Comparison of Canopy Cover Estimation Using
Densiometer and Photo Analysis Software
Katherine Fergot
Introduction
Methods
Results
Canopy cover can heavily influence the understory and other biological
components of ecosystems, such as productivity, composition and structure.
The densiometer has been viewed as an acceptable measurement tool by
the US Department of Agriculture since the 1950’s. However, advancements
in technology have led to new methods of canopy cover assessment that
are less expensive, and less time consuming, while maintaining precision
and decreasing observer bias. One such method involves capturing photos
of the canopy with a digital camera with a basic lens and then using photo
analysis software to determine proportion canopy. The program CanopyDigi
yields results that are consistent between camera types and produces
results that are correlated with the estimates produced by a moosehorn,
ocular estimations, and a sighting tube. There are currently no studies that
report on the relationship between regular photo estimates like those from
CanopyDigi and those of the densiometer.
Data was originally collected from 2011 through 2013 in the Loomis
State Forest and the Okanogan National Forest in north-central
Washington
No outliers were found using the dfbetas
Densiometer Data
Estimates of canopy were determined for each plot by taking the
uncovered, or sky measurements, and subtracting it from 96 to get
canopy estimates, then averaging across the four cardinal directions
for each plot and dividing by 96.
Photo Data
Images were converted into greyscale bitmaps and then uploaded
into the CanopyDigi program. The program ran eight different
thresholds on each image. Then the canopy cover estimate for each
image was calculated from the most appropriate threshold
Analytical Methods
The CanopyDigi Software takes the black and white bitmap and assigns
pixels a greyscale value between 1 and 255 before converting the bitmap
into the false color image based on a given threshold
Ran dfbetas to determine if outliers existed
Wilcoxon signed rank test to determine if means were equal
Ran Pearson’s product-moment correlation
Determined if data met assumptions of linear regression
Determined if transformation was appropriate based on qq plots
and Shapiro-Wilk normality test
Ran offset model and 95% confidence interval on slope to
determine in relationship significantly different than 1 to 1 given
non zero intercept
Objectives
• To estimate the canopy cover for 220 plots using both the densiometer
data and the canopy photographs
• To determine if there is a difference between the densiometer estimates
and the photo estimates
• If there is a difference, determine the relationship between the methods
and if conversion is possible
Hypotheses
Null: There is not a linear relationship between the densiometer estimates of
canopy cover and the photo analysis software estimates of canopy cover
Alternative: There is a linear relationship between the densiometer
estimates of canopy cover and the photo analysis software estimates of
canopy cover
Fiala, A. C. S., Garman, S. L., & Gray, A. N. (August 15, 2006). Comparison of five canopy cover estimation techniques in
the western Oregon Cascades. Forest Ecology and Management, 232, 188197.
Gonsamo, A., D'odorico, P., & Pellikka, P. (2013). Measuring fractional forest canopy element cover and openness
definitions and methodologies revisited. Oikos, 122(9), 12831291.
Goodenough, A. E., & Goodenough, A. S. (January 01, 2012). Development of a Rapid and Precise Method of Digital
Image Analysis to Quantify Canopy Density and Structural Complexity. Isrn Ecology, 2012, 3, 111. Korhonen, L.,
Korhonen, K., Rautiainen, M., & Stenberg, P. (January 01, 2006). Estimation of forest canopy cover: a comparison of
field measurement techniques. Silva Fennica,40, 4.
Symstad, A. J., Wienk, C. L., & Thorstenson, A. D. (2008). Precision, repeatability, and efficiency of two canopycover
estimate methods in northern great plains vegetation. Rangeland Ecology & Management, 61(4), 419429.
Zagalikis, G., Cameron, A. D., & Miller, D. R. (2005). The application of digital photogrammetry and image analysis
techniques to derive tree and stand characteristics. Canadian Journal of Forest Research, 35(5), 12241237.
Densiometer estimate = 17.216 + Photo estimate
I would like to thank Aaron Wirsing and Laurel Peelle for their guidance and support. I would
also like to thank Laurel for the use of her densiometer data and canopy photographs
Pearson’s product moment correlation
T=23.3147, p-value<2.2e-16, cor=.8789
The correlation was not equal to zero and the data had strong
positive linear relationship
Wilcoxon signed rank test
W=22708, p-value<2.2e-16
The means for densiometer estimates and photo estimates were
not equal
Offset Anova
Fail to reject that the slope was different than 1
Intercept was different than 0
Source
Estimate
Intercept
Photo
estimates
17.216
-0.03984
Offset Anova
Standard t value
Error
2.498
7.085
0.0412
-0.967
p value
1.89e-11
0.334
Discussion
As a result of the linear relationship and the high correlation
between the densiometer and photo estimates, it was possible
to estimate a conversion equation between the two methods.
The equation that best fit the data was
Densiometer estimate = 17.216 + Photo estimate.
There was no significant difference between the slope and 1,
however, there was a significant difference between the
intercept and zero. This suggests that the difference between the
densiometer estimates and the estimates produced from the
photos is relatively constant over the range of canopy cover
measured. The need for an intercept different than zero suggests
that the photo estimates are underestimating canopy cover
when compared to the densiometer. It was expected that the
photo estimates would underestimate canopy cover as methods
it is correlated with, such as the moosehorn, also produce lower
estimates of canopy cover than a densiometer. While there was
slight departure from normality, the robust sample of 220 comes
from an approximately normal distribution because of the
Central Limit Theorem. However, if the sample size was below
50, normality can not have been assumed and the conversion
equation may not have been valid. As a result of the large
sample size and the other assumptions of a linear relationship
met, conversion between the two methods is possible. Thus,
canopy photos may replace the densiometer in the field and
reduce time spent collecting canopy data. Conversion between
methods allows for the photo estimates to be converted to
densiometer estimates that are more holistic, less variable and
more similar to animal perspectives.