Forest disease mapping and modeling using GIS
Monitoring the spread of Sudden Oak Death
Table of Contents:Introduction to Sudden Oak Death The oomycete Phytophthora ramorum
is an exotic pathogen threatening coastal forests of California and
Oregon. Known to infect over 60 plant genera (Jones 2006), P.ramorum causes non-lethal leaf blights or dieback in most hosts. On others, particularly in tanoak, Lithocarpus densiflorus, and red oaks, Quercus
section Lobatae, the pathogen causes bleeding stem cankers and eventual
mortality (Rizzo and Garbelotto 2003). This later disease has been
titled “Sudden Oak Death” (SOD). Management of aggressive, exotic
pathogens is inherently a spatial problem, especially once the pathogen
has spread beyond its initial introduction area. Any management
decisions must be made with detailed information about where the
pathogen currently is, the patterns of spread, and which areas are most
at risk. Answering these questions can be accomplished through the
remote sensing of infested and uninfested forests.
Citations: Articles: public monitoring and the use of WebGIS Kearns, F.R., Kelly, M., Tuxen, K.A. 2003. Everything happens somewhere: using webGIS as a tool for sustainable natural resource management. Frontiers in Ecology and Evolution. 1(10):541-548 Landscape scale monitoring programs can be accomplished with public participation (e.g. see the Audubon Christmas bird survey). GIS is a useful tool for visualizing and analyzing spatial data, esp. as it relates to public policy management; however its use in developing public policy has limited power for it can require specialized software and hardware or professional training. These drawbacks may limit the quality of data gathered by the public or its accessibility when, under ideal situations, it may be used to gain support for management options. In response to this, webGIS (public participation GIS; PPGIS) has been developed. WebGIS uses common web language (e.g. http and Java), eliminating the need to own and be able to use GIS software. It is aimed at providing access to the data present on website. Users may submit personal inquires and, in some programs, input potential data of interest. Although it is a new technology, webGIS has been used to map the spread of P.ramorum within California. Developed in 1999, ‘OakMapper’ (see websites) provides information on the currently confirmed and potentially positive infections as collected from data gathered by the public and academics. This program is most effective as an outreach tool in two ways: 1) an interactive query system that will allow the public to generate their own customized maps of areas of interest and 2) show were there has been effective public outreach. Kelly, N.M., Tuxen, K. 2003. WebGIS for monitoring “Sudden Oak Death” in coastal California. Computers, Environment, and Urban Systems. 27:527-547 WebGIS integrates classic community-based environmental monitoring with modern GIS technology. An effective real-time, internet-accessed monitoring program incorporates data input, analysis of monitoring data, presentation of the results, and a database that may be queried. Individual tree infection is gathered from multiple sources, including the public. Regionally mortality is gathered through aerial surveys, and larger scale coverage are being monitored through more extreme remotely sensed data (e.g. Airborne Data Acquisition and Registration with 1 m spatial resolution and Thematic Mapper at 30m spatial resolution). Software used to present the data for OakMapper is ArcIMP version 3.1, a webGIS software developed by ESRI. Data are stored as vector shapefile (all point and polygon data; i.e. individual trees and patches, respectively) or raster GRID format (regional). Public reports made with a specific address are geocoded into a spatial reference (projected Teale-Albers Equal Area) and automatically submitted on the website as symptomatic. Lab results will confirm and change the status of the submission. There have been approx. 5 submissions / week; the most common trees reported are coast live oak and tanoak. Users can search and query by county, zip code, district, symptoms or species, and can zoom into areas of interest. Hyperlinks to images of the tree are provided. This paper is the most comprehensive treatment of the OakMapper program. Kelly, M., Tuxen, K., Kearns, F. 2004. Geospatial informatics for management of a new forest disease: Sudden Oak Death. Photogrammetric Engineering and Remote Sensing. 