GIS Applications in Forestry

Geography 560 – Joseph Kochlacs

GIS Applications within Forestry

As I re-enter the world of academic geography and GIS at the graduate level at Oregon State University, my primary interest is in Geographic Information Systems applications in forestry. The following articles and studies provide relevant insight into how GIS is being used for solutions to complex problems within the natural world, and in particular, forestry.

The study, "Modeling Forest Fire Spread Using Machine Learning-Based Cellular Automata in a GIS Environment" examines the efficacy of a new model for predicting fire spread by using Cellular Automata, or how cells interact with their surrounding cells in a grid while incorporating LSSVM (least squares support vector machines). There are rules for if and how quickly fire will spread from cell to cell using dependent variable data (fire point) and independent variable data (slope, aspect, elevation, relative humidity, normalized vegetation index). These rules help the LSSVM classify whether a cell is burning (1) or not (0), but with another delineation of (0,1) for what is essentially a forest fire cell almost or just starting to burn. Using the LSSVM-CA model, the authors tested the model against a real forest fire in China and found the model results were within 97% of the number of land (grids) burned in actuality. I find this fascinating as these models can help in predicting the direction of fire, speed, as well as the volume of land expected to be burned. It can help in prevention by establishing where local fires can best be fought, using local weather patterns and ground data to create the best possible places for establishing fire roads and fire responder plans. The model however is only as good as the local data as so many factors play into its accuracy.

Xu, Yiqing, et al. "Modeling Forest Fire Spread Using Machine Learning-Based Cellular Automata in a GIS Environment." Forests, vol. 13, no. 12, Nov. 2022, p. NA. Gale In Context: Environmental Studies, link.gale.com/apps/doc/A745708804/GRNR?u=s8405248&sid=bookmark-GRNR&xid=5973b5de. Accessed 2 Dec. 2023.

In this study, Xiao and McPherson seek to quantify the health of campus trees at UC Davis using remote sensing. The methodoly used can help provide a good template for using GIS within the field urban forestry. A tree inventory layer was created based on a tree survey which included location, species, and dimensions. A raster layer was used containing NDVI (vegetation index) that used Infrared and near infrared (NIR) sensing equipment to determine what ground cover was vegetation because in the NIR, vegetation has a high reflectance. Using the previously made survey, the authors were able to determine what vegetation corresponded to which trees. Then different types of trees were assigned different percentages of their canopy range which needed to have “healthy” pixels to correspond with leaf type and other delineating factors. The tree health index defines health by ratio of healthy pixels to entire tree crown pixels for a given species, because healthy vegetation has a higher NDVI value than unhealthy vegetation. A tree was designated unhealthy if 30% or more of the pixels were unhealthy and the average NDVI value was less than the threshold for healthy per the tree species. 86% of the trees were deamed healthy using this methodoloy. The methodology found in this study could be used in a variety of forestry and urban forestry applications, to learn about the health of trees in a given area. One could use the health findings along with other data in GIS to determine what factors lead to tree health, such as soil type, local hydrology, slope, aspect, pollution and other features. The findings could be used to plan for urban forestry planting, or even larger projects such as planting redwoods into Oregon as climate change progresses and habitat types change.

Xiao, Qingfu, and E. G. Mcpherson. "Tree health mapping with multispectral remote sensing data at UC Davis, California." Urban Ecosystems, vol. 8, no. 3-4, 2005, pp. 349-361. ProQuest, https://oregonstate.idm.oclc.org/login?url=https://www.proquest.com/scholarly-journals/tree-health-mapping-with-multispectral-remote/docview/758418058/se-2, doi:https://doi.org/10.1007/s11252-005-4867-7.

