Biocomplexity report (Dec. 2006- Dec. 2007)
Research and Education Activities
Atmospheric measurements and modeling (Fenn, Bytnerowicz, Tonnesen):
Air pollution monitoring with passive samplers and active monitors was continued for the
southern California forested sites with a special emphasis on the San Bernardino
Mountains. Results of the measurements were used for development of maps of
distribution of the pollutants using the ArcGIS Geostatistical Analyst. The 2007 season
completed five years of air pollution monitoring in the study area.
Nitrate leaching is the cardinal symptom of ecosystem N saturation. Based on field data
collection (streamwater nitrate concentrations and annual N deposition in throughfall)
from sites with a range of N deposition inputs, we have developed empirical relationships
between atmospheric N deposition and nitrate concentrations in streamwater of forested
catchments in southern California. From these relationships a critical load (CL) of N
deposition leading to elevated nitrate concentrations was determined. Simple soil and
plant nutrient status indicators were also used to determine the critical values of these
indicators, that can be used to identify forested sites that are likely in excess of a N
deposition CL. Similarly, we have developed the relationship between N deposition and
soil pH, although the focus of the current studies is on the eutrophying effects of N as
opposed to the acidifying effects. The empirical CLs determined in these studies are
being compared to CL determined from the Daycent biogeochemical model and from the
simple mass balance model, widely used to calculate CL in other parts of the world.
We have also developed the empirical relationships between N deposition and lichen
concentrations in lichen thalli along with changes in lichen communities. Based on their
response to N enrichment, lichen species can be considered as acidophytes, neutrophytes,
or nitrophytes. Acidophytes are highly sensitive to even low levels of N. By contrast,
neutrophytes are either tolerant or, in some cases, enhanced by N. Neutrophytes are not
considered weedy like the nitrophytes, however, which are fast-growing species
associated with ammonia deposition as well as substrates with relatively high pH.
Biogeochemical and hydrologic modeling and data collection (Meixner, Simunek):
In the past year we finished DAYCENT simulation at two mixed coniferous forests. Data
collection, compilation, and first model run for Coastal Sage Scrub (CSS) at Santa
Margarita Ecological Reserve (SMER) have been conducted. Field data at SMER include
multiple years of flux measurement using eddy covariance technique by Dr. Walter
Oechel’s group at California State University – San Diego. This data set is unique for
DAYCENT parameterization of CSS ecosystems.
The outputs from forest simulations had been partially presented as a poster in the 7th
SAHRA Annual Meeting held in the University of Arizona, and will present in the AGU
Fall Meeting held in San Francisco, CA. And a manuscript discussing forest ecosystem
capacity of retention deposition N is in drafting.
During the 7th SAHRA Annual Meeting, HYDRUS and DAYCENT work groups met
together and had a discussion on linking two models for the project. A subroutine adapted
from HYDRUS will be incorporated into DAYCENT. DAYCENT work group met with
others working on remote sensing and landscape transition modeling at the project
meetings. Progress was reported and model integration was discussed.
We have started working on implementing water flow routines from the HYDRUS-1D
software package in the CENTURY model. In the first step, the computer program
HYDRUS-1D was adapted and simplified for the CENTURY model. The simplification
involved removal of subroutines simulating solute and heat transport, hysteresis in the
soil hydraulic functions, and boundary conditions that were irrelevant for the coupled
model. All input, output, and actual simulation routines were grouped into three large
subroutines, so that implementation into the CENTURY model should be simplified. The
simplified HYDRUS code still considers the effects of precipitation, infiltration,
evaporation, plant water uptake, soil moisture storage, and water accumulation at the
ground surface and in the vadose zone in a much more robust way than the flow module
of CENTURY. The final adopted HYDRUS-1D code thus simulates only onedimensional water movement in variably-saturated porous media and considers water
uptake by plant roots that has a great effect on water in the root zone. This greatly
simplified code was recently forwarded to our collaborators and UA in Tucson.
