A few things you should know
right away!
If nothing seems to be happening
when you click, it's because the browser window (in which the output should
have appeared) is already open/iconized and is hiding behind other windows
or sitting iconized.
In MS Internet Explorer (earlier versions
of 4.0), there is a bug that does not let you jump to the links in the
same page correctly. This could happen when you click
button. If so, please scroll to the bottom of the page where the target
is.
If the screen appears garbled or
missing
content (when you resize the screen for example), RELOAD (REFRESH in
Internet Explorer) !!
If you get JavaScript errors, your
browser is old. Close all browser-windows that you may have opened,
go back to the starting (index) page, and click on "Old Browser".
Also be aware that some older versions of Netscape 4.0 and Internet Explorer
4.0 have bugs! Unfortunately it is very difficult to make DHTML compatible
on all versions of all browsers!!
If you are using old browser, you
may not be able to get context-sensitive help to work for
the distribution maps - however the same help is available in the table
on the main species page.
If you don't understand the acronyms/abbreviations
used, be sure to click
page. This also appears on all the Species Pages.
FAQ
(Frequently Asked Questions)
Navigating this web site
How do I get back to the starting
atlas web page? I seem to have gotten rid of that window and don't know
how to get back.
You'll have to get to the Launch-Pad Window which
has a link back to the starting web atlas page. So, click on the New
Launch-Pad Window button in the a_species page,
and then click on the Main
Atlas Page.
Methods
Where can I get information
on the methodology used?
The best place to look for more information on the methodology
used is to refer to the publication in Ecological Monographs. An on-line
version of the paper is available.
What is RTA (regression tree
analysis) and where can I get some background information?
The best place again is this
section of our publication and also the references cited there.
Why did you choose county
as the level of resolution?
We were forced to since the best
resolution for many of the predictors was at a county level. Also, if we
had used FIA plot-level, we would have had to deal with spatial interpolation
and geostatistical complexities that would dominated the entire project!
Why have you
included some strange variables like CROP.LND GRAZE.PST, DIST.LND etc.
in you predictors?
Since we knew that FIA data were
influenced by anthropocentric pressures (see this FAQ
item), and also that the distribution of certain species can be
influenced to some extent by humans, we decided to include many of these
landscape-modifier variables to help the model fit the FIA data better.
We tested the model by including climate/soil/physiographic variables only
and found that the addition of landscape-modifier variables improved the
fit.....the limitation of this approach is that we can't predict how these
variables would have changed in the future GCM scenarios.
Can you provide R2
values for the model?
The equivalent of R2 is
provided for each species in the RTA tree diagram which you can
access as follows. In the a_species page, click on Current-FIA
link under Distribution Maps heading. Then click on the "Geographic
Predictors" map's legend to bring up the RTA diagram, whose
heading contains the R2. There is also a blurb on its meaning
for RTA in the help button provided.
What software did you use to
develop the models?
We used S-PLUS (Mathsoft)
for statistical modelling and Arc/Info (ESRI)
for GIS. We used a lot of Unix shell tools and Perl
scripts for database manipulation, calculation, and web-translation. We'd
like to make special mention of the "rpart" module (classification and
regression tree software) developed for S-PLUS by Terry Therneau and Elizabeth
Atkinson of the Mayo Clinic,
Rochester, Minnesota (biostat-info@mayo.edu). Please note that we are NOT
endorsing these products in any way - we merely used them.
Tree species data - range and distribution
What is FIA - can you give me
some background information?
The USDA Forest Service periodically determines the extent,
condition, and volume of timber, growth, and removals of the Nation's forest
land by the work of six Forest Service Forest Inventory and Analysis (FIA)
units. Four FIA units produced a data base of standard format called the
Eastwide Data Base (EWDB) for the 37 states from North Dakota to Texas
and east. These data are stored in three record types (Hansen et al. 1992):
county data, plot data, and tree data. Plot locations are not precisely
located but county location was provided for each plot. We used the data
from more than 100,000 plots and nearly 3 million trees to summarize the
desired county-level information needed for this study.