The article “Constraints on Mechanized Treatment Significantly Limit Mechanical Fuels Reduction Extent in the Sierra Nevada” seeks to find how much land can be accessed with large machinery to reduce fuels in the Sierra Nevada Mountain Range. The methodology used factors such as slope, sensitive species habitat, riparian zones, and other protected areas as constraints, which were mapped using GIS. Another consideration was the possibility of timber sales to offset the cost of using heavy machinery, so species and sizes were added to the equation. Because certain machinery cannot come within close distance of streams and rivers, some steeply sloped areas were ruled out, but an assessment was made that mechanized fuel reduction can work as an anchor point for hand crews in such areas. Overall a very small percentage of forest lands in the Sierra Nevada were suitable for mechanized fuel production because 46% of the land has a high level of constraint, 37% a moderate level of constraint, and only 20% a low level of constraint. The findings suggest mechanized fuel abatement alone is not a promising way to reduce fuels for forest fire, but that machinery can create base areas which are well protected from which hand crews and prescribed burns start from. I find the methodology and analysis interesting because the amount of constraints an area may have is more than I imagined. Finding all the layers for things such as owl habitat, streams, rivers, types of trees, their size, density, slope, prevailing wind, areas under study, etc seems like it would be an enormous task requiring conferring with multiple agencies and experts. It is nonetheless inspiring to see what can be done with GIS and forestry.

North, Malcolm, et al. “Constraints on Mechanized Treatment Significantly Limit Mechanical Fuels Reduction Extent in the Sierra Nevada.” Journal of Forestry, vol. 113, no. 1, 2015, pp. 40–48, https://doi.org/10.5849/jof.14-058.

In the article, “The Effects of Spatial Patterns on the Accuracy of Forest Vegetation Simulator (FVS) Estimates of Forest Canopy Cover,” the authors seek to find the difference between Forest Vegetation Simulator (FVS) estimates and a model for estimating canopy cover using GIS. The GIS model uses a stem plot (each tree is a point) whereas the FVS doesn’t use spatially explicit information. The FVS model does have inputs of the number of trees, their rough size and their rough distribution, but doesn’t know exactly where the trees are. The distribution pattern types are clustered, random, and even or regular. After running the GIS data which has the actual locations of the canopies versus the FVS, the findings show an 11% difference for regular distribution, where the FVS underestimates. In the case of clustered stands of trees, the FVS overestimates by 2%. This seems to be because a tree that is not sharing a canopy sprawls out to gain more sunlight, while trees in denser stands have more compact canopies. The models used in this show how important it is to use actual data, but also that the FVS does a pretty good job, as 11% is not that far off when you consider the model doesn’t actually plot known trees. Both of the models could be used in my work in the future depending on the data available with an acceptably high degree of accuracy.

Christopher, T. A., and J. M. Goodburn. “Effects of Spatial Patterns on the Accuracy of Forest Vegetation Simulator (FVS) Estimates of Forest Canopy Cover.” Western Journal of Applied Forestry, vol. 23, no. 1, 2008, pp. 5–11, https://doi.org/10.1093/wjaf/23.1.5.

In this article, the author, Justin Kurland analyses the spatial pattern of redwood burl poaching in the Redwood National and State Parks through the lenses of crime pattern theory and environmental GIS. Redwood burls have been poached with greater frequency over the last couple of decades, and this article sets out to figure out what factors lead to certain areas being more susceptible. The findings are interesting in that they suggest usual factors for crime, such as criminal disposition, or familiarity with the area of their crime are less important than usual in crime theory. The input factors used include sites of known poaching/cutting, the likelihood of ranger patrolling, proximity to roads, size of tree, and proximity to a burl shop/potential buyer. The findings suggest that possible locations of trees with burls were less important than proximity to a road and a buyer, and especially potential trees in close proximity to multiple burl shops. Suggestions for decreasing poaching include not only added ranger patrolling but also the addition of campsites and parking lots, which would draw in tourists as guardians, and ticketed gates with cameras into the high-risk areas. I find this paper fascinating because it also lays out how other conservation crimes can be mapped, analyzed and prevented, using surveillance, citizens as guardians and a variety of other methods. I think one can even use the results and models created to use GIS to reverse engineer the location of illicit flora and fauna shops or dealers which could curb poaching of highly sought after items such as rhino tusks and certain succulents. Authorities could then search those areas for dealers.

