Thank you, Ecolog! I posted yesterday about how to estimate canopy coverage from GoogleEarth images, and your collective wisdom has lots of good info – don't let anyone say you're just a job board or aggregator of ecological non sequiturs. :)
Several people asked me to pass along the information I received, so here it is. -Jeff My original post: Hello Ecolog – My colleagues and I are studying the reforestation of a reclaimed landfill - from essentially bare-ground to a reasonably dense forest from 1991 to present. I am interested in quantifying changes in percent canopy cover over time using GoogleEarth images. Their archived images include good–resolution growing season photos taken in 1995, 2001, 2005, 2007, and 2010. Does anyone have suggestions as to how to quantify canopy coverage in each photo? Many thanks! -Jeff ________________________________ A very simple software I've used is called i-tree canopy. It uses google images and user classification to quantify canopy cover It could be helpful. The website is here: http://www.itreetools.org/canopy/ ________________________________ I don't know if this will help, but some folks created a methdology called View-IT, using Google Earth to verify some satellite imagery for looking at forest cover Virtual Interpretation of Earth Web-Interface Tool (VIEW-IT) for Collecting Land-Use/Land-Cover Reference Data, Matthew L. Clark 1,* and T. Mitchell Aide 2 http://www.mdpi.com/2072-4292/3/3/601 (open access). ________________________________ Jeff, Since you're working on a fairly small spatial scale, I assume you're looking for a relatively low-tech solution, not a fancy object-oriented processing scheme? I've got colleagues using various methods. One good one that I've seen is to generate a random array of dots in your study area, and just manually go through and count how many of the dots are or are not tree canopy. Another that might be applicable is to use ImageJ (open-source) to outline the tree canopy (either manually or having the software do some of it automatically by color) and calculate the percent of the image. Depends on how much precision you need and how much time you're willing to invest. Some ImageJ info and tips are below, courtesy of Jennifer Soltis. ________________________________ Hi Jeff, Since you're mostly dealing with digital versions of air photos, (as opposed to histogram matched and calibrated satellite data) you're immediately going to be either hand extracting or "photo-interpreting" the extent of the canopy layer within each of the images or (potentially) using an automated (machine segmentation program) like eCognition, which, depending on a range of factors contained within each image set (lighting, texture, time of year, resolution, parallax, etc....).... Either way, you're best off by having the same person(s) perform the analysis so that areas considered to be "FOREST" can be more uniformly assessed and mapped. (at some point you have to decide "how big does a tree or forested patch have to be, to be considered "FOREST".... with a virtually uniform gradient, spatially as well as temporally, from BARE SOIL to MATURE FOREST, that break line can be tricky to hit. You'll have to be aware of a number of error factors within the work; how well each year\image epoch was "georeferenced", (or superimposed within a real-world feature space) both to real world features as well as each other... whether elevation factors were included in the correction process (ala orthocorrection processes to remove terrain displacement), etc... All of these can significantly affect the area metrics you compile for each image set. And that's all assuming that you can somehow thread the data of the respective image layers that are served up via Google Earth, into an appropriate image processing or GIS capture software array. In my opinion, you certainly will not want to attempt the exercise within Google Earth itself. While single point, line or polygon feature creation is supported in the software, to attempt to extract large areas of adjacent feature layers would be an extraordinary challenge (it's really not intended to serve as a GIS, but as a simple "geobrowser" or tool to look at things). But if you have access to software like ArcGIS, there very likely may be a way to gain access to the imagery (often in a higher resolution and "clearer" version) as an online feature or map service which NYS and other agencies host. How big is the area you are looking at? There are a number of satellite based end products as well as potential input layers (e.g. Landsat) which, at 30m resolution, you might be able to get some sense of the general trend (there are also versions of certain products that measure change in land cover over time...). ________________________________ There are several methods you can use. See Nowak, D.J., R.A. Rowntree, E.G. McPherson, S.M. Sisinni, E.R. Kerkmann, and J.C. Stevens. 1996. Measuring and Analyzing Urban Tree Cover. Landscape and Urban Planning 36:49–57. for an older approach that still gets a fair amount of use today, examples include: Nowak, D. J., & Greenfield, E. J. (2012). Tree and impervious cover change in U.S. cities. Urban Forestry & Urban Greening, 11(1), 21–30. doi:10.1016/j.ufug.2011.11.005 and Nowak, D. J., & Greenfield, E. J. (2012). Tree and impervious cover in the United States. Landscape and Urban Planning. doi:10.1016/j.landurbplan.2012.04.005 Are you familiar with NDVI? You may consider using Google Earth Engine to calculate NDVI for you selected images. Your academic institution can request a free license, I believe. ________________________________ Hi Jeff, I went through this process conceptually and as well as using existing riparian forest restoration sites for the Bay/Delta HCP. In the end it was decided that the permitting agencies would accept ground sampling and I provided estimated height growth rates. Given soil differences that would lead to variation in growth rates and expected patchiness due to a range of establishment issues we opted out of using imagery. 1) Your cover metric does not include the overlap over sub-canopy cover. LiDAR is typically used to obtain that with varying degrees of success. 2) As John Mickelson pointed out, there are all kinds of spatial decisions that will affect your results as well as classification errors introduced by different observers and different imagery. 3) If you are only interested in doing this in a qualitative way you could use Photoshop to convert "green" to black pixels and everything else to white pixels and run a processing program, I used NIH Image, to get the relative amount of green vs other. If you know the area of the restoration project. This assumes that the terrain is almost flat and that the green pixel size is significantly smaller than the patchiness of the forest. This might be good enough for monitoring the success of the restoration in a coarse way for permit purposes if you can convince the agencies. 4) Note that Microsoft also offers similar imagery and sometimes the dates of the imagery could supplement your GoogleEarth imagery. 5) There is a vast forestry literature on measuring and modeling the growth of even age stands that you may want to take a look at but typically single species are involved and the statistics requires some kind of time sequence data. Talk to your foresters back east. ________________________________ This article may be helpful. Citation: Duhl, T.R., A.B. Guenther, and D. Helmig, 2012: Estimating urban vegetation cover fraction using Google Earth® images. Journal of Land Use Science, 7, 311-329, DOI: 10.1080/1747423X.2011.587207. ________________________________ You can try Google's planetary-scale platform for environmental data & analysis. http://earthengine.google.org/ Google Earth Engine brings together the world's satellite imagery — trillions of scientific measurements dating back almost 40 years — and makes it available online with tools for scientists, independent researchers, and nations to mine this massive warehouse of data to detect changes, map trends and quantify differences on the Earth's surface. Applications include: detecting deforestation, classifying land cover, estimating forest biomass and carbon, and mapping the world’s roadless areas ________________________________ @Jeff_d_corbin Image J software should work. It's free:) ________________________________ ************************************ Jeffrey D. Corbin Associate Professor Department of Biological Sciences Union College Schenectady, NY 12308 (518) 388-6097 http://jeffcorbin.org ************************************ ________________________________
