Progress 10/01/19 to 09/30/20
Outputs
Target Audience:Forest land managers, natural resource professionals, private industrial forests, and other forestry stakeholders are the main target audience for this. These stakeholders may include people working for agencies (e.g., US Forest Service, US Fish and Wildlife Service, National Park Service, county land departments, state DNRs, non-governmental organizations (e.g., The Nature Conservancy, Ecoforestry Institute Society, American Forest Foundation), or private consultants (e.g., private consulting foresters) with a natural resource focus. A second audience is university students. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?Over the course of the reporting period, a graduate student and an undergraduate student were trained on how to combine remote sensing raster data with ground-based forest inventory data for modeling forest structure and forest defoliation from insect stressors. PhD candidate Thapa became proficient at modeling a time series of insect defoliation ground data with multi-temporal satellite sensor data to map this forest disturbance through time. She was also trained and has become and expert on how to use, run, and calibrate the landscape disturbance model Landis-II, for running management scnarios for understanding linkages between spruce budeworm outbreaks and host abundance on a landscape, Undergraduate Melanie Bogert was trained on combining low-density lidar and conifer biomass data for modeling coniferous, understory ladder fuels across a landscape using an iterative exclusion partial least squares regression approach. How have the results been disseminated to communities of interest?We have reached our target audiences through publications andconference presentations. For this reporting period I have four papers published and one paper in review, have authored or co-authored presentations with more planned for 2021, and results of this work are also available via USFS websites: http://www.nrs.fs.fed.us/disturbance/fire/extreme_fire_effects_mn/ What do you plan to do during the next reporting period to accomplish the goals?In the next reporting period, we will continue to focus on tasks across the two goal categories (spruce budworm [SBW] dynamics and wildfire fuels). For SBW, we will validate our satellite-based change detection approach using winter Landsat sensor data for another conifer defoliator, jack pine budworm (JPBW, Choristoneura pinus), in the region that has a different conifer host (Pinus Banksiana). Canopy disturbances caused this latter defoliator are far easier to detect remotely, since, unlike SBW, JPBW host trees occupy overstory positions. Once the change detection approach is validated, we will apply the methods to known SBW defoliation events. Results of this work and prior host structure work will be used together to calibrate and test our SBW landscape model, Landis-II, for exploring landscape configurations of host that mitigate the trajectory of SBW damage in this landscape. With respect the second category of research, modeling and mapping wildfire fuels, resulting canopy bulk density fuel layers will be combined with understory conifer ladder fuel biomass data layers to run fire behavior sensitivity tests within FlamMap6 to gauge the usefulness of these locally-derived fuel variables.
Impacts
What was accomplished under these goals?
IMPACT - The ability to model and map forest disturbance risk by fire and defoliating insects is critically important to sustain both valuable ecosystem services and to protect the economic health of our forest resources, especially in lieu of climate change. This project period, we mapped understory coniferous ladder fuels across the Superior National Forest (SNF). Thesedata are currently in use by SNF fire managers to develop and implement fuel reduction treatments. FIRE - We aim to improve forest fire modeling and risk assessment over and above that which is currently possible in the upper Midwest by deriving local estimates of a key forest parameter (canopy bulk density, CBD) universally needed to model forest fire behavior. National estimates