Kurland, Justin, et al. “The Spatial Pattern of Redwood Burl Poaching and Implications for Prevention.” Forest Policy and Economics, vol. 94, 2018, pp. 46–54, https://doi.org/10.1016/j.forpol.2018.06.009.

It is widely accepted that manmade climate change is occurring as a result of increases in carbon emissions, and that trees can help offset these emissions by storing carbon in the form of wood. In this article, Helen M Cox discusses the benefits of sequestering carbon in on-campus trees at universities. Students at CSU Northridge used GIS to map the 3,900 campus trees and calculate the amount of carbon being sequestered. The tables kept track of growth over time, as well as the condition or health of the tree and other factors such as species, height, diameter at breast height, etc. The trees were made into points in vector format and overlaid over a photo raster base layer to help visualize which canopies belong to what trees. The Urban Foresty Research Carbon Tree Calculator was used to determine how much carbon each tree sequestered. A total of 1000 hours went into collecting the data, storing it in easily accessible databases, and calculating/analyzing the data. The total amount of carbon sequestered was 154 tons per year, which was found to be only a fraction of the commutes of 150 students. This is a discouraging finding, however Northridge is not the best growing environment with very low average rainfall. The findings don’t go into how much the trees protected the urban environment of the campus from urban heat island effect, so the presence of trees may do more to combat climate change than simply by sequestering carbon. I find this study fascinating and would love to do similar research on my local University. UC Santa Cruz, which is heavily forested. The methodology is inspiring, regardless of the results on this particular campus. I also have a keen interest in knowing how much carbon is offset in urban forests as I have owned and operated a business milling downed urban trees into lumber for the last 10 years. I would love to be involved in tree planting programs in the future to continue to sequester carbon in urban forests.

Cox, Helen M. “A Sustainability Initiative to Quantify Carbon Sequestration by Campus Trees.” Journal of Geography (Houston), vol. 111, no. 5, 2012, pp. 173–83, https://doi.org/10.1080/00221341.2011.628046.

California forests have long been subject fire, so much so that many species are evolved to withstand or even flourish with occasional fire. For example, certain tree cones need fire to open and germinate, but too hot of a fire can kill even those, so it is a complicated balance for the ecosystem. This study by Richard G. Everett uses burn scars on tree rings or their exteriors, mapped in GIS, to learn how fires effected the landscape of the San Jacinto Mountains. This method of using tree rings to record natural history is called fire-scar dendrochronology. The burn scar samples were often taken from standing dead trees and then cross referenced at a research center in Redding CA to learn what year time of year the fire happened to see if burn scars from different trees corresponded with the same fire. By doing so, the authors were able to map the size of a given fire that had happened long in the past, to gain knowledge of how large natural fires in the area were. The results found a 653 years long chronology, indicating a point mean fire return interval of 5.2 years, and an area wide grand mean fire interval of 32.2 years while the fires were usually at or under 18 hectares in size. I find this work fascinating because you can use the methodology to learn what to expect of a fire in a given area. This can help with planning for forest fires including how and where to fight, along with prevention using other local factors such as prevailing wind, slope, fuel etc. The results can also be compared to modern fires, which tend to be more devastating due to fuel build up and higher temperatures or drought as a result of climate change. Analyzing and mapping all these variables can help plan for the future in both physical and policy contexts such as where to build roads and hold controlled burns. Forestry practice such as thinning, fuel removal and leaving behind larger trees could help forests survive if we know the past of local fires.

