Raza, Mansoor -|
|Serr, Kindra -|
|Gokhale, Bhushan -|
Submitted to: Rangeland Ecology and Management
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: July 12, 2012
Publication Date: January 2, 2013
Repository URL: http://handle.nal.usda.gov/10113/58174
Citation: Weber, K.T., Chen, F., Booth, D.T., Raza, M., Serr, K., Gokhale, B. 2013. Comparing two ground-cover measurement methodologies for semiarid rangelands. Rangeland Ecology and Management. 66(1):82-87. Interpretive Summary: Images such as those from the Landsat satellite, provide extensive views of the earth but little detail of plant communities or soil erosion; whereas, very-high resolution aerial images may show individual plant leaves but the field-of-view (FOV) is small (square yards), requiring users to balance FOV and resolution needs. To help users of aerial-images users understand the tradeoffs between FOV and resolution, we compared common ecological characteristics measured on the ground with similar data measured from images having resolutions of 50 millimeters per pixel (mmpp). Particular care was given to matching locations of ground plots and aerial images. We found (1) using state-of-the-art equipment, location still accounted for 43% or the error between ground and aerial measurements and, (2) 50 mmpp imagery is not adequate for accurate ground-cover measurements. We conclude (1) that while the location accuracy achieved in this project was quite good, the level of accuracy required to match ground and aerial sample points represents an unrealistic expectation with currently available positioning technologies and, (2) the balance between resolution and FOV is best addressed by using multiple cameras to simultaneously acquire at two or three resolutions including 1-mmpp imagery for accurate ground-cover measurements.
Technical Abstract: The limited field-of-view (FOV) associated with single-resolution very-large scale aerial (VLSA) imagery requires users to balance FOV and resolution needs. This balance varies by the specific questions being asked of the data. Here, we tested a FOV-resolution question by comparing ground-cover measured in the field using point-intercept transects with similar data measured from 50 millimeters per pixel (mmpp) VLSA imagery of the same locations. Particular care was given to spatial control of ground and aerial sample points from which observations were made, yet percent cover estimates were very different between methods. An error budget was used to calculate error of location and error of quantification. Budget results indicated location error (0.435) played a substantial role, compared to quantification error (0.216); however, significant quantification error was present. We conclude that 1) while the georectification accuracy achieved in this project was actually quite good, the level of accuracy required to match ground and aerial sample points represents an unrealistic expectation with currently available positioning technologies, 2) 50 mmpp VLSA imagery is not adequate for accurate ground-cover measurement, and 3) the balance between resolution and FOV needs is best addressed by using multiple cameras to simultaneously acquire nested imagery at two or three VLSA resolutions. We recommend ground-cover be measured from 1-mmpp imagery and that the imagery be nested in lower resolution, larger FOV images simultaneously acquired.