of CBD used to model fire behavior in northern Minnesota produce inadequate results; requiring frequent empirical calibration changes to enable duplication of past fire events. Hence, we have developed a highly accurate method of mapping CBD using remote sensing-based predictor variables. To our knowledge, this marks the first successful attempt at mapping CBD reported to date in this region, which represents an enormous impact in the field of fire behavior modeling.CBD models were used with salient image predictors to seamlessly map CBD across the study area. SPRUCE BUDWORM (SBW) - The current trend of forest disturbance from SBW is outside the range of natural variability due to an expansion of host (Abies balsamea and Picea glauca) over the last 100 years resulting from past fire suppression policies and subsequent forest management methods. It is believed that modification of host density and spatial arrangement (Robert et al. 2018), via adaptive management, has the potential to return the SBW disturbance regime back to historically normal levels. By doing so, we hope to curb annual, economic losses to the forest products industry, both locally and regionally. Other beneficial outcomes include an eventual reduction of wildfire risk to people and property through a reduction of spruce SBW host (primarily balsam fir) connectivity across the landscape. Providing quantifiable evidence of links between forest structure and SBW outbreak mechanics has been a challenge due to a lack of host information. Hence, our focus was to develop both spatially explicit SBW host maps (past and present) and a repeatable methodology to detect defoliation disturbance using space-borne assets. Host and disturbance maps are needed to enable the design of pest resistant landscapes via rigorous future-cast scenario modeling. We have succeeded in developing a method to map SBW host using Landsat sensor data and USFS inventory data. Reliable remote detection of SBW disturbance, on the other hand, is proving to be a more recalcitrant problem. However, we have identified a method of detecting host disturbance and we are currently in the validation phase of this work. FIRE OBJ 1 -USE SATELLITE SENSOR DATA TO QUANTIFY AND MAP FOREST BIOPHYSICAL VARIABLES NEEDED TO MODEL FIRE BEHAVIOR. We successfully used ground data (direct tree dimension data and indirect canopy gap fraction data) with satellite data (Landsat-8, Palsar-1, Sentinel-1, & Sentinel-2) to calibrate models for mapping canopy bulk density (CBD). We modeled both CBD_CGF and CBD_FuelCalc with high accuracy (adj. R^2 = 0.97 and 0.96, RMSE = 0.07 and 0.16 kg*m^-3), respectively. Each of the resulting CBD models were used with salient image predictors to seamlessly map CBD across the study area. FIRE OBJ 2 - USE SPATIALLY EXPLICIT BIOPHYSICAL FOREST FUEL PARAMETERS TO IMPROVE FIRE RISK MANAGEMENT AND FIRE BEHAVIOR MODELING FOR DECISION SUPPORT IN THE UPPER MIDWEST. Canopy bulk density (CBD) results from objective one are currently being used with other remote sensing-derived estimates of forest structure in the US Forest Service's FARSITE modeling framework (via FlamMap6) to run hind-cast simulations for the 725 ha 2006 Redeye Lake fire. Duplicating the boundary of this 2006 fire using our estimates of CBD is a critical first step in determining the value of these space-based estimates of CBD for future fire risk assessment in this region. In a parallel effort, we used low-density Lidar data and vertically explicit understory fuels biomass data (2015-2016) to predict and map understory coniferous ladder fuels across the Superior National Forest (SNF). Our understory ladder fuels data are currently in use by SNF fire managers to develop and implement fuel reduction treatments. FIRE OBJ 3 - DEVELOP A SET OF PROTOCOLS AND RECOMMENDATIONS FOR MAPPING FOREST FUEL PARAMETERS IN THE UPPER MIDWEST. To capture fuel structures beyond optical detection limits, as suggested by some researchers (Keane et al. 2005, 2006), we used synthetic aperture radar (SAR) in combination with optical satellite sensor data, coupled with both FuelCalc-based estimates and indirect