To parameterize HYDRUS, Julie Scanlan, a PhD student, set up her two experimental
sites in the Joshua Tree National Park. These are in desert creosote scrub vegetation, one
with hi gh and one with low nitrogen deposition site. Nine access tubes for nuclear probe
measurements of water content were installed at each site. Water content measurements
as well soil cores are regularly collected at these two sites. Julie has also continued her
modeling work with HYDRUS-1D for two sites for which data were collected
previously, the Coastal Sage Shrub and Chaparral sites.
Remote Sensing (Gong):
1. We improved the performance of three programs including TopoRadCor (radiometric
correction for relief topography), TopoSunIncidence (solar angle correction under relief
topography), and TopoSunShadow (shadow detection and correction), so they can be
used to calculate the relative radiance for any time interval such as one day and one year
for a certain area. The total radiance energy or its change over the season calculated with
these algorithms may be used to help us classify the main ecological unit at different
areas with the help of DEM. The programs can also be used in other models that require
surface energy calculations.
2. Using the preprocessing flow developed last year, we calculated the topographiceffect-removed radiance and reflectance products for 2 seasons in 2005 and 1 season for
the years 2002, 1998, 1995 and 1993, respectively. They are masked by the boundary of
our study area. Furthermore, some traditional classification results for each season are
obtained such as the unmixing result of leaves, dry tresses and soil background. Some
seasonal differences for 2005 and annual differences for other years are calculated and
compared to see the primary change features in the area. For all these parameters, we still
need to obtain ground truth data. All calculated results have already been provided to
Professor Larry Li’s landscape modeling group.
3. We applied the CASM and IARS model, which were originally developed by us for
hyperspectral image analysis, to the classification and nonlinear-unmixing of the 5-band
multispectral TM data under the support of DEM and its derivatives. We already began
the experimental test that involves three steps. First, the spectral feature of vegetation and
soil in our study areas must be limited to less than 5 parameters; second, a near-linear
simplified mixing model is required; third, a method should be found to make relative
processing and then validate the result in the field. Now our work is at the first step and
we have done some tests for the second step. Some original results are obtained and show
some positive effect, but the results also show that we should further improve the
algorithms. A process only using CASM model, which considers the spectra as a function
of a special base spectrum, is also under development. We hope that such a processing
flow will be helpful for classifying different vegetation types, especially for the invasive
grasses.
4. We also analyzed the original results obtained above. Our initial analysis shows that
some background features or background changes in different years can be discovered or
highlighted after processing the data with our algorithms. For example, the difference
between the summer data of 1995 and 1998 show that in the west part of Joshua Tree
National Park, a special change in a certain area that may be caused by a fire.
Vegetation type conversion and fire modeling (Allen, Minnich, Sadler):
The following tasks are in progress or completed:
We assessed the accuracy of the % cover classification of the California Department of
Fish and Game 2005 Western Riverside County Vegetation Map and completed the
Weislander (1933) vegetation map classification. We calculated environmental variables
to add to our existing WRC map points dataset to model the process of invasion of exotic
grasses on a landscape scale. We are calculating fire regimes from the 1930’s to 2006
FRAP data. Using Greg Miller’s (project graduate student with Minnich and Sadler) fire
tool (see below) we can calculate for each unique fire regime polygon the following
variables: ave, min and max fire interval, # fires, time since most recent burn, average
size of burn. These data will be compared with Gail Tonnesen’s current model layer of N
deposition at a 4 km grid, including monthly and annual values for dry wet deposition of
oxidized and reduced forms of N. We have monthly, annual, and rainfall year
precipitation amounts available from some weather stations and from Oregon State
University’s PRISM model. Rainfall will be back calculated to the 1930’s and related to
fire events. Biomass data are provided from MODIS data product MOD1&A3 with
annual net plant productivity at 1 km resolution for 2000-2006. We can calculate plant
productivity in grams of carbon per square meter. Some vegetation clipping data are
available from Minnich (15 years) and Allen’s groups (13 years) at two sites. Miller
imported agency fire data for other areas to test software tools and replicate published
findings concerning fire hazard as a function of fuel age.
Next we plan to construct and evaluate logistic regression models predicting categories of
% conversion to grassland from coastal sage scrub using the environmental variables
listed above. We can compare models using AIC values. We will include climate and
topographic variables in our models that have already been calculated for the study area.