Hansen, M. H., T. Frieswyk, J. F. Glover, and J. F.
Kelly. 1992. The eastwide forest inventory data base: users manual. General
Technical Report NC-151. USDA Forest Service, North Central Forest Experiment
Station. St. Paul, Minnesota.
Are there any limitations/problems
in using FIA data in your models?
FIA sampled data is on existing
natural forests and plantations - so if a county is poorly forested due
to anthropocentric pressures (cropland, urban-land etc.), the importance
value of any species in that county could be artificially low (since we
aggregate plot-level data to county-level in our analysis - see this FAQ
item). This can cause misleading correlations with climatic, edaphic
and physiographic variables in our model and could cause errors in the
model output - thus influencing the way they would respond to climate change
in the GCM scenarios.
What is IV (Imp. Value) - how
was it derived?
We generated importance values (IV) for each species
as follows:
IV(x) = 100*BA(x) / BA(all
species) + 100*NS(x) / NS(all species)
where x is a particular species on a plot, BA is basal
area, and NS is number of stems (summed for overstory and understory
trees). In monotypic stands, the IV would reach the maximum of 200. The
IVs were rounded to decimal numbers with one exception. If the IV was greater
than zero but less than one, it was assigned to one. This decision was
taken since rounding would have falsely turned species-present counties
to species-absent counties.
How can I download the database
for Importance Value and Species for all/some of the eastern states?
You can download the IV-Species data for current-FIA
and all the GCM model scenarios. On the launch pad page click on "Outputs
for ALL 80 Species Combined" -> "Ranked Species List" -> "Current" (or
any of the 5 scenarios). At the bottom of the eastern US image map, there
is an option to download the entire database (about 3-4 MB). This should
give you a sorted SpeciesName-IV list for each county for all the eastern
states OR alternatively you can click on any state you want and get the
database for all counties in that state.
On what basis did you select
the 80 tree species?
Sample restrictions made us select only 80 of the 196
tree species. We selected species that have a recorded minimum IV of 3.0
in each of at least 100 counties. Thus our sample represents the more common
tree species in the eastern US.
Why do some species have more
"No Data" areas than others?
In the ideal world, we would have data for each pixel
and all that would based on perfect estimation. Alas in reality it is not
so. The NoData areas refer to the non-availability of data, which
is a straight-forward paucity of data - however, some species have
more NoData than others for several reasons: (1) there were no data
for any tree species for those counties; this occurs for many counties
in the prairie states, especially Oklahoma and Texas along with a spattering
of counties elsewhere in the western part of the region, and in the southern
tip of Florida; (2) one or more of the four FIA units (northeastern, north
central, southeast, southern), charged with the collection of the FIA data,
did not report the species in their data base, most likely because it was
not present in the region; this occurs for several distinctly northern
(e.g., Betula papyrifera, Populus tremuloides) or distinctly southern
(e.g., Pinus elliotti, Pinus palustris) species; or (3) one or more
of the FIA units do not recognize a particular species name as present
in the unit, even though it undoubtably is present but called something
else; this occurs especially in the Carya genus, where some units
lump species into Carya sp., while others identify them to species.
Unfortunately, it is not always easy to distinguish (2) from (3). If the
species distribution is wholly encompassed with some white (importance
value < 1.0) zones around the range and outside the gray zone, it likely
is scenario (2) where the species is not present in the gray zone; if the
distribution abruptly ends at the gray boundary, it is probably a taxonomic
confusion (scenario 3).
What are Little's boundaries?
These are boundaries that Elbert Little Jr. delineated
from various sources for several tree species. These distribution ranges
are well known and frequently cited (e.g., in the Silvics manuals) and
hence provides a useful comparison to the FIA distribution maps.
Little, Elbert L., Jr. 1971. Atlas of United States
Trees (Vol. 1. Conifers and Important Hardwoods). Miscellaneous Publication
No. 1146, U.S. Dept. of Agriculture. Forest Service. Washington D.C.