canopy gap fraction (CGF via LAI-2200C) based field estimates of CBD. The presence of understory balsam fir (e.g., ≤ 3m tall) attenuated the view of our LAI-2200C optical instrument when measuring CGF, which caused substantial bias in predicting CBD. We concluded that C-band (5.6 cm wavelength) SAR satellite sensor data are not capable of adequately penetrating the upper canopy layers to enable adequate characterization of the understory fuel layer, while these fuels are largely invisible at L-band wavelengths (24 cm). The maximum range of CGF-modeled CBD was highly sensitive to the zenith angle ranges of the LAI-2200C instrument. The importance of spatial variability among the various CGF-derived CBD estimates on actual fire behavior remains untested. BUDWORM OBJ 1-- DEVELOP RELIABLE SATELLITE-BASED METHODOLOGIES FOR DETECTING THE MIXED AND OFTEN CONFOUNDING SIGNATURE OF SPRUCE BUDWORM DAMAGE IN THE UPPER MIDWEST. We used the US Forest Service's Forest Inventory and Analyses (FIA) data with near concurrent Landsat sensor data to estimate and map spruce budworm (SBW) host species abundance from before and after an early 1990s SBW infestation. To our knowledge, mapping the spatial distribution of SBW disturbance has not been successful in the upper Midwest compared to work in the relatively pure spruce-fir forests of the Atlantic maritime regions. SBW defoliation detection in the upper Midwest, where forests are mixed and host is often in the understory, is a much more recalcitrant problem. We have found that use of winter Landsat sensor data affords an optimal view of SBW host is the SNF region. Hence, we developed a satellite-based change detection methodology using exclusively winter Landsat data to quantify SBW host defoliation, which we are currently validating using another conifer defoliator in the region (jack pine budworm, Choristoneura pinus). Success of our change detection method for detecting SBW host defoliation will mark the final data gap in our efforts to model SBW dynamic across the upper Midwest. Our research has successfully filled one critical SBW host species knowledge gap (Thapa et al. 2020) and is poised to fill the SBW disturbance mapping gap (Thapa et al. in prep). Resulting SBW host maps now serve as input calibration data to spin-up a landscape disturbance simulation model, LANDIS-II (Scheller et al. 2007). LANDIS-II uses weather data, SBW life cycle data, and our host species data to run simulations on various management practices aimed at reducing SBW outbreak severity, duration, and frequency into the future; up to a 1000 years. We are currently working with US Forest Service and Canadian collaborators to better parameterize the LANDIS II model to implement SBW host manipulation scenario trials. The SBW disturbance and host management scenario modeling results will be used to inform federal, state, county, and private forest managers on the best management practices for mitigating losses to SBW activity.
Publications
- Type:
Journal Articles
Status:
Under Review
Year Published:
2020
Citation:
Wolter, P.T., Olbrich, J., & Johnson, P. (2020). Modeling sub-boreal forest canopy bulk density in Minnesota, USA using
synthetic aperture radar and optical satellite sensor data. Fire Ecology (in review).
- Type:
Journal Articles
Status:
Published
Year Published:
2020
Citation:
Robert, L.-E., Sturtevant, B. R., Kneeshaw, D., James, P. M. A., Fortin, M-J., Wolter, P.T., Townsend, P. A., and Cooke, B. J. (2020). Forest landscape structure influences the cyclic-eruptive spatial dynamics of forest tent caterpillar outbreaks. Ecosphere, 11(8), e03096.
- Type:
Journal Articles
Status:
Published
Year Published:
2020
Citation:
Thapa, B., Wolter, P.T., Sturtevant, B.R., and Townsend, P.A. (2020). Reconstructing past forest composition and abundance by using archived Landsat and national forest inventory data. International Journal of Remote Sensing, 41(10), 4022-4056.
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Engelstad, P. S., Falkowski, M., Wolter, P., Poznanovic, A., & Johnson, P. (2019). Estimating Canopy Fuel Attributes from Low-Density LiDAR. Fire, 2(3), 38.