Some of these variables include different temperature parameters, elevation, soils. From
this we will calculate how much coastal sage scrub has converted to non-native grassland
in WRC. We will compare the 1930s Weislander map to the CDFG 2005 vegetation map
(we will screen for recent fires).
Analytical Software Development: GIS-Matlab-Python Toolbox– Miller and Sadler
wrote GIS-Matlab tools to capture and analyze historic fire distribution data and model
data; tools compute both traditional and novel (from Box Springs study) statistics and
prepare maps and diagrams; Miller is now using Python to write a Graphical User
Interface that will insulate users from command-line interfaces and maximize use of
platform-independent shareware programs.
Modeling: Sadler and Miller are analyzing form of hazard functions (probability of
burning as a function fuel age and level) in chaparral fire mosaics, given complete spatiotemporal information of fire and vegetation history; systematic examination of impact of
changing productivity, weather, topography, and fire suppression on emergent mosaic
and hazard functions.
With empirical and model data Miller is studying the impact of censored and incomplete
data on traditional and novel fire history statistics. The goal is to write algorithms and
user interfaces that minimize the misleading impact of missing data, while maximizing
the utility of the data and ensuring that any biases are explicit. Currently, the best
approach seems to be to make the censoring and its impacts explicit in all graphical
summaries.
To determine ignition probabilities, Sadler is redesigning model algorithms to increase
simplicity and flexibility of functions while providing better separation of probability
distributions with fuel age and level before and after canopy closure.
Further development of model out put is underway: 1) to feed maps from successive steps
into GIS database and toolbox; 2) to distinguish probability of burning for those model
pixels actually reached by fire; 3) to calculate and display hazard functions by fuel age
and fuel level.
The landscape modeling (Li) has focused on theoretical and modeling methodological
development:
1. We modified the classic competition-colonization model by an explicitly linked
resource competition model for explaining plant species diversity and exotic species
invasions at local spatial scales.
2. The mechanistic theory of plant competition for belowground resources has been based
on the concept of local interactions of individual plant with their environment and
interactions between plants are indirect. We explore a simple mechanistic model for
plant-to-plant direct competitive interactions by overlapping their resource depletion
zones.
3. We developed a theoretical modeling framework for post-fire ecological succession,
mainly for Mediterranean vegetation, based on a three-level hierarchy of successional
causes. Within this proposed framework, fire effects are considered by associating them
with the number of burned sites that open-up and specific changes at the burned sites
relative to unburned sites. Three distinct site-specific neighborhoods are constructed;
changes of neighborhood structure allow temporal replacement of plant species by
another plant species with greater maximum size, age and lower maximum growth rates
and dispersal abilities. The proposed framework can be used for developing a spatially
explicit individual-based model, which will be useful for monitoring and predicting
successional changes and hence for restoring native grasslands or shrublands.
We are just starting to work with some of the data to test our theories and models and are
working with the remote sensing group for some scaling analyses of imaging data.
Findings
Atmospheric measurements and modeling:
High levels of all measured air pollutants (O3, NO2, HNO3 and NH3) characterize the
study area, as well as high levels of ozone both in high and low elevation sites (SBM,
Banning Pass and JTNP). These levels have a strong phytotoxic potential. High levels of
NO2, HNO3 and NH3 are mostly found along the Interstate 10 corridor (and much lower
concentrations are at high elevation sites). Strong N deposition gradients (W-E) are
present in the San Bernardino Mountains and Joshua Tree NP. Deposition of N to plant
surfaces in the Banning Pass is very high. This phenomenon results from very high
concentrations of HNO3 and NH3 and steadily high winds. High values of N deposition
to plants in winter can result from lower rates of NH4NO3 volatilization compared to
summer
The empirical N deposition critical load (CL) for nitrate leaching in mixed conifer forests
was estimated to be 17 kg N/ha/yr and the areas exceeding the CL within the San
Bernardino National Forest have been identified. Simulations with the Daycent model are
being compared to the empirical results and the model is being used to determine the
number of years of N deposition until the CL is exceeded and to evaluate the various
environmental and biological factors affecting the CL. Results suggest that the widely
used simple mass balance model is not effective in determining the “N as a nutrient” CL
for the mixed conifer forests in southern California.