Little, Elbert L., Jr. 1979. Checklist of United States
trees (native and naturalized). U.S. Department of Agriculture, Agriculture
Handbook 541. Washington, DC. 375 p.
I need only current
data - I don't care about the global change scenarios.
Much of the data pertain to the current distributional
status of the species, or of current ecological or life history attributes
of the species. For current distribution information, just look at
the current (FIA actual data) or current-predicted (modeled
output for current condition) scenarios in all our outputs. For current
attributes of the species, see the next FAQ question.
Tree species data - attributes
How do I get attribute data
about particular tree species?
-based on GIS overlay of environmental
data and actual distributions
Click on the species of interest, then the box:
Besides statistical summaries by species, its rank among all 80 species
is given.
How do I get attribute data
about particular tree species?
-based on literature survey
of life history attributes and disturbance responses
Click on the species of interest, then the box:
How do I get attribute data
about particular tree species?
-based on a link to the "Silvics
of North America" manuals
Click on the species of interest, and on the 4th line:
Silvics
Manual:
Environmental data
Can you give me some background
information on the predictor variables you have used in your model?
Once again it is best to refer you to this section
of our Ecological Monographs publication. A large amount of effort went
into deriving the data base as well as thinning the number of variables
to 33 (from over 100).
Can you give me summary statistics
on the predictor variables you have used?
Yes we can - click
here please.
I don't understand the "Geographic
Predictors" map nor its legend.
The predicted maps give the overall scenario of distribution
for the species. In addition, if you are interested in finding what predictor-variables
are driving the distribution according to the RTA model,
you can figure that out in the "Geographic Predictors" map. Since RTA is
a rule-based model based on recursive partitioning, this map is calculable
(with some difficulty!) and is of considerable interest. For more
details on the interpretation please refer to the context-sensitive
help links provided. Also, you can look at the regression
trees and the results-discussion
section
of our publication in Ecological Monographs.
Global change scenarios
What is GCM?
GCM means Global Circulation Model. The more recent ones are coupled
ocean-atmospheric models that are used to predict climate change under
various "man made" scenarios. We are using the climate scenarios under
doubled carbon dioxide of several GCMs to predict the distribution of species.
According to the Intergovernmental Panel on Climate Change, we are looking
at a doubled carbon dioxide level in our atmosphere by the year 2100.
Are you using the latest GCM
models?
The RTA model uses equilibrium
2xCO2 GCM conditions to predict potential future distributions.
It is essentially a model predicting potential future suitable habitat
for each species, so that the assumption is made that the species will
be able to colonize all suitable sites. There is no real time component
to the model, although predictions from the Intergovernmental Panel on
Climate Change show that, if carbon dioxide emissions were maintained at
1994 levels, the 2xCO2 level could be reached by the end of
the 21st century (Houghton et al., 1996).
We have used GCM models for which temperature, precipitation
and PET values were already calculated for the United States. This condition
means that we may not necessarily have access to the latest models. However
it is easy to plug in the latest scenarios when the above three values
are estimated, at a reasonable spatiial resolution, for the United States.
Five climate scenarios were used to evaluate possible
future species distributions: (1) the Geophysical Fluid Dynamics Laboratory
(GFDL) model (Wetherald and Manabe, 1988); (2) the Goddard Institute of
Space Studies (GISS) model (Hansen et al., 1988); (3) the United
Kingdom Meteorological Office (UKMO) model (Wilson and Mitchell, 1987);
(4) the Hadley Centre for Climate Prediction and Research (Hadley) model
(Mitchell et al., 1995); and (5) the Canadian Climate Centre (CCC) model
(Boer et al., 2000 and Kittel et al., 2000). The Hadley and CCC scenarios are transient scenarios;
for these, 30 year climatic averages were estimated for the period 2071-2100
(Neilson, personal communication). These five scenarios give a good range
of possible outcomes in equilibrium climate at 2xCO2 (Table
1). The Hadley scenario has the least severe change in temperatures, especially
January temperature, while the UKMO predicts a very large change in January
temperature. Precipitation shows little change in each scenario except
UKMO and Hadley with significant increases in predicted precipitation .