Richard G. Everett, Dendrochronology-based fire history of mixed-conifer forests in the San Jacinto Mountains, California, Forest Ecology and Management, Volume 256, Issue 11, 2008, Pages 1805-1814, ISSN 0378-1127, https://doi.org/10.1016/j.foreco.2008.04.036. (https://www.sciencedirect.com/science/article/pii/S0378112708003411)

Forest Fires are common in the western United States, and trees of course die as a result, but not always at the time of the fire or even within the first year post-fire. Matthew J Reilly attempts to use remote sensing techniques to assess how many trees die in the 5 years following a fire. Because tree health can be mapped using infrared and near infrared bands to determine if they are photosynthesizing, the percentage of living trees can be recorded as data. In this study, percent of live canopy before the fire, 1 year after fire, and 5 years after fire to track post-fire delayed tree mortality. 900 plots representing over 7,000 hectares of land were studied for 6 different large western fires to gauge different species' reactions to fire over the subsequent 5 years post-fire. Generally, the results found delayed mortality was most prevalent when in cold (22-41%) and wet (30%) forest types than in dry (1.7-19%) types. The numbers also reflected higher survivability in ponderosa pine and douglas fir than in broad leaf trees. I find this study relevant to my interests because I have been following the recovery of forests in Bonny Doon and Big Basin State Park, both in Santa Cruz County, California after the CZU complex fire came through in 2020. There are areas I frequent where some large madrone trees appeared to survive but now after 3 years clearly have died or look to be declining further. Most redwoods I see on the other hand are sprouting more and more fresh offshoots. It would be interesting to map offshooting trees and whether or not they survive another 3 years, as offshoots can also indicate a tree’s last attempt at survival. I hope to be able to use this methodology when I work for the CA State Parks in the Santa Cruz District in the near future, to gain insight and help in conservation efforts.

Reilly, Matthew J., et al. "Characterizing post-fire delayed tree mortality with remote sensing: sizing up the elephant in the room." Fire Ecology, vol. 19, no. 1, Dec. 2023, p. NA. Gale In Context: Environmental Studies, link.gale.com/apps/doc/A770485276/GRNR?u=s8405248&sid=bookmark-GRNR&xid=489ecfbe. Accessed 9 Dec. 2023.

In this study by Shane Coffield, we learn about efforts to quantify the amount of carbon being stored by forests in California by carbon-offset projects using remote sensing and GIS analysis. The models used for estimating carbon stored in a given area of land using remotely sensed data were eMapR, LEMMA, and project-reported carbon over time. The findings revealed several differences including the fact that the projects report considerably faster accumulation of carbon than estimates derived from eMapR or LEMMA. The time-series of remotely sensed imagery were matched with plot data using an algorithm based on spectral, climate, topographical, and disturbance history attributes to create maps of the data metrics on the annual level. Canopy height was also used to estimate biomass in the model. These methods using remote sensing to estimate carbon stocks have some shortcomings but are overall still reasonably accurate for the purpose of detecting changes on carbon offset project lands. The study found that timber companies overestimated their rate of carbon accumulation by putting aside less productive forests for the projects and were allowed to count wood products leaving on the assumption that wood products continue to sequester carbon for 100 years. Both eMapR and LEMMA may underestimate incremental growth in closed-canopy forest because they have trouble modelling high density forests such as redwood. In general this study suggests the carbon cap projects of California need refinement to sequester the most carbon and create the healthiest forests which can survive fire. I found this study interesting because I had thought remote sensing models were more lacking in their ability to estimate carbon in a forest, though I had come at it from the angle of board feet rather than carbon. I would like to learn more about the methodology of the eMapR and LEMMA models to understand how they make their estimates. Though the models focus on carbon stored in an area, one may be able to calculate board feet by subtracting branches and leaves from the equation using standard percentages for a given species.

Coffield, Shane R., et al. “Using Remote Sensing to Quantify the Additional Climate Benefits of California Forest Carbon Offset Projects.” Global Change Biology, vol. 28, no. 22, 2022, pp. 6789–806, https://doi.org/10.1111/gcb.16380.