70(9):1001-1004 Spatial aspects of SOD that became of interest to spatial analysts soon after the disease was recognized include the patterns of disease impact (esp. the scale of impact at advanced stages), the mechanism of disease spread (and resulting patterns of patch mortality at the landscape scale), spatial distribution of the disease (i.e. where is the disease the severest, or where might it be in the future?), and the use of webGIS as a spatially referenced community outreach program used to aid management. This paper is a formal presentation of OakMapper as a community-input, web-based, GIS tool, as well as the information found at the California Oak Mortality Task Force (COMTF) website (see websites). Static maps by county and state are freely available on the web (JPEG, PDF, and TIFF formats) through COMTF, or customized maps may be made with OakMapper. OakMapper also allows community members (even those without GIS experience) to report symptomatic trees with the following information: host name and symptoms (pictures provided in form), a rating of the reporter’s familiarity with forest and forest diseases, and the tree location (GPS coordinates, address, or location on a map). This spatial information is often used in conjunction with other remote sensing data, including the mapping the distribution of host trees, mapping the changes in water stress of host trees, and the predictive modeling of disease spread and / or severity over time. Articles: remote sensing and modeling of SODS and other hardwood forest diseases Condeso, T.E., Meentemeyer, R.K. 2007. Effects of landscape heterogeneity on the emerging forest disease sudden oak death. Journal of Ecology 95:364-375 The movement and establishment of an invasive pathogen are inherently a spatial problem greatly influenced by landscape heterogeneity. The spatial distribution of pathogen habitat (the conglomerate of terminal host, reservoir host, and non host species) can mediate disease spread, especially as they affect microclimate suitability. This paper answers two questions: how does various hosts habitat (composition and configuration) influence disease severity and on what scale; and how does this composition create microclimate conditions suitable for disease development? Pixels of remotely sensed images (ADAR) were classified (supervised, based on field data) into host-woodland and nonhost-woodlands classes. Landscape patterns were defined by woodland area, perimeter : area ratio, and patch cohesion. Plot-level variables include abundance of bay, canopy cover, distance to forest edge, and elevation (based on USGS 10m DEM); calculated microclimate variables included average daily number of hours in optimal temperature range (10-25degrees C) for P.ramorum reproduction and that for temperatures greater than 25degrees C, average daily number of hours at high relative humidity (>95%), and number of hours in the optimal temperature range, and those lower than 10degrees C during and up to 24 hours after a rain event of at least two consecutive days. P.ramorum is severest in plots surrounded by a high percentage of woodland habitat, though the landscapes effect on disease severity is scale-dependant (greatest for patches with radius 200m). The authors further hypothesize that the disease severity at this radius represents a maximum dispersal distance of P.ramorum (though they admit that the proportion of woodland on their plots did not change at scales greater than 200m, so further studies are needed to explore long-range dispersal possibilities). Contiguous forest likely contains greater number of bays, providing an important inoculum source and/or an increased probability of successful infection by dispersing propagules. Still, the disease could still be found in isolated forest patches. No variation in microclimate measurements were attributed to distance from forest edge, rather microclimate conditions are more influenced by topography than by landscape patterns. Disease severity in these regions are possibly attributed to greater wind velocities and thus greater spread of inoculum. Everitt, J.K., Escobar, D.E., Appel, D.N., Riggs, W.G., Davis, M.R. 1999. Using air-borne digital imagery for detecting oak wilt disease. Plant Disease 83:502-505 The disease oak wilt (causal agent Ceratocystis fagacearum) causes the loss of function of water-conducting tissues in oak species, resulting in dieback of upper crown and browning of foliage (much like what is observed by P.ramorum). Color-infrared (CIR) aerial photography can be used to distinguish between live, infected and dead trees. Reflectance measurements to do so were made in the visible green, visible red, and near-infrared spectral bands. Foliage from healthy branches had lower visible green and red reflectance and higher near-infrared reflectance than that from diseased and dead trees (the inverse is true for dead trees compared to diseased and healthy ones; diseased trees reflectance lies in-between the two classes). Near-infrared can be correlated with leaf density, allowing these aerial photographs to be used to determine tree health and thus monitor disease severity on the landscape scale. Guo, Q., Kelly, M, Graham, C.H. 2005. Support vector machines for predicting distribution of sudden oak death in California. Ecological Modeling 182:75-90. Most methods used to map the potential and predicted spread of diseases require presence/absence data or an understanding of ideal environmental conditions (as for the case when pathogens are restricted by temperature, moisture, or other environmental factors), neither of which may be available in the case of a resent introduction of an invasive. Alternative methods, which use only presence data include the BIOCLIM, DOMAIN and ENFA models; a new alternative is support vector machines (SVMs) which incorporate machine learning algorithms used to distinguish between classes within sets of remotely sensed data. This paper contrasts the use of a one-class SVM model (with presence only data) to that of a two-class SVM model (both presence and pseudo-absence data). 