The outputs from the GFDL and GISS models were acquired
in 10 x 10 km format (U.S. Environmental Protection Agency, 1993).
The Hadley, CCC, and UKMO data were obtained from the USDA Forest Service
Laboratory in Corvallis, Oregon in 0.5 x 0.5o
format (Neilson and Drapek, personal communication). Importantly,
the latter three data sets had relatively higher PET values compared to
the original three data sets. The later PET values were calculated using
a slightly different model compared to the earlier data sets (Neilson,
personal communication). However, we believe the impact is minimal,
and we report the model outcomes here because of the consistent response
across all scenarios, and because in most of the models , the PET-related
variable comes out low in the binary regression tree, and as such, would
impact few counties.
Hansen, J., Fung, I., Lacis, A., Rind, D., Lebedeff, S.,
Ruedy, R., 1988. Global climate changes as forecast by Goddard Institute
for Space Studies three-dimensional model. Journal of Geophysical Research
93, 9341-9364.
Houghton, J.T., L.G. Meira,Filho, B.A. Callander, N.Harris,
A. Kattenberg, and K. Maskell. 1996. Climate Change 1995: The Science of
Climate Change. Cambridge University Press, Cambridge, UK. 572 pp.
Kittel, T.G.F., J.A. Royle, C. Daly, N.A. Rosenbloom, W.P. Gibson, H.H. Fisher, D.S. Schimel, L.M. Berliner, and VEMAP 2 Participants (1997). A gridded historical (1895-1993) bioclimate dataset for the conterminous United States. Pages 219-222, in: Proceedings of the 10th Conference on Applied Climatology, 20-24 October 1997, Reno NV. American Meteorological Society, Boston.
Kittel, T.G.F., NA Rosenbloom, C. Kaufman, JA Royle, C. Daly, H.H. Fisher, WP Gibson, S. Aulenbach, R. McKeown, D.S. Schimel, and VEMAP2 Participants (2000). VEMAP Phase 2 Historical and Future Scenario Climate Database. Available online at [http://www-eosdis.ornl.gov/] from the ORNL Distributed Active Archive Center, Oak Ridge National Laboratory, Oak Ridge, Tennessee, U.S.A.
Boer GJ, Flato GM, Ramsden D. 2000 . A transient climate change simulation with historical and projected greenhouse gas and aerosol forcing : Projected climate for the 21st century. Climate Dynamics 16: 427 - 451.
Mitchell, J. F. B., Johns, T. C., Gregory, J. M., Tett,
S., 1995. Climate response to increasing levels of greenhouse gases and
sulphate aerosols. Nature 376, 501-504.
Wetherald, R. T., Manabe, S., 1988. Cloud feedback processes
in a general circulation model. Journal of Atmospheric Science 45, 1397-1415.
Wilson, C. A., Mitchell, J. F. B., 1987. A doubled CO2
climate sensitivity experiment with a global climate model including a
simple ocean. Journal of Geophysical Research 92 (D11), 13315-13343.
Can you show me quickly how
the GCM models compare among themselves and with the Current scenario for
the climate variables you have used?
Yes, we can - please click here
to view the boxplot comparison.
Why have you used
outputs of so many GCM models - what are the differences among them?
Since climate change is a dynamic and uncertain topic,
each model has its inherent strengths and weaknesses. By using the output
of several GCMs we can hopefully arrive at a range of distribution scenarios,
whose average may represent reality better. Differences among them is best
described by the following statistics:
Area-weighted averages for the Eastern US of each climate variable
in the RTA tree models, for current climate, and for five GCM
scenarios (2 x CO2 equilibrium runs).