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Zlonis, E.J., Walton, N.G., Sturtevant, B.R., Wolter, P.T., & Niemi, G.J. (2019). Burn severity and heterogeneity mediate avian response to wildfire in a hemiboreal forest. Forest Ecology and Management, 439, 70-80.
|
Progress 10/01/18 to 09/30/19
Outputs
Target Audience:My research efforts center on two distinct goals: (1) improved fire behavior modeling and (2) enable and perform spruce budworm host management scenario modeling; both within the upper Midwest. Regarding the first project goal, burgeoning fire risk in northern coniferous forests, where private properties are tightly intertwined with publically managed forestland, demands that we enhance our preparedness via fire scenario modeling and targeted fuels reduction in critical areas of the landscape reduce fire risk and fire severity. Hence, the target audiences of my research are those engaged in fire risk management: U.S. Forest Service, state- and county-level fire managers, and interested private citizens living in these landscapes. These audiences are reached via a combination of media that includes peer-reviewed research findings, oral presentations at national and international meetings, and online information including webinars. Regarding the second project goal, northern coniferous landscapes have been altered over the past century (via fire suppression and clear-cut harvesting) so that they now supports an abundance of spruce budworm (SBW) host trees (fir and spruce) that is far outside the range of natural variability. Understanding the relationship between elevated SBW outbreak periodicity and forest management is key to dampening both the environmental and economic impact of such outbreaks. Hence, the audience for this research is international in scope. Both U.S. and Canadian researchers, forest managers, and private landowners seek strategies to reshape these landscapes to natural and pre-fire suppression conditions that will minimize the impact of this endemic insect. Hence, outlets for this research mirror those of the first project goal listed above. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?This project is training a PhD student and a budding undergraduate forestry student in remote sensing techniques for estimating and mapping forest biophysical variables with satellite sensor data. The PhD student is also receiving advanced training in the spatially explicit landscape forecast modeling framework known as LANDIS-II, which was developed by Dr. David Mladenoff and subsequent collaborations and updates by Dr. Rob Scheller from the early 1990s to present. My PhD student continues to be the go to person by the USFS for modeling the behavior of spruce budworm in the upper Midwest and also the maritime regions of eastern Canada. This project is training a PhD student, a MS student, and a budding undergraduate forestry student in remote sensing techniques for estimating and mapping forest biophysical variables with satellite sensor data. The PhD student is also receiving advanced training in the spatially explicit landscape forecast modeling framework known as LANDIS-II, which was developed by Dr. David Mladenoff and subsequent collaborations and updates by Dr. Rob Scheller from the early 1990s to present. My PhD student continues to be the go to person by the USFS for modeling the behavior of spruce budworm in the upper Midwest and also the maritime regions of eastern Canada. The masters student received similar training, but with a focus on fire behavior modeling. This student received advanced training (via Iowa State University and the US Forest Service) in the FARSITE modeling framework employed by the US Forest service. This MS student has becoming an expert spatial analyst using both ArcGIS and the Python programming language. How have the results been disseminated to communities of interest? We have reached our target audiences through publications, presentations, and formal classroom instruction. In total, six papers were either published or in review in 2019, seven presentations were provided, and the results of this work has been disseminated through websites and webcasts. I approximate that Pagami Fire presentations alone have collectively reached over 2000 individuals. Modes of information dissemination include a maintained US Forest Service web site (http://www.nrs.fs.fed.us/disturbance/fire/extreme_fire_effects_mn/), presentation of results at national conferences, invited seminars at major universities, and webinars. What do you plan to do during the next reporting period to accomplish the goals?In the next reporting period we will focus on two primary tasks across the two overarching goals of the study. First, we will continue to perform change detection tests to detect SBW defoliation using the aforementioned methods above using Landsat or Sentinel-2A satellite sensor data. Second, we will validate our SBW disturbance detection methods on damage caused by jack pine budworm (JPBW) in northern Wisconsin. The idea is that the JPBW mechanism of disturbance is much more easy to detect from satellite sensors (primarily overstory) than that of SBW disturbance (primarily understory), so we hypothesize that mapping JPBW disturbance patterns should be a relatively simple task. Tests of our multi-sensor change detection methodologies will be performed within an area of northwest Wisconsin that was partially defoliated between 2004-2006, and for which collaborators had collected concurrent ground data (B. Sturtevant, P. Townsend). Results of these analyses will serve as a validation of our basic methodologies, but also to showcase how difficult mapping SBW disturbance events from space in the upper Midwest is compared to Canadian's Atlantic maritime provinces where host typically dominates the overstory. With regard to forest fire fuels mapping and fire behavior modeling, we will investigate the combined used of optical, SAR, and Lidar data for mapping coniferous understory ladder fuels in the upper Midwest to inform adaptive fuels management by the US Forest Service.