Critical loads determined from lichen responses were very low because of the highly
sensitive response of lichens to N enrichment. The CL based on lichen community
responses ranged from 2.8 to 5.2 kg N/ha/yr depending on which lichen response is
considered. This work was done over a range of sites in the Sierra Nevada, because
lichen communities in the San Bernardino Mountains in the Los Angeles Basin
experience N deposition loadings well in excess of these CL values. The significance of
the lichen CL is that at sites with N deposition below the CL for lichen impacts, one can
be relatively certain that the ecosystem will be protected from the detrimental effects of
chronic N deposition. The lichen-based CL establishes a lower bound for the ecosystem
CL for N deposition.
Biogeochemical and hydrologic modeling:
DAYCENT modeling pursued understanding the mechanistic fate of N deposition in
forested and CSS ecosystems. In forests litter is an important ecosystem component.
Articles have reported that litter C/N can be as low as 26~30 under high-N deposition at
the Camp Paivika site. Besides litter decomposition and respiration (i.e., C loss), N
absorption by litter materials could be another mechanism of ecosystem N retention.
This hypothesis was demonstrated and tested in the process of DAYCENT
parameterization.
In the studied forests, plants, litter and SOM were shown to retain 80~90% of N
deposition, while gaseous N emission, that occurred in or after the winter season, could
explain most of the remaining deposited N. N leaching losses were highly variable
depending on precipitation. DAYCENT was applied to predict the critical N load (CL) to
forested ecosystems in terms of “N saturation”, one of the issues addressed in this project.
DAYCENT could be calibrated to simulate CSS monthly (or weekly) CO2 exchange, as
demonstrated in initial runs. DAYCENT however was inadequate at capturing seasonal
CSS biomass dynamics, because it does not represent the seasonal quick re-growth and
dieback of woody tissues characteristic of CSS ecosystems. Model modification to
accommodate this type of plants will be needed for this project, and will contribute to
improvement of the DAYCENT model.
Vegetation type conversion, N deposition, and fire:
Nitrogen fertilization promotes an increase in invasive grass biomass in desert (Allen et
al in press) and coastal sage scrub (Allen et al in preparation) that may provide additional
fuel for fire. Biomass analyses from these studies are being used to compare with
biomass levels found along N deposition gradients to determine the impact of nitrogen on
fire frequency. Invasive grasses occur under shrubs across the desert where soil N and
other resources are high. But in areas of high N deposition, the interspaces are also high
in N and there is sufficient dry grass fuel to carry a fire from shrub to shrub.
Laboratory N mineralization studies of desert soils show that initial soil N is correlated
with total N mineralization, and that N mineralization is correlated with invasive grass
cover. However, there is no relationship between soil N and cover of native forbs.
Fire model: Box Springs Mountain case study: The historic information coverage and the
spatial resolution are sufficient that the invasion of exotic grasses can be recognized in
the fire history. This provides a fire regime “fingerprint” that may be applied to other
regions without recourse to vegetation data so that we may make testable predictions
from fire regime change. The change in fire regime is found to be markedly
asynchronous between north- and south-facing slope aspects. The time span of the
information is a little short to determine whether and under what conditions coastal sage
may recover from grass invasion. There are indications that it can.
Fire Parcels: Miller’s GIS-Matlab data management system tracks spatio-temporal
“parcels.” These are contiguous elements of the landscape that share the same fire
history though time. A single time slice through these parcels reveals the familiar patch
mosaic. The parcels track the ancestry of patches in terms of the fires that have modified
them. We find this to be not only the most logical and efficient data management scheme
but also allows us to determine typical life histories for patches when traced through
time. In other words, we will be able to extract the relationship between the antecedent
fire-size distributions and an instantaneous “snapshot” of patch size distributions. Fires
may generate new large patches, leave new “island” patches, and reduce the size of preexisting patches. The dominant characteristic of patches, of course, is that they are
smaller than the fires and become smaller with time. We have the tools to search for
characteristic patch size histories; they are intimately related to questions of fire return
interval and fire overlap.