JANT JULT AVGT MAYSEPT PPT PET JARPPET
CURRENT -1.68 23.53 11.60 20.63 1043.02 64.91 1.06
GISS 3.21 27.03 16.10 24.60 1067.62 104.02 0.79
GFDL 3.14 30.76 17.01 26.45 998.57 139.21 0.30
HADLEY 0.77 25.86 14.30 23.29 1284.86 179.00 0.50
UKMO 6.53 30.19 19.12 27.62 1159.37 267.30 0.28
CCC 4.88 28.52 17.19 26.04 1082.55 215.51 0.26
**********************************************************************
Change from current climate conditions as predicted by five GCMs:
absolute change for temperature variables, percent change for others.
JANT JULT AVGT MAYSEPT PPT PET JARPPET
GISS 4.89 3.50 4.50 3.97 2.36% 60.25% -25.47%
GFDL 4.82 7.23 5.41 5.82 -4.26% 114.47% -71.70%
HADLEY 2.45 2.33 2.70 2.66 23.19% 175.77% -52.83%
UKMO 8.21 6.66 7.52 6.99 11.16% 311.80% -73.58%
CCC 6.56 4.99 5.59 5.41 3.79% 232.01% -75.47%
Legend:
AVGT Mean annual temperature (EC)
JANT Mean January temperature (EC)
JULT Mean July temperature (EC)
PPT Annual precipitation (mm)
PET Potential evapotranspiration (mm/mo)
MAYSEPT Mean May-September temperature (EC)
JARPPET July-August ratio of precipitation to PET
|
Where can I get
information on the GCM models you have used?
See above. and also search under
Google (http://www.google.com).
Citations
How can I cite
this web page?
How to cite this web page:
Prasad, A. M. and L. R. Iverson. 1999-ongoing.
A Climate Change Atlas for 80 Forest Tree Species of the Eastern United
States [database]. http://www.fs.fed.us/ne/delaware/atlas/index.html, Northeastern
Research Station, USDA Forest Service, Delaware, Ohio.
Hard copy version of the climate change tree atlas:
Iverson, L. R., A. M. Prasad, B. J. Hale, and
E. K. Sutherland. 1999. An atlas of current and potential future distributions
of common trees of the eastern United States. General Technical Report
NE-265. Northeastern Research Station, USDA Forest Service. 245 pp.
(Contact Anantha
Prasad for a copy)
Published articles related to the work:
DeHays, D. H., G. L. Jacobson, P. G. Schaber, B. Bongarten, L. R. Iverson, and A. Kieffenbacker-Krall. 2000. Forest responses to changing climate: lessons from the past and uncertainty for the future. Pages 495-540 in R. A. Mickler, R. A. Birdsey, and J. L. Hom, editors. Responses of northern forests to environmental change. Springer-Verlag, Ecological Studies Series vol. 139, New York, NY.
Easterling, M. M., D. R. DeWalle, L. R. Iverson, A. M. Prasad, A. Z. Rose, A. R. Buda, and Y. Cao. 2000. The potential impacts of climate change and variability on forests and forestry in the Mid-Atlantic Region. Climate Research 14:195-206.
Hansen, M.H., Frieswyk, T., Glover, J.F., and Kelly, J.F. 1992. The Eastwide
forest inventory data base: users manual. General Technical Report NC-151,U.S.
Department of Agriculture, Forest Service, North Central Forest Experiment
Station. St. Paul, MN. 48 pp.
Hanson, A. J., V. Dale, C Flather, R. P. Neilson, P. Bartlein, L. Iverson, and D. Currie. 2001. Global change in forests: interactions among biodiversity, climate, and land use. BioScience 51(9):765-779.
Iverson, L. R. and A. M. Prasad. 2001. Potential changes in tree species richness and forest community types following climate change. Ecosystems 4:200-215.
Iverson, L. R. Prasad A. M. and M. W. Schwartz.
1999. Modeling potential future individual tree-species distributions in
the Eastern United States under a climate change scenario: a case study
with Pinus virginiana. Ecological Modelling 115:77-93.