Impacts
What was accomplished under these goals?
My research efforts center on two distinct goals: (1) fire behavior modeling plus forest fuels mapping and (2) understanding spruce budworm (SBW) host management scenario modeling; both within the upper Midwest. FIRE - Regarding the first project goal, burgeoning fire risk in northern coniferous forests, where private properties are tightly intertwined with publically managed forest land, demands that we enhance our preparedness via fire scenario modeling and targeted fuels reduction in critical areas of the landscape reduce fire risk and fire severity. To do so, the U.S. Forest service uses an integrated decision tool known at the Wildland Fire Decision Support System (WFDSS), which combines data from a national vegetation and fuels data program (LANDFIRE) with a modeling framework (FARSITE) to develop management recommendation for adaptive forest management to reduce risk. However, while this system works reasonably well for other coniferous regions across the United States, it performs poorly --at best-- in the upper Midwest; where trained USFS professionals are unable to recreate the behavior of recent fire events given the latest input parameters supplied via the LANDFIRE effort. Because LANDFIRE fuels data are largely generated via analyses of optical Landsat satellite sensor data, we hypothesized that optical detection of understory coniferous canopy fuels is partially confounded by overstory canopy elements. Hence, inclusion of active satellite sensor data such as synthetic aperture radar (SAR, C- and L-band) would afford a more complete view of understory structures. Results of our 2019 investigations show that inclusion of both optical and SAR satellite sensor data drastically improved mapping of canopy bulk density form moderate (Engelstad et al. 2019) to excellent (Wolter et al. in prep). Moreover, modeling of fire behavior also showed signs of improvement compared to LANDFIRE-calibrated fire behavior models (Oblrich et al. in prep). The impact of this research effort includes recommendations for reshaping LANDFIRE procedures and protocols for processing a wider suite of satellite image data in the upper Midwest. Objectives met: 1) Use satellite sensor data to quantify and map forest biophysical variables needed to model fire behavior 2) Use these spatially explicit forest biophysical parameters to improving fire risk management and fire behavior modeling for decision support in the upper Midwest. 3) Develop a set a set protocols and recommendations for mapping forest fuel parameters in the upper Midwest. SPRUCE BUDWORM - regarding our second project goal, historical forest landscape disturbance patterns by spruce budworm (SBW), needed for hind-cast calibration, are currently lacking at sufficient scales to be useful in understanding mechanics between SBW host (spruce and fir trees) abundance and disturbance. The Wolter lab endeavored to fill these historical information gaps using archived satellite sensor data (optical and radar). While we have not yet been able to recreate past spruce budworm disturbance events, we have identified critical limitations of national forest inventory data for ground-to-satellite calibration of forest composition models (Thapa et al. accepted 2019, published 2020). Objectives that remain: 1) Develop reliable satellite-based methodologies for detecting the mixed and often confounding signature of spruce budworm damage in upper Midwest forests.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Olbrich, J. & Wolter, P. (2019). Understanding fire in a complex landscape. Oral presentation at the 6th Fire Behavior and Fuels Conference, 29 April 3 May, Albuquerque, NM.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Sturtevant, B.R., Cooke, B.J., Robert, L.-E., Miranda, B.R., Thapa, B., and Wolter, P.T. (2019). Modelling insect-forest interactions across landscapes: Heart of the Continent advances the state of the art. Heart of the Continent Partnership, Science Symposium, Duluth, MN, April 8-9.