14 variables were used to predict the potential distribution of P.ramorum: annual mean temp., annual mean precip., mean temp. and precip. for Jan., April, July and Oct., annual mean solar radiation, distance to main roads, distance to edges of patches of hosts, and elevation (climate extracted from DAYMET; see websites). Seasonality of climate variables is considered key in limiting range of the pathogen. The authors champion SVM as a new technique citing the ease of use (less fine tuning needed), the ability to handle non-normal feature distribution data, and the lack of density restriction. The true-positive rate of the one-class model is greater than that of the two-class model; however the one-class SVM predicted a greater risk area. In comparison to two-class SVMs, one-class SVMs are computationally more efficient (as they don’t need to generate pseudo-absence data and thus don’t require as many replicates) and are more applicable when it makes little sense to generate absence data (e.g. in the case of a new invasive species). Kelly, M., Meentemeyer, R.K. 2002. Landscape dynamics of the spread of sudden oak death. Photogrammetric Engineering and Remote Sensing 68(10):1001-1009 This paper is an early attempt at integrating remotely sensed oak mortality attributed to P.ramorum with spatially distributed environmental variables in an effort to better understand factors influencing the spatial pattern and spread of the disease. Imagery was originally analyzed with a combination of unsupervised classification, spectral enhancement, and manual methods (est. 92% accuracy). This was modeled against eight landscape variables, five of which (elevation, slope, topographic moisture index, and summer and winter solar insolation) were derived from a 5m resolution DEM. The remaining three include density of California bay, Umbellularia californica, distance to the forest edge, and distance to nearest trail. An estimated rate of mortality was calculated between two years, as well as the distance between a dead crown in the second year to the nearest dead crowned neighbor of the previous year. The distance and bearing between nearest neighbor dead crowns found dead in either year were randomly distributed, though the distribution of dead crowns show significant clumping (between 100-300 m). For all combinations of variables, trees most at risk include those that are less than 6.1m from the forest edge, occur with bay, and are within 76m of a trail. The lower risk trees are those away from forest edges, on topographically moist slops, and with high summer solar insolation. While other papers have outlined the potential for inoculum to spread in the soil of hikers’ boots, this paper suggests that more important risk factors do not indicate human vectors but instead include forest structure. Kelly, M., Shaari, D., Guo, Q., Liu, D. 2004. A comparison of standard and hybrid classifier methods for mapping hardwood mortality in areas affected by “sudden oak death.” Photogrammetric Engineering and Remote Sensing 70(11):1229-1239 This paper discusses the relative accuracy of multiple classification methods (supervised, unsupervised and hybrid) used to distinguish between live trees, dead or dying tree crowns, and bare ground within remotely sensed images. Distinguishing between dead crowns and non-forested ground has been problematic with previous classification methods. In general, image classification divides data pixel by pixel into classes with defined spectral values. Supervised classification requires the input of training pixels from a defined classification scheme combined with a decision rule (e.g. minimum distance, maximum likelihood) and is subject to variation depending on the selection of training pixel structure and user subjectivity. Unsupervised classification requires less training data as an algorithm is used to generate natural groupings around user defined thresholds. Either methods have variable success over different scales and vegetation types; both methods had very little success in distinguishing between bare ground and dead crowns. In a hybrid classification there is an initial spectral stratification of the image into clusters, followed by an assignment of the cluster to the user-defined classes, and the application of a decision rule. Hybrid classification was more successful at distinguishing between the three general categories, though alternative methods are being developed. Liu, D., Kelly, M., Gong, P. 2006. A spatial-temporal approach to monitoring forest disease spread using multi-temporal high spatial resolution imagery. Remote Sensing of Environment 101:167-180 In response to the shortcomings of pixel-by-pixel, single date approaches to modeling forest disease epidemic