|
| |
Johns Hopkins University: Jonathan Patz (PI), Hugh Ellis (Co-PI)
Robert Gilman, Thaddeus Graczyk, Greg Gurri-Glass, Subhash Lele, Marc Parlange, Brad Sack, Clive Shiff
University of Maryland: Rita Colwell, Anwaral Huq, Estelle Russek-Cohen
Pennsylvania State University: Ann Fisher, James Shortle, James Lynch, Ray Najjar, Barry Evans, Egide Nizeyimana, Patricia Kocagil
University of Delaware: Laurence Kalkstein
Georgia Institute of Technology: Ann Bostrom
Science Communication Studies: Joan Aron
NOAA, National Climatic Data Center: David Easterling, Thomas Karl
NASA, Ctr. Health Applications of Aerospace Related Technologies: Byron Wood
University of South Florida: Joan Rose
US Department of Agriculture: Dana Focks
University of Texas, Houston