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Engelstad, P. S., Falkowski, M., Wolter, P., Poznanovic, A., & Johnson, P. (2019). Estimating Canopy Fuel Attributes from Low-Density LiDAR. Fire, 2(3), 38.
- Type:
Journal Articles
Status:
Published
Year Published:
2019
Citation:
Zlonis, E.J., Walton, N.G., Sturtevant, B.R., Wolter, P.T., & Niemi, G.J. (2019). Burn severity and heterogeneity mediate avian response to wildfire in a hemiboreal forest. Forest Ecology and Management, 439, 70-80.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Olbrich, J. & Wolter, P. (2019). Sensitivity analysis of canopy fuel loads on crown fire behavior in northeast Minnesota. Oral presentation at the Forestry and Wildlife Research Review, 10 January, Cloquet, MN.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Thapa, B., Wolter, P.T., Sturtevant, B.R., and Townsend, P.A. (2019). Reconstructing Historical Forest Composition and Abundance by Using Archived Landsat and National Forest Inventory Data (oral presentation). The 2nd Heart of the Continent Science Symposium. 8-9 April, Duluth, MN.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Thapa, B., Wolter, P., & Sturtevant, B. (2019). Oral presentation: Detecting spruce budworm induced change through remote sensing. Annual Ecological Society of America symposium, 11-16 August 2019, Louisville, KY.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Thapa, B., Wolter, P., & Sturtevant, B. (2019). Oral presentation: Detecting spruce budworm induced change through remote sensing. American Association of Geographers symposium, 3-7 April, Washington, DC.
|
Progress 05/01/18 to 09/30/18
Outputs
Target Audience:Forest land managers, natural resource professionals, private industrial forests, and other forestry stakeholders are the main target audience. These stakeholders may include people working for agencies (e.g., US Forest Service, US Fish and Wildlife Service, National Park Service, county land departments, state DNRs, non-governmental organizations (e.g., The Nature Conservancy, Ecoforestry Institute Society, American Forest Foundation), or private consultants (e.g., private consulting foresters) with a natural resource focus. A second audience is university students. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?This project is training a PhD student in remote sensing techniques for estimating and mapping forest biophysical variables with satellite sensor data. This student is also receiving advanced training in the spatially explicit landscape forecast modeling framework known as LANDIS-II, which was developed by Dr. David Mladenoff and subsequent collaborations and updates by Dr. Rob Scheller from the early 1990s to present. My PhD student is becoming the go to person by the USFS for modeling the behavior of spruce budworm in the upper Midwest and also the maritime regions of eastern Canada. A masters student is also receiving similar training, but with a focus on fire behavior modeling. This student is receiving advanced training (via Iowa State University and the US Forest Service) in the FARSITE modeling framework employed by the US Forest service. This student is also becoming an expert spatial analyst using both ArcGIS and the Python programming language. How have the results been disseminated to communities of interest?We have reached our target audiences through publications, conference presentations, and formal classroom instruction. For this reporting period I have one paper published and two papers in preparation, have authored or co-authored three presentations with four more planned for 2019, and results of this work are also disseminated through websites (http://www.nrs.fs.fed.us/disturbance/fire/extreme_fire_effects_mn/), invited seminars at major universities (e.g., South Dakota Society of American Foresters), and national conferences (Soil Science Society of America, Ecological Society of America). What do you plan to do during the next reporting period to accomplish the goals? In the next reporting period we will focus on two primary tasks across the two overarching goals of the study. First, we will perform change detection tests to detect SBW defoliation using the aforementioned methods above using Landsat or Sentinel-2A satellite sensor data. Tests of change detection methodology will be performed within an area of Minnesota defoliated last year; and for which we collected concurrent ground data. Successful methods for detecting this recent SBW defoliation event will then be applied to detect the 1991 SBW defoliation event. Concurrently, we will continue to refine and calibrate the LANDIS-II model for use in Minnesota. Result of these efforts will be presented at two national scientific conferences: Thapa, B., Wolter, P., & Sturtevant, B. (2019). Oral presentation: Detecting spruce budworm induced change through remote sensing. Annual Ecological Society of America symposium, 11-16 August 2019, Louisville, KY. Thapa, B., Wolter, P., & Sturtevant, B. (2019). Oral presentation: Detecting spruce budworm induced change through remote sensing. American Association of Geographers symposium, 3-7 April, Washington, DC. Second, we will continue to test FARSITE model calibrations on past forest fire events using our canopy bulk density (CDB) information derived during this reporting period. Result of these efforts and our forest CBD modeling work will be presented at no less than two scientific conferences: Olbrich, J. & Wolter, P. (2019). Sensitivity analysis of canopy fuel loads on crown fire behavior in northeast Minnesota. Oral presentation at the Forestry and Wildlife Research Review, 10 January, Cloquet, MN. Olbrich, J. & Wolter, P. (2019). Understanding fire in a complex landscape. Oral presentation at the 6th Fire Behavior and Fuels Conference, 29 April - 3 May, Albuquerque, NM.