The Lake Tahoe Basin is one of the premier tourist attractions of California and Nevada, containing a large alpine lake, forests, mountains and other landscape features. Over the past 150 years the area has seen great change as European Americans settled the area. The authors of “Change in the Forested and Developed Landscape of the Lake Tahoe Basin, California and Nevada, USA, 1940–2002” sought to assess land use changes using historical maps and imagery, often using GIS. They set out to make this assessment using a variety of factors derived through manual interpretation of images, image processing, and GIS data integration using data from the following years: 1940, 1969, 1987, and 2002. The results show that while deforestation and development were high in the mid 20th century, then began to slow down and forest density increased. Following a few severe fires, forest density decreased through the present day as a result of intentional thinning. As the local population began to realize the negative impacts of development on the natural systems of the Lake Tahoe Basin, they passed reforms to make development harder. The result has been that there is more public ownership than in the early 20th century, and development has been drastically curbed since the 1960s in particular. I find this historical analysis very pertinent because I was offered a forestry worker job in neighboring Truckee last summer. The job would have been thinning to get the forest back to its natural density before human fire suppression caused the forests to become overly dense, and thus subject to more damage as a result of any fire that came through because of dense fuel which can kill even larger trees. I would love to be part of planning for the future in this area, helping natural systems recover and mitigating the damage of wildfire. This historical study gives good insight into the effects of humans developing an area, and then suppressing fire.

Raumann, Christian G., and Mary E. Cablk. “Change in the Forested and Developed Landscape of the Lake Tahoe Basin, California and Nevada, USA, 1940–2002.” Forest Ecology and Management, vol. 255, no. 8, 2008, pp. 3424–39, https://doi.org/10.1016/j.foreco.2008.02.028.

In timberlands in Oregon, as well as other surrounding areas, degradation of marketable trees has taken place as a result of black bears. Black bears emerge from hibernation and peel bark off trees, which affects their marketability in the timber industry. Canopies of bear-killed trees are red, which can be easily detected with remote sensing. The red canopies are factored in as a percentage of total canopy cover, but without better estimates of the size of the trunks, it is very hard to estimate the loss of board feet. Other factors lead to dying trees, such as pests and drought, while some trees have degraded trunks without actually dying and having their canopies turn red, making estimation even more difficult. Under the simulations, bear peeling resulted in less than 1 percentage of loss of standing conifers per year. Bear damage to private timberlands caused annual loss of revenue by $0.9M-$88.5M annually, and bear peeling caused job losses between 11 and 1,012 in western Oregon each year. This massive range shows there is a need to better estimate the sizes of trees or produce a better way of collecting the data. I found this study discouraging given its inconclusive findings. I am extremely interested in creating a model for estimating board feet in a tree using remote sensing, but a simple analysis of “these canopies are red” will not suffice. It would be interesting to note if any bands or methods used in remote sensing can more accurately detect tree diameter and height to estimate board feet in a tree or stand of trees. I imagine the data would have to be coupled with data taken from lower angles, such as drones, to make those estimations.

Taylor, Jimmy, et al. Estimating the Total Economic Impact of Black Bear Peeling in Western Oregon Using GIS and REMI. 2014.

This study from 1996 researched the effects of air pollution and climate change on the health of Ukrainian forests over the course of 1989 to 1995 using GIS. Inputs included remote sensing data as well as on the ground observations to determine defoliation in stands of coniferous and deciduous trees. The findings were however rather inconclusive because although there was actually less air pollution in Ukraine in the stands studied, the stands still experienced defoliation. At the same time any effects of climate change were negligible. The writers attribute this to other human-caused factors such as logging and the introduction of pests. I found this to be another great reason to distinguish the difference between causation and correlation. Not every study over a short time period can use the tools at hand to discover meaningful findings without using more data on factors such as introduction of pests, or other human induced problems which could effect outcomes.