development over time, this paper presents an alternative algorithm that classifies images based on both spectral and spatial information. Two seasons of ADAR 5500 (Airborne Data Acquisition and Registration) images were acquired for the infested study site (resolution = 1m); images were classified into three categories, bare ground (B), dead oak (D), and live canopy (F) for each year by a trained SVM. Using Markov Random Fields (MRF) models transitions between years are mapped by overlaying the two resulting images prohibiting the transition from B-D, D-B, and D-F. Classification errors involved with confusing bare ground with dead oaks are largely identified by these transitions. An iterative algorithm is use to generate a combined single framework for analysis. Mahon, L.C., Fischer, C., Levien, L., Mai K. 2002. The use of remotely sensed data for the detection, mapping and monitoring of Sudden Oak Death. Proceedings of the 9th Biennial Forest Service Remote Sensing Conference. San Diego, CA. April 8-12 The CA Dept. of Forestry and Fire Protection (CDF) and the U.S. Forest Service (FS) use Lansat Thematic Mapper satellite imagery (30m resolution) to assess the severity of SOD infection, comparing images taken in 1996 (before the epidemic was recognized) to that in 2000. The ultimate goal of this project is to determine the “hot spots” of oak and tanoak mortality, and verify these locations with more finely scaled sensed data (e.g. airborne imagery, aerial surveys, and ultimately field data). Effective collaboration between these two methods is required in SOD sensing as the affected area is too large to detect and monitor through ground sensing alone. A single vegetation layer was created predominately from the Classification and Assessment with Landsat of Visible Ecological Groupings (CALVEG) and Wildlife Habitat Relationships (WHR) classification schemes. Changes in landcover across most of costal CA between the two years for differences in brightness (measures overall reflectance), greenness (vege tation cover), and wetness (canopy or soil moisture). These discrete values of change were then transferred into one of nine change classes, ranging from large decrease to non-vegetation change). Results for China Camp area (a common SOD study site N of San Francisco) were compared to finer-scale imagery produced by ADAR. Areas suspected to have infection but for which the resolution was too coarse to detect were labeled at “potential SOD”. Accurate detection was largely influenced by the input layer data sources (e.g. CALVEG). The methods used were able to accurately detect changes in hardwood canopy cover, likely due to SOD (values not specified; see map attached to paper). While this paper outlines a useful large scale monitoring method, it must be used in conjunction with finer scale monitoring methods for assured verification. Mai, J.A., Mark, W., Fischer, L., Jirka, A. 2005. Aerial and ground surveys for mapping the distribution of Phytophthora ramorum in California. USDA Forest Service Tech. Report. In Press. This paper is the result of the collaborate monitoring efforts partaken by researchers with the USDA Forest Service region-5 Forest Health and Protection (see websites) and Cal Poly since 2001. Since 2002 a digital aerial sketch mapping system (DASM) was used to aerially digitize hardwood mortality, ultimately used to map polygons of high-mortality areas (ranging in size between 0.3 to 22,297 acres large; centroid coordinates calculated in a UTM NAD83, zone 10 or 11). Fixed wing aircrafts surveyed much of coastal CA and the Sierra foothills (variable locations throughout the years); more detailed covered was obtained with helicopter in more high priority polygons. For an example of survey scope, in 2002 winged units flew 14,5000 miles and surveyed 20,000,000 acres (20% of the state of CA; this was the max. amount of area of all years). Of these 149,000 acres were considered to have SOD suspicious mortality. The amount of remote sensing that may be done is sensitive to funds and especially to timing, before deciduous leaves turn brown and may otherwise be mistaken for mortality. Approximately 400 of 1,337 sites have been visited by field crews to verify pathogen presence. Both flight paths and field data collected by hand are provided to OakMapper. While only a small percentage of potential sites are positive for SOD, the project has detected new infection foci as well has documented intensification of infestation over the years it has been operational. Meetnemeyer, R., Rizzo, D., Mark, W., Lotz, E. 2004. Mappig the risk of establishment and spread of sudden oak death in California. Forest Ecology and Management 200:195-214 The potential for foliar hosts to support rapid sporulation of P.ramorum, the ability for these propagules to disperse aerially, and the broad range of potential hosts makes SOD a major threat to coastal forests. Monitoring and preventative control measures are often only practically applied to limited areas, therefore the need to predict the potential range of the pathogen, and thus the best areas of treatment, is required. This