The Analytical data are the initial model outputs generated in the production workflow. These data are best suited for users who will carry out their own detailed statistical and uncertainty analyses on the dataset and place lower priority on the visual appearance of the dataset for cartographic purposes. Datasets for the nominal years of 2011 and 2016 are available.
The Cartographic products mask the initial model outputs to improve the visual appearance of the datasets. These data are best suited for users who prioritize visual appearance of the data for cartographic and illustrative purposes. Datasets for the nominal years of 2011 and 2016 are available.
The NLCD data are the result of further processing of the masked data. The goal was to generate three coordinated components. The components are (1) a dataset for the nominal year of 2011, (2) a dataset for the nominal year of 2016, and (3) a dataset that captures the change in canopy cover between the two nominal years of 2011 and 2016. For the NLCD data, the three components meet the criterion of “2011 TCC + change in TCC = 2016 TCC”. These NLCD data are best suited for users who require a coordinated three-component data stack where each pixel’s values meet the criterion of “2011 TCC + change in TCC = 2016 TCC”. Datasets for the nominal years of 2011 and 2016 are available, as well as a dataset that captures the change (loss or gain) in canopy cover between those two nominal years of 2011 and 2016, in areas where change was identified.
These tree canopy cover data are accessible for multiple user communities, through multiple channels and platforms, as listed below:The Hawaii TCC NLCD change dataset is comprised of a single layer. Percent canopy gains are represented as values 0 to 100. Percent canopy losses are represented as values 101 to 200 such that a percent canopy loss of 25 would be represented by the value 125. The background is represented by the value 227 and data gaps are represented by the value 210 since this is an unsigned 8-bit image.
These data were collected using funding from the U.S. Government and can be used without additional permissions or fees. If you use these data in a publication, presentation, or other research product please use the following citation:
USDA Forest Service. 2019. NLCD 2011 to 2016 Percent Tree Canopy Change (Hawaii). Salt Lake City, UT.
Appropriate use includes regional to national assessments of tree cover, total extent of tree cover, and aggregated summaries of tree cover.
The random forests regression algorithm (R Core Team 2017; Cutler et al. 2007; Breiman 2001) employed in creating this product calculates the mean of squared residuals along with percent variability explained by the model for assessing prediction reliability. The random forests models consisted of 500 decision trees, which were used to determine the final response value. The response of each tree depended on a randomly chosen subset of predictor variables chosen independently (with replacement) for evaluation by that tree. The responses of the trees were averaged to obtain an estimate of the dependent variable. Because the random forests bias correction option was used, it was possible to obtain estimates less than 0 or greater than 100. These estimates were reset to either 0 or 100. The estimates were also rounded to the nearest integer. The standard error is the square root of the variance of the estimates given by all trees.
References
Breiman, L. 2001. Random forests. Machine Learning 45:15–32.
Cutler, R.D.; Edwards, T.C.; Beard, K.H.; Cutler, A.; Hess, K.T.; Gibson, J.; Lawler, J.J. 2007. Random forests for classification in ecology. Ecology 88 (11):2783-2792.
R Core Team. 2017. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL www.R-project.org.
Step 1: Disturbance data from the USFS Forest Inventory and Analysis (FIA) program were analyzed to estimate the “percent of area disturbed” between the nominal years of 2011 and 2016, for Hawaii.
Step 2: Pixels were sampled from the 2011 and 2016 “FS-Analytical” TCC products for Hawaii, at 2,000 locations. TCC and standard error values were collected for each sampled pixel.
Step 3: The compiled TCC and standard error values from each sampled pixel were input to a numerical optimization tool. The target of the optimization was the “percent of area disturbed”, as derived from the FIA data described in Step 1 above. The optimization tool was used to determine the condition such that (A) the percentage of the Hawaii sampled pixels identified as loss between the nominal years of 2011 and 2016 matched (B) the “percent of area disturbed” between the nominal years of 2011 and 2016, as derived from FIA data. The conditions estimated using the optimization tool, sampled data from the 2011 and 2016 datasets, and the target were two multiplier values that were subsequently used to determine whether canopy cover had changed or not, between the nominal years of 2011 and 2016 for individual pixels.
Step 4: The Hawaii multiplier values, output from the optimization tool, were applied with an ERDAS model across Hawaii. Specifically, upper and lower bounds of a range of TCC were computed for each pixel, in each year, by (A) using the pixel’s TCC value as the midpoint of the range, (B) setting the lower bound of the range to be the pixel’s TCC value minus the product of the multiplier value and the pixel’s standard error value, and (C) setting the upper bound of the range to be the TCC value plus the product of the multiplier value and the pixel’s standard error value. If the range computed for a pixel in 2011 did not overlap with the range computed for a pixel in 2016, the pixel was identified as changed. If the ranges overlapped, then the pixel was identified as “No change”.
Step 5: For pixels identified as changed, the pixel value in the percent tree canopy cover change dataset was set to the simple difference between the 2016 and 2011 “FS-Cartographic” TCC values for the pixel.
Step 6: For pixels in which confidence in actual change was low or non-existent (i.e., the ranges computed in Step 4 above overlapped and the pixel was identified as “No change”), the pixel value in the percent tree canopy cover change dataset for the 2016 NLCD was set to zero.
Step 7: Results were reviewed by geospatial analysts, especially in areas of severe change (e.g., severe fires, harvest and regrowth in timber areas, etc.), and compared to high-resolution aerial photography and imagery.
Step 8: Spatial filtering was also applied to clean up noise and speckle. While the integrated data stack was achieved, minor visual artifacts (e.g., very small islands of “No Change” surrounded by pixels identified as change) may still be present within the tree canopy cover products included in the overall 2016 NLCD Product Suite.