Impacts
What was accomplished under these goals?
IMPACT: The ability to map and model forest disturbance and future disturbance risks by fire and defoliating insects is critically important to sustain both valuable ecosystem services and to protect the economic health of our forest resources, especially in lieu of unprecedented climate change. Hence, a targeted suite of forest management practices are sorely needed to adapt to and mitigate burgeoning forest disturbance threats. This research strives to address these forest management deficiencies, while also expanding our knowledge of forest ecosystem dynamics. First, we strive to improve forest fire modeling and risk assessment/prediction over and above that which is currently possible in the upper Midwest. We explore satellite-based estimation of a key forest parameter (canopy bulk density, CBD) universally needed to model forest fire behavior. While the current methods of estimating CBD in other regions of the country are considered sufficient for assessment and modeling of fire risk, such efforts breaks down in northern Minnesota; requiring frequent empirical calibration changes to enable duplication of past fire events. In an effort to improve this process, we have developed a highly accurate method of mapping CBD using remote sensing-based predictor variables. To our knowledge, this marks the first successful attempt at mapping CBD reported to date in this region, which represents an enormous impact in the field of fire behavior modeling. Second, with regard to defoliating insects, we strive to better understand the changing dynamics of the spruce budworm (SBW, Choristoneura fumiferana) in the Upper Midwest. The impacts of the SBW are considered outside the range of natural variability due to the build-up of SBW host over the last 100 years. This build-up of host is the result of past fire suppression policies and subsequent forest management methods that evolved in parallel. It is believed that modification of the host density and spatial arrangement (Robert et al. 2018), via adaptive management, has the potential to return the SBW disturbance regime to that of more normal level. To enable this, we SBW host maps are needed and a repeatable methodology to detect SBD disturbance using space-borne assets. Both host and disturbance maps are needed to design a future pest resistant landscape. So far, we have developed a method to map SBW host using extant Landsat sensor data and US Forest Service national inventory. We have found that FIA data must be used with care if accuracy is to be maximized. We suggest that FIA ground collection protocols be modified going forward so that they integrate better with 30-m Landsat sensor data, as substantial federal commitments exist to continue the collection of both sources of data (FIA and Landsat). FIRE OBJ 1 Use satellite sensor data to quantify and map forest biophysical variables needed to model fire behavior: Ground plot data (n=62) from our northern Minnesota study area, consisting of tree species counts, corresponding tree dimension measurements, and canopy gap fraction (CGF) measurements (via LI-COR LAI2000), were used with remote sensing data (i.e., Landsat-8, Palsar-1, Sentinel-1A, and Sentinel-2A) to calibrate models for estimating and mapping canopy bulk density (CBD) across the study. Two methods were used to convert ground data (tree data and CGF data) into separate ground estimates of CBD. First, CBD at our ground plot locations was determined using tree dimension data and the US Forest Service program FuelCalc (Reinhardt et al. 2006). Second, plot-wise CBD was also estimated using our CGF measurements and a transformation equation developed by Keane et al. (2005). Each set of ground estimates for CBD (i.e., CBD_FC and CBD_GF) served as the dependent variable for two separate calibrations with remote sensing data. We modeled