paper presents a rule-based model, incorporating the effects of spatial and temporal variability of multiple factors, which is used to predict the spread and establishment risk in California. Host distribution (CalVeg, see websites) and monthly weather conditions (PRISM, assumes elevation is the most important landscape factor controlling climatic variation within the landscape. see websites) suitable to disease development were analyzed in a GIS to generate a per-variable, per-month risk ranks. A prediction of spread risk was then generated by combining the variables over the reproductive time period of the pathogen (6 months between December and May). At publication, P.ramorum was not yet found at 48% of field samples mapped moderate risk or higher; these areas include coastal forests North of Sonoma County to the Oregon border and limited areas in the Sierra Nevada foothills which may remain uninfected due to lack of exposure. 21% of the areas where P.ramorum is found were mapped as low risk, probably due to patchiness of host vegetation (CalVeg mapping unit = 1 ha), or incorrect vegetation designation. Further information needed to increase the power of the model includes the mode of spread of various hosts (e.g. turbulence aerial dispersal, rain splash, or vertebrate vectoring, which influence long distance dispersal events), land usage history / fire history in response to disease establishment, and the spatial arrangement of infested versus non-infested stands (e.g. the likelihood that small and isolated stands of host vegetation will become infected versus a large and nearly continuous stand). Pu, R., Foschi, L., Gong, P. 2004. Spectral feature analysis for assessment of water status and health level in coast live oak (Quercus agrifolia) leaves. International Journal of Remote Sensing. 25(20):4267-4286 Remote sensing of SOD largely relies on the detection of dead crowns; however the method does not serve to prevent the spread of the pathogen as sporulation may take place before the entire canopy dies. This research was done under the supposition that the foliage of oak trees infected with P.ramorum have a different water stress level than healthy leaves even when they are still green. Previous work on plant water status has correlated the spectral absorption features of water (shortwave infrared) in leaves to the relative water content (RWC, a common measure of plant water stress) of the leaves. This work verifies these findings in detached Q.agrifolia leaves, outlining differences between the magnitude and wavelength position in absorption patterns of leaves with different RWC. Currently, however, this study has limited application for aerially sensed data due to the influence of atmospheric water vapor. An algorithm to correct this problem for this application has yet to be developed. Wilson, B.A., Lewis, A., Aberton, J. 2003. Spatial model for predicting the presence of cinnamon fungus (Phytophthora cinnomomi) in sclerophyll vegetation communities in south-eastern Australia. Austral Ecology 28:108-115 A GIS spatial database was used to develop predictive associations between spatial variables and sclerophyll infection by P.cinnomomi. Seventeen site variables were used including aspect, slope, altitude, distance, distance and elevation from roads or tracks, relative elevation of site, road categories and frequency of use, average height of trees, foliage cover, soil profile, drainage characteristics, and others. The spatial database was developed for basic visualization and mapping, estimation of spatial variables compared between infested and uninfested sites (e.g. slope depicted in an interpolated DEM), and the spatial extrapolation of regression models for the probability of pathogen presence. 97.1% of infested sites were correctly classified as infested in a regression model; when the spatial regression is extrapolated over an entire area (beyond the study plots) with the DEM and a sun-index model, the results show increasing probability of infection in lower cachements (reasonably, as P.cinnomomi is water spread) and in locations with a higher sun index. 3rd SOD Science Symposium presentation summaries Following are summaries of the ten minute presentations related to GIS and landscape monitoring of SOD. They were given by the first authors at the Sudden Oak Death Science Symposium III, held in Santa Rosa, California March 5-9, 2007. While these summaries are drawn from my own notes taken during the presentation, the title and page numbers refer to the unpublished abstracts supplied by the authors to the symposium’s abstract booklet. In some instances numbers do not agree between the two sources; I present my own notes on the supposition that what was presented is a more recent re-assessment of the researcher’s findings. Condeso, T.E., Meentemeyer, R.K. Landscape connectivity influences the establishment of Phytophthora ramorum. Sonoma State University and University of North Carolina at Charlotte, respectively. abstract pg 18 This presentation is a summary of findings of the paper Condeso and Meentemeyer (2007) with particular