Hazard Function Modeling: Hazard functions are not a simple product of (or estimator
for) flammability as a function of vegetation age or fuel level. The probability of burning
depends upon two basic considerations: 1) the flammability of a vegetation patch once
fire reaches it; and 2) the probability that fire will reach a vegetation patch. The latter is a
function of the emergent fire-modulated mosaic pattern and not simply determinable
from fuel flammability characteristics. Fire begins with ignition in flammable vegetation
and progresses primarily downwind or upslope through flammable vegetation to reach
new areas. Because recent fires generate areas in which ignitions are less likely, fire
starts are rarer in these patches and fire does not carry readily to vegetation islands left in
the burn area or to vegetation at their down wind limits. For this reason, some model
runs produce hazard functions which have a weak harmonic fluctuation based on the age
of canopy closure.
Modeling of Persistent High-Risk Weather
Models indicate that, faced with persistent high winds and low humidity in the fire
season, natural vegetation mosaics adjust to reduce fire and patch size. A mosaic would
emerge that consists of small patches that are elongated perpendicular to the dominant
wind direction. This emergent mosaic depends upon vigorous flank fires. It is another
indicator that natural systems may self-organize to limit fire size.
Analogies with Fluid Mechanics: Just as theoretical physicists use highly stylized “forest
fire” models of turbulence and self-organized criticality, so the dimensionless-product
approach of fluid mechanics provides a promising means to analyze fire propagation
across a vegetation patch mosaic. The patch mosaic may usefully be compared with bed
roughness or the topography on a sandy sediment-water interface across which fluid flow
is locally impeded.
Remote Sensing: In activities 2 and 3 listed above, we found that the random noise and
strip noise have relatively large effect on the final result over different years. It is
important to find a way to control the noises.
According to our original analysis, 1998 may be quite different from other years. It may
be because it was a high precipitation year, so we will pay more attention to this year in
subsequent analysis.
Landscape modeling: 1. For the competition-colonization model, we showed that there
exist resource concentration thresholds, determined by exotic and native species’ traits,
which allow exotic species to successfully invade and displace native species. It revealed
that successful exotic plant species need not necessarily have superior competitive ability
relative to native species, but have higher seed production and survival rates. The model
is also explored to determine a limit to the number of species that stably coexist at
competitive equilibrium. We found that a constant resource input rate defines a limit to
that number and therefore, the resource quantity in the habitat maintains plant species
diversity.
2. For the mechanistic theory of plant competition for belowground resources. We show
that R*-value characterizations of competitive ability do well when plants directly
compete in strongly resource limited environments. A specific plant trait is found that
confers the variation in R* among all directly competing plants, that is the plant with
higher resource capture efficiency, keeps lower resource concentrations in its nonoverlapping depletion zone, accesses a lower number of overlapping zones within the
neighborhood of interactions, and is superior in direct plant competition; the best
competitor takes over an overlapping depletion zone by excluding all others at
competitive equilibrium by depleting the resource to the lowest as in its non-overlapping
depletion zone. Specifically, this trait would be a useful proxy measure for R* that does
not necessarily require the establishment of equilibrial field monocultures.
Training and Development
Four post-doctoral fellows and four graduate students are being trained (as listed under
personnel). They are being trained in fire and landscape modeling, DAYCENT and
HYDRYS modeling, and data collection related to N cycling and vegetation analysis. In
addition, one high school student (Thomas Bytnerowicz) worked in E. Allen’s lab, is now
an undergraduate at UCSD, and continues to work vacations. Three other undergraduates
(Ania Wrona, Cecilia Osorio, Jose Chavez) have worked on the project. Osorio is
working on a manuscript for publication.
Outreach Activities
Allen and Fenn will participate in the workshop “Nitrogen Deposition Effects on
Ecosystems in the US” Feb. 26-28, 2008, Baltimore, MD, sponsored by the USDA Forest
Service and US EPA. The objective of this workshop is to determine critical loads of N
deposition for impacts on ecosystems based on data sets from researchers across the US.