International Symposium on Air Pollution and Climate Change Effects on Forest Ecosystems (1996 : Riverside, Calif.), and Pacific Southwest Research Station. Proceedings of the International Symposium On Air Pollution And Climate Change Effects On Forest Ecosystems, February 5-9, 1996, Riverside, California. Berkeley, CA: U.S. Dept. of Agriculture, Forest Service, Pacific Southwest Research Station, 1998.

In “Use of GIS in Optimizing Timber Thinning Strategies in the Eastern Sierra Nevada,’ the researches aim to use GIS to choose the optimal stands of trees for thinning. Their motivation seems mostly fire prevention based, but there is a financial incentive and a financial limitation as well. Inputs regarding financial limitations include tree type, size, and hill slope or steepness where a stand is located, access to the stands via roads, etc. Natural limitations include proximity to streams and rivers. There is also an older human induced factor, which was the logging of larger pines like ponderosa during the second half of the 19th century, but the loggers left less valuable and damaged trees behind, which effects which stands are valuable enough to bring equipment for. All of these factors used to be very hard to map, but thanks to GIS mapping, were easier to map, analyze and thus make decisions. The authors note that as a result of their models and findings, thinning efforts are being made. I find this very relevant to my life as I am involved in an upcoming exemption timber harvest (for defensible space) where it will be my job to plot trees, their diameter at breast height, total height, etc. The goal is to do a small logging operation that leaves the stands of trees with a larger diameter than they currently are, while reducing density to an old growth forest while decreasing fire risk.

Knapp, Paul A., et al. “Use of GIS in Optimizing Timber Thinning Strategies in the Eastern Sierra Nevada.” The Professional Geographer, vol. 45, no. 3, 1993, pp. 323–31, https://doi.org/10.1111/j.0033-0124.1993.00323.x.

This study looks at the Niepolomice forest in Poland, which is near Krakow and an iron work factory which has emitted a significant amount of so2 or sulfur dioxide since the 1950s. In a rare instance of a positive effect of anthropogenic industrial pollution, this study found that certain metal pollution actually fertilized the forest. The writers analyzed the prevailing wind and soil samples to determine where the pollution was gathering on the ground, then conducted analysis on the forest type and size. They found that trees in areas with high so2 were actually larger and healthier than areas with less pollution, but that this coincided heavily with heavy metal dust containing zinc, iron, and others. This is an earlier application of GIS by today’s standards, so probably not including as many factors as a modern GIS dataset. It is nevertheless enlightening from a forestry standpoint, and would be interesting to see alongside a study of the effects of the same pollutants on water quality and fish, for example.

This prescient article on the uses of remote sensing in forestry from 1998 outlines the problems and future uses of remote sensing. The author, Henry M Lachowski makes plain problems we know from the time period, such as the expense of remotely sensed data from the time, as well as the database availability to help organizations in need of the data. He goes on to say he believes it will become more readily available and cheaper. He lays out the current uses at the time, such as plotting vegetation types and change of an area over time. As we know now, his article predicted the expansion of national databases, less expensive data, more types of sensors to collect data beyond what the eye can sense, and other developments for the use of remote sensing within GIS as it applies to forestry. Remote sensing is now used to gather all kinds of geographical data helping with environmental restoration and mapping to predicting fire spread and disaster mitigation. I believe we should continue to look to the future, as we now know more 3d modeling can be completed using drones, cell phones, and other vessels for remote sensing devices to sense from angles traditional satellites and planes never could. The technology for mapping and analyzing will continue to improve alongside improvements in sensors. I think the biggest issues will be in privacy, as mapping every inch, including in the 3rd dimension, becomes easier.

Remote Sensing Applied to Resource Management1 California. 2 Program Leader, Integration of Remote Sensing, Engineering Staff, USDA Forest Service, 2222 West 2300 South, Salt Lake City, UT 84119. Henry M. Lachowski

This is your Education description. Concisely describe your degree and any other highlights of your studies. Make sure to include relevant skills, accomplishments, and milestones gained. Don’t forget to adjust the timeframe in the subtitle.