emphasis on the patch dynamics of successful pathogen dispersal and establishment. After accounting for plot-level variables of host abundance, elevation, canopy cover, and microclimate, disease severity of SOD was lower in forest fragments than larger, contiguous stands. The model generated was non-significant for patches with less than a 50m radius, and best described patterns seen in patches with a radius of 200m. They compare these results to patch dynamics as expected under island biogeography theory, whereby smaller patches are 1) less likely to be successfully colonized by dispersing propagules; and 2) are more likely to experience local extinction of pathogen populations. Personal interpretation: while 200m is concluded to be the effective limit of dispersal distance, this is related only to CA bay foliar hosts to Quercus bole hosts. Extrapolation to other systems is suspect given the differential production of sporangia on different foliar hosts, amount of inoculum needed to infect any host (e.g. tanoak which is much more susceptible to P.ramorum and probably requires less inoculum to initiate infection, thus the resulting effective dispersal distance would be greater). Hunter, R. Predicting the spread of sudden oak death in California: Spatio-temporal modeling of susceptible-infectious transitions. University of North Carolina at Charlotte. (no abstract) A cellular automaton-like model, based on the transition between an uninfected 250m plot to an infected unit in weekly time-steps, has been generated to predict the spread of SOD. Probability of infection was guided by the following parameters: host availability within the plot (generated from CALVEG, see websites); ranks of host susceptibility and sporulation capacity; human density (assumes humans are epidemiologically important vectors); and either an inverse power dispersal model or a negative exponential dispersal model operating with potential long distance dispersal events upwards of 3 km. A short term goal of this project is to contrast the distribution predicted by each dispersal curve to actual observed field data. From the time of initial infection (est. 1990, epicenter in the San Francisco bay area) both models over-predicted spread the pathogen; however the negative exponential model fit the field data better than the inverse power dispersal model. Kelly, M., Tuxen, K. Understanding the spatial component of sudden oak death. University of California Berkeley. presentation and poster abstract pg 115. The OakMapper project has been outlined in numerous papers thus far, however this presentation introduced new presentation formats, outlined the downfalls of webGIS as a monitoring program, acted as a call for free sharing of information by SOD researchers. In addition to the layer-interactive OakMapper, SOD infection info is now also available through GoogleMaps (also interactive, though with fewer options), and GoogleEarth (see websites). Images from the latter two programs, through they are increasingly being offered at resolutions capable of remotely sensing SOD, are not easily used to analyse spread. Metadata associated with each image is not well presented by Google (e.g. when was that image last updated?) and old images are not accessible. They do, however, serve to provide a small snapshot of areas of infection. OakMapper, as a webGIS system focused on civilian use, is not random sample (with an over-sampling of public lands with easy access), does not report negative confirmations (to minimize false negatives), and does not include nursery data. These serve to provide a more conservative picture of where P.ramorum may be found. Most importantly, the resolution of SOD infection is very incomplete. For example, a site may be heavily infested, though only one or two trees within this site will have a positive confirmation (indicated by a single shapefile in the database). Polygons need to be created to better represent continuous areas where SOD might be found. Magarey, R. Climate-host mapping of Phytophthora ramorum, causal agent of sudden oak death. USDA APHIS PPQ Center for Plant Health Science and Technology. Presented by Glenn Fowler. (no abstract). This presentation is an expansion on a drafted report on SOD climate prediction by the NCSU/APHIS Plant Pest Forecast (NAPPFAST) System (see websites) entitled: Climate and host risk map for sudden oak death risk (Phytophthora ramorum) by Magarey, R, Fowler, G., Randall-Schadel, B., USDA_APHIS_CPHST_PERAL and Clunga, M., Michigan State University. CPHST NAPPFAST SOD Risk Mapping Report Oct. 06, 2006. Assuming that P.ramorum is climatically limited by temperature for growth and moisture requirement for infection, a national climate and host risk map was generated to predict areas of potential invasion. For each 10 km pixel across the country, individual days are assigned a binary value (0, unfavorable for infection, or 1, favorable for infection), which were then added over the entire year. A layer of pixels for which the sum was greater than 60 favorable days was generated and averaged with a susceptible host area layer. A new