both CBD_GF and CBD_FC with high accuracy (adj. R^2 = 0.97 and 0.96, RMSE = 0.07 and 0.16 kg*m^-3), respectively. Each of the resulting CBD models were used with salient image predictors to seamlessly map CBD across the study area. FIRE OBJ 2. Use these spatially explicit forest biophysical parameters to improving fire risk management and fire behavior modeling for decision support in the upper Midwest: The CBD results from objective one are currently being used with other remote sensing-derived estimates of forest structure (e.g., forest basal area and crown closure) in the US Forest Service's FARSITE modeling framework to run hindcast simulations a past forest fire from 2006: the Redeye Lake fire. This 2006 Redeye Lake fire was 725 ha in size and occurred ca. 25 km west of our study area. Duplicating the boundary of this past fire using our estimates of CBD is a critical first step in determining the value of these space-based estimates of CBD for future fire risk assessment in this region of Minnesota. FIRE OBJ 3. Develop a set a set protocols and recommendations for mapping forest fuel parameters in the upper Midwest: Nothing to report. BUDWORM OBJ 1: Develop reliable satellite-based methodologies for detecting the mixed and often confounding signature of spruce budworm damage in upper Midwest forests: We used the US Forest Service's Forest Inventory and Analyses (FIA) data with near concurrent Landsat sensor data to estimate and map SBW host species abundance from before and after an early 1990s SBW infestation. To our knowledge, mapping the spatial distribution of SBW disturbance has not been successful in the upper Midwest compared to the maritime regions in Canada for two primary reasons. First, the preferred host species (Abies balsamea) exists primarily in the forest understory hidden from optical satellite sensors by numerous non-host hardwood and conifer species. Second, SBW host density in Minnesota is substantially lower than in the less species rich eastern boreal forests. Hence, detection of Studies in other regions have used aerial sketch maps of disturbance (Candau et al. 1998), remote sensing with canopy gap fraction field data (Leckie et al. 1988, 1989), and more conventional satellite sensor (i.e., Landsat and SPOT) data (Franklin et al. 2008). The latter two studies have demonstrated the potential of remote sensing to detect the cumulative defoliation, however, they haven't explicitly implemented such data to produce disturbance maps. Furthermore, if mortality is not severe during the first five years of an outbreak, as is typically the case in Minnesota (Robert et al. 2018), the impacts of SBW defoliation may remain relatively invisible with respect to Landsat or other optical satellite sensors due to spectral confusion with burgeoning coniferous understory regeneration through time. Hence, we are in the process of testing different methods of SBW disturbance detection: (1) simple 1985 to 2005 change detection using vegetation indices known to be sensitive to forest disturbance (Healey et al. 2005, Jin & Sader 2005); (2) a temporal autocorrelation function (ACF) used on time-series Landsat images between the two dates to detect subtle SBW-related spectral changes in forest canopy (Shumway & Stoffer 2011); and (3) dynamic time wrapping (DTW), which calculates the Euclidian distance between two-time series while minimizing the cumulative distance (Keogh et al. 2005). This latter method identifies the time series that are similar in their spectra and then classifies them into k-groups.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2018
Citation:
Robert, E., Sturtevant, B., Cooke, B., James, P., Fortin, M-J., Townsend, P., Wolter, P., and Kneeshaw, D. (2018). Landscape host abundance and configuration regulate periodic outbreak behavior in spruce budworm Choristoneura fumiferana. Ecography, 41, 1556⿿1571.
|
|