Data collected as part of this Biocomplexity grant will be presented.
UCR web page lists Minnich and Sadler as press contacts for wild fire questions.
Minnich handled numerous national and international press enquiries concerning ongoing wild fires in Southern California; Sadler responded to questions of computer
application, long-term fire behavior, and relationship to this biocomplexity project.
Sadler presented a lecture on Oct. 24, 2007 at the San Bernardino Co. Museum titled
“Chaparral Wildfire”
As part of the UCR Copernicus Project, Sadler met with Riverside Community College
faculty Carlos Tovares and Lori Keeler to plan demonstration classes in physical
geography that combine hands-on experiments with readings (Nov. 30, 2007). First class
will distribute a simplified version of Sadler’s fire mosaic software and readings about
controversial aspects of fire suppression in Southern California. Students will test ideas
with two model runs, changing parameters of their choice. Instructors will discuss results
and methods.
Publications:
Note: These were listed in the 2006 report as Accepted for publication
Breiner, J., Gimeno, B.S., and Fenn, M. 2007. Calculation of theoretical and empirical
nutrient N critical loads in the mixed-conifer ecosystems of southern California.
TheScientificWorldJOURNAL 7(S1), 198-205. DOI 10.1100/tsw.2007.65.
Wood, Y.A., Fenn, M., Meixner, T., Shouse, P.J., Breiner, J., Allen, E., and Wu, L. 2007.
Smog nitrogen and the rapid acidification of forest soil, San Bernardino Mountains,
southern California. TheScientificWorldJOURNAL 7(S1), 175-180. DOI
10.1100/tsw.2007.74.
New publications:
Bytnerowicz, A., Arbaugh, S. Schilling, W. Fraczek, D. Alexander, P. Dawson (2007)
Air pollution distribution patterns in the San Bernardino Mountains of southern
California: a 40-year perspective. TheScientificWorldJOURNAL 7(S1), 98–109. DOI
10.1100/tsw.2007.57.
Bytnerowicz, A., Arbaugh, S. Schilling, W. Fraczek, D. Alexander (2008) Ozone
distribution and phytotoxic potential in mixed conifer forest of the San Bernardino
mountains, southern California. Environmental Pollution (accepted).
Saito, H., J. Šimůnek, J. W. Hopmans, and A. Tuli, Numerical evaluation of the heat
pulse probe for simultaneous estimation of water fluxes and soil hydraulic and thermal
properties, Water Resour. Res., 43, W07408, doi:10.1029/2006WR005320, 2007.
Šimůnek, J., M. Th. van Genuchten, and M. Šejna, Development and applications of the
HYDRUS and STANMOD software packages, and related codes, Vadose Zone Journal,
Special Issue “Vadose Zone Modeling”, (submitted April 23 2007, revised July 20 2007,
accepted September 16 2007).
Šimůnek, J. and M. Th. van Genuchten, Modeling nonequilibrium flow and transport
with HYDRUS, Vadose Zone Journal, Special Issue “Vadose Zone Modeling”, (accepted
September 16 2007).
Sirulnik, A.G., E.B. Allen, T. Meixner, M.E. Fenn, M.F. Allen. 2007. Impacts of
anthropogenic N additions on nitrogen mineralization from plant litter in exotic annual
grasslands. Soil Biology and Biochemistry 39:24-32.
Cox, R.D. and E.B. Allen. 2008. Composition of soil seed banks in southern California
coastal sage scrub and adjacent exotic grassland. Plant Ecology, in press.
Egerton-Warburton, L.M., N.C. Johnson, and E.B. Allen. 2007. Mycorrhizal community
dynamics following nitrogen fertilization: a cross-site test in five grasslands. Ecological
Monographs 77:527-544.
Sirulnik, A.G., E.B. Allen, T. Meixner, M.E. Fenn, M.F. Allen. 2007. Changes in N
cycling and microbial N with elevated N in exotic grasslands of Southern California.
Applied Soil Ecology 36:1-9.