Iverson, L. R. and A. M. Prasad. 2001. Potential tree species shifts with five climate change scenarios in the Eastern United States. Forest Ecology and Management 155(1-3).
Iverson, L. R. and A. M. Prasad. 1998. Predicting
abundance of 80 tree species following climate change in the eastern United
States. Ecological Monographs 68:465-485. (Ecological
Monographs paper)
McNulty, S. G., J. A. Moore, L. R. Iverson, A. Prasad, R. Abt, B. Smith, G. Sun, M. Gavazzi, J. Bartlett, B. Murray, R. A. Mickler, and J. D. Aber. 2000. Application of linked regional scale growth, biogeography, and economic models for southeastern United States pine forests. World Resources Review 12:298-320.
Prasad, A. and Iverson, L. R. 1997. Modelling
tree distributions in eastern United States using Arc/Info GIS and S-PLUS
statistical package. Pages (http://www.esri.com/library/userconf/proc97/PROC97/TO200/PAP200/P200.HTM)
in Proceedings, 1997 Arc/Info Conference. Environmental Systems Research
Institute, Inc. Redlands, California.
Prasad, A. M. and L. R. Iverson. 2000. Predictive vegetation mapping using a custom built model-chooser: comparison of regression tree analysis and multivariate adaptive regression splines. In. Proceedings CD-ROM. 4th International Conference on Integrating GIS and Environmental Modeling: Problems, Prospects and Research Needs. http://www.colorado.edu/research/cires/banff/upload/159/index.html, Banff, Alberta, Canada.
Schwartz, M. W., L. R. Iverson, and A. M. Prasad. 2001. Predicting the potential future distribution of four tree species in Ohio, USA, using current habitat availability and climatic forcing. Ecosystems.
Sutherland, E. K., B. J. Hale, and D. M. Hix. 2000. Defining species guilds in the Central Hardwood Forest, USA. Plant Ecology 147:1-19.
Caveats
Are you really
serious when you show that some species will disappear from the US? What
confidence do you place on your assertions?
Ah Ha! We knew you were going to ask the hard question
and so tried to bury it lower down - but you are persistent! Well,
what we mean is: given our assumptions and the inherent limitations
of a macro-scale, multiple-species modelling approach, in addition to the
intra-model limitations as well as input data errors, we predict that species
x will "potentially" disappear from the United States. Remember what
we lose, our friendly northerly neighbor, Canada gains! For more
details on assumptions, refer to this section
of our publication. As you can imagine, it is hard to place "confidence"
to this modelling approach. Even if we concocted something, it would be
highly misleading - so we ask you to treat this as a pioneer investigation
of the "potential" suitable habitat of 80 species under a changed climate,
given the present state of data, using a regression modelling approach.
Enough said.
Your maps depict
potential future distributions, or potentially suitable habitat for 80
species, based on GCM scenarios for roughly 100 years into the future.
These models also assume the species will get to the potentially suitable
habitat in that time frame. What might really happen over the next 100
years if migration were inhibited by fragmented habtitat or dispersal limitations?
We are working on that. Mark Scwartz (1992) developed
a model that simulates tree migration across fragmented habitats. We have
applied this model, SHIFT, to one species (Pinus virginiana) and
published the paper in Ecological Modelling (Iverson et al. 1999). In that
work, we combine the outputs from the RTA work shown on this web with that
of SHIFT to produce more logical estimates of what might really happen
over the next 100 years. We are proceeding to do this work for many other
species, as well as concentrate on SHIFT model sensitivity for four species
in Ohio. Stay tuned for more on this and visit our global change page often (SHIFT Model)
Iverson, L. R. Prasad A. M. and M. W. Schwartz. 1999.
Modeling potential future individual tree-species distributions in the
Eastern United States under a climate change scenario: a case study with
Pinus virginiana. Ecological Modelling 115:77-93.
Schwartz, M.W. 1992. Modelling the effects of habitat
fragmentation on the ability of trees to respond to climatic warming. Biodiversity
and Conservation 2:51-61.