addition to the layer, included in the presentation, includes the understory host composition. As expected, risk is high along the western coast forests, with marginal risk to the Sierra foothills. Risk is also very high along the Appalachian Mountains, though the disease has not yet been found outside of nurseries in these areas. The speaker also presented a global risk map, generated from climate data gathered by the International Panel on Climate Change (IPCC, see websites) with 55 km resolution. Out of ten years, the frequency of years for which at least two months were favorable for the pathogen denoted relative risk. Regions for high risk include areas in all continents (expect Antarctica); for example regions with a classic Mediterranean climate (e.g. South Africa, coastal CA), the whole of Western Europe, New Zealand, Japan, and SE China. These are rough maps only, but outline potential areas to be considered when looking for the origin of P.ramorum. Meentemeyer, R., Anacker, B., Rank, N., Cushman, J.H. Influence of historical woodland expansion on the establishment of Phytophthora ramorum. R.M. University of North Carolina at Charlotte; rest Sonoma State University. abstract pg 47-48 This presentation expands slightly on the conclusions of Condeso and Meentemeyer (2007)— which found evidence for greater disease severity in continuous forest that smaller patches— by explaining how humans have influenced forest composition between 1942-2000. They hypothesized that the change in fire regime between these times resulted an increase of woodland density and decrease in gap (distances between forest patches) size, thus exacerbating the disease. Digitized aerial photographs of their study plots were compared to photos of the same area in 2000 (min map unit = 8 square m); finding include a 25% increase in woodland area (at the expense of chaparral and grassland), a twofold increase in development, and no change in the amount of land devoted to agriculture. Survey plots that were included in the expanded area (nonwoodland in 1942, but woodland in 2000) had a greater number of infected bay leaves and density of bay stems. Shoemaker, D., Meentemeyer, R., Oneal, C., Rizzo, D. Quantification of sudden oak death tree mortality in the Big Sur Ecoregion of California. D.R. University of California Davis; rest University of North Carolina at Charlotte. abstract pg 64-65 This study was performed to assess the impact of invasives, particularly P.ramorum, on the landscape scale of forest biodiversity, structure, and function. Using remotely sensed imagery (collected April-May 2005, 1/3m resolution), they hoped to quantify the number of trees killed since the start of the epidemic, total basal area loss, and spatial distribution of susceptible trees. Vegetation was classified into host classes: redwood-tanoak, mixed broad leaf evergreen, and non-host (grassland, chaparral, etc.) by a combined object-base classification and manual editing (accuracy est. 83%). They estimated that 214,134 trees have been removed (2/3 of which include tanoaks in the redwood-tanoak habitat), at a total 262,154 square meters of lost basal area. 17% of the regions host habitat displayed some mortality attributed to SOD. The methods used likely underestimate mortality as remotely sensed images cannot detect understory vegetation (small tress, stump sprouts etc that are also infected) and are sensitive to shadow and terrain effects. The following are site links that were referenced in the above papars and presentations, as well as a few references that are especially essential to the monitoring and research done on P.ramorum. Regional and University Operated Sites California Oak Mortality Task Force (COMTF) Monitoring Sudden Oak Death
Direct Link to OakMapper webGIS application: University of California, Berkeley Forest Pathology and Mycology Laboratory Kelly Research and Outreach Lab Geospatial Imaging and Informatics Facility (GIIF) North Carolina State University Plant Pest Forecasting System (NAPPFAST) SOD in Oregon Nurseries Government Agency Operated Sites USDA Animal and Plant Health Inspection Service (APHIS): Phytophthora ramorum USDA Forest Service region-5 Forest Health Monitoring (FS-FHM) Program USDA Forest Service region-5 Remote Sensing Lab Geo-referenced Climate Resources Daily Surface Weather and Climatologically Summaries (DAYMET) International Panel on Climate Change (IPCC) Parameter-elevation Regressions on Independent Slopes Model (PRISM) The Future in Disease Mapping: the Phytophthora Database Phytophthora database Registered researchers can input genetic sequences, isolate morphotypes, and species locations into the database. Currently they database contains 2,428 sequences and 1,128 isolates, representing 85 distinct species. Any individual can access these records, as well as detailed descriptions of species phylogeny, morphology, disease characteristics, life cycle, and photos of symptoms on susceptible hosts (as in all cases, when known). Researchers may also input their own genetic sequences of an unknown Phytophthora to search for the closest genetic match.
|