Land Cover Trends Project


Overview/Summary  •   Ecoregion Framework  •   Sampling Strategy  •   LandSat Source Data   •   Classification System  •   Interpretation Process  •   References

Sampling Strategy

LULC composition and change was determined using a probability sample of 2,688 blocks randomly selected within 84 EPA Level III ecoregions. The probabilistic nature of the sampling strategy created derived estimates categorically "representative" of the population (Kish, 1987). The size of each sample block and the number of samples in each ecoregion was established by trying to strike a balance between estimating change in LULC area and estimating change in landscape pattern. The initial design used a 20-km square sample size. Nine ecoregions were analyzed using 9 to 11 randomly selected 20-km by 20-km sample blocks per ecoregion. The remaining 75 ecoregions used 25 to 48 randomly selected 10-km by 10-km sample blocks because a greater number of smaller samples would favor better estimates of change in area and type over better characterization of change in landscape pattern.

The number of samples was determined by an expert panel based on expected land-change characteristics and a desire to measure change within +/-1% margin of error at an 85% confidence level. Sample selection involved overlaying a regular grid of 10-km by 10-km (or 20-km by 20-km) blocks over the conterminous U.S. and assigning blocks to a specific ecoregion following a centroid spatial allocation rule. The blocks within each ecoregion were then assigned a unique numerical value from 1 to N. A random number generator was used to select actual sample blocks one at a time until the total number of desired blocks was met.

Accessibility FOIA Privacy Policies and Notices

Take Pride in America logo logo U.S. Department of the Interior | U.S. Geological Survey
Page Contact Information: WRG Web Team
Page Last Modified: Thursday, 07-Jul-2016 11:19:52 EDT (mg)