home > data and maps > foundational data > canopy cover detail

ForestERA Data Layer Details - Percent Canopy Cover

Note to data users: Please carefully review the metadata provided with each layer. We request that users consult with the ForestERA project in advance of using these data in publications and/or presentations to ensure that the strengths and limitations of the data are considered.

Description

Canopy Cover is a measure of the amount of surface area that is covered by overhanging vegetation. This differs from canopy closure which is a more detailed analysis of the canopy that, when undertaken, includes small gaps within areas of continuous canopy in cover calculations. These small gaps are not taken into account when determining canopy cover. This layer is provided in percent and has a resolution of 90m (0.8 ha or 2 acres).

Purpose

This data layer was created as part of the ForestERA project to support landscape-scale forest restoration planning efforts by a broad group of stakeholders including federal and state agencies, academic institutions, and non-governmental entities.  These data are intended for regional analyses over spatial extents on the order of tens to hundreds of thousands of acres, and were not developed for use at finer spatial scales, although they may be useful for some applications at finer scales.

Development

The canopy cover layer was developed from mosaics of Digital Orthophoto Quads (DOQs) using advanced exploratory data analysis (Xu, et al., in review).  DOQs are a high spatial-resolution data source (1 m ground cell resolution) that have been used for years in generating base maps or reference datasets for a variety of spatial analysis projects (Hohle, 1996; Greenfeld, 2000; Huang, et al., 2001; Xu, et al., in review).  Our methodology, built on the basis of a normal distribution model and a fractal distribution model, allows for discrimination between areas of tree crown, shadow, and ground vegetation over a large area based on characteristic of the signature from the DOQ imagery. 

This layer was initially developed at 1m resolution, each of pixels of which was labeled tree crown, shadow, or other.  Canopy cover maps at different scales then can be generated based on the 1m resolution thematic map. For the landscape level assessments required by ForestERA, canopy cover layers with resolutions of 30m and 90m were selected as the final products. Most of the DOQs used to create the canopy cover layers were generated in 1997, but in a few areas the imagery is dated 1992.  Due to this, recent changes in forest structure in some areas (e.g., due to fire or timber harvest) will not be represented in the canopy cover layers.

Accuracy Assessment

We assessed the accuracy of this layer in two ways, using ground plot data, and by visual comparison of the layer with the original DOQ imagery.

1. Accuracy assessment using ground plot data
   
   We used canopy cover estimated from two hundred 17.6m radius plots of ground data.  These data were collected by Norris Dodd of the Arizona Game and Fish Department and his field crew at 18 study sites spread across ForestERA’s 2 million acre Western Mogollon Plateau focal area in northern Arizona. Canopy cover estimates from ground data can vary dramatically depending on methodology used to collect the data and the number of samples taken. We used this dataset due to the strict methodology used in its collection and its distribution across a large landscape.  For the ground data canopy cover estimates, observers measured canopy cover by vertical projection, using a staff-mounted, self-leveling sighting periscope (Ganey and Block 1994, Dodd et al. 2003).  Observers recorded periscope cross-hair interceptions with canopy foliage >2 m height at 100 locations on each of the 200 plots.  For more information on this methodology see Dodd et al. (2003).
   
   For the assessment the center of each ground plot was georeferenced using a Trimble GPS unit.  We then buffered each point location using the 17.5m radius to identify the area in which the ground data had been taken.  The canopy cover estimates from these areas were then compared to estimates from our DOQ-derived layers over exactly the same area.
   
   Linear regression was used to determine the relationships between the canopy cover as estimated by the ground data and the canopy cover as estimated by the DOQ imagery.  As the two measures of canopy cover are expected to be equal we used the best-fit line of a regression with intercept constrained to zero.  Using this line results in a lower r² value than that obtained from using the overall best-fit line.  However, analysis of the slope of the line with intercept constrained to zero allows for detection of bias in the predictive dataset.  Under conditions of no bias, the slope of this line should be equal to one.
   
   The results of this analysis indicated that a highly significant relationship exists between the two measures (Fig. 1; r² = 0.545, P < 0.0001).  The slope of the line from this relationship is 1.02 indicating that the DOQ derived canopy cover estimate is nearly unbiased with relation to the ground estimate.  Based on the actual differences in the data between the two estimates we have determined that values for 50% of the pixels in the DOQ derived canopy cover layer will be within 9% of the actual values for canopy cover in those areas, and values for over 80% of the pixels will be within 16% of the actual values.  Actual accuracy may be higher as this assessment is based on the assumptions that the ground data values are absolutely correct, and that the ground and DOQ datasets are perfectly co-registered. Neither of these assumptions is likely to be completely true in the real world, and as a consequence the accuracy of DOQ derived canopy cover layer will be an underestimation.
   
   

   Figure 1:  Linear regression showing the relationship between our DOQ-derived canopy cover estimates and ground estimates of cover at the same locations.  Solid line is the best-fit line with intercept constrained to zero.  Dashed lines are the 95% confidence intervals for the best-fit line.

   
   The area encompassed by a 17.5m radius circle (962 m2) is almost exactly the same the area covered by a 30m pixel (900 m2), so the accuracy using this methodology should very closely approximate the accuracy of each 30m pixel.  Accuracy of the 90m resolution canopy cover layer would be expected to be equally high or higher.
   
   
2. Accuracy assessment by visual inspection
   
   DOQs are a type of imagery that can be used to visually identify tree crowns with relative ease.  Therefore, the canopy cover layer can be overlaid on screen with the original DOQs for comparative purposes.   Visual evaluation of the canopy cover layer by experienced aerial photo interpreters and local forest exports suggest the quality is high.

Sources of errors

We believe the canopy cover is the best vegetation structure data layer available for the Western Mogollon Plateau.  Errors are likely to come from only three sources.  First, small portions of the landscape may have changed since the time the DOQs were taken due to fires, forest management, or other disturbances.  Our canopy cover estimates reflect what was present at the time the images were taken and will not reflect changes since the imagery was taken.  Second, saplings and large shrubs are easily mistaken for trees in DOQs.  Thus, canopy cover will be overestimated in areas with large numbers of shrubs and/or saplings.  Finally even with the advanced measures used here, some areas of shadow cannot be completely differentiated from tree crowns.  In areas where shadow and crown are mixed together actual canopy cover may be overestimated or underestimated slightly depending on the actual ratio of shadow to crown.

Recommendations

We recommend that this layer be used at a minimum resolution of 30m (approximately 0.1 ha or 0.25 acres) for purposes of analysis and display.  However, ForestERA data layers were not designed for analyses at the level of individual pixels, and uncertainty in the data will generally decline over greater spatial extents.  Therefore, we recommend using larger analysis units, with groupings of at least 50 cells (4.5 ha or 12.5 acres) being optimal.  Finally, we reiterate that ForestERA data layers were developed for the purpose of regional landscape-level planning, and we suggest that the analyses be applied over spatial extents of tens to hundreds of thousands of acres.  We recognize, however, that this layer may be useful for analyses over smaller spatial extents depending on the type and purpose of those analyses.

References

Dodd, N. L., J. S. States, and S. S. Rosenstock.  2003.  Tassel-eared squirrel population, habitat condition, and dietary relationships in north-central Arizona.  Journal of Wildlife Management 67: 622-633.

Ganey, J. L. and R. P. Balda.  1994.  Habitat selection by Mexican Spotted Owls in northern Arizona.  Auk 17: 162-169.

Hohle, J.  1996. Experience with the production of digital orthophotos.  Photogrammetric Engineering & Remote Sensing 6: 1189-1194.

Greenfeld, J.  2001.  Evaluating the accuracy of digital orthophoto quadrangles (DOQs) in context of Parcel-based GIS.  Photogrammetric Engineering & Remote Sensing, 67: 199-205.

Huang, C., L. Yang, L., B. Wylie, and C. Homer.  2001.  A strategy for estimating tree canopy density using Landsat 7 TM+ and high resolution images over large areas, proceeding of the Third International Conference on Geospatial Information in Agriculture and Forestry, Denver, CO.  CD-ROM.

Xu, B., P. Gong, and R. Pu.  2003.  Crown closure estimation of oak savannah in a dry season with landsat TM imagery: comparison of various indices through correlation analysis.  International Journal of Remote Sensing 24: 1811-1822.

Xu, Y., J. W. Prather, H. M. Hampton, E. N. Aumack, and T. D. Sisk.  (in review). Advanced exploratory data analysis for mapping regional canopy cover. Photogrammetric Engineering & Remote Sensing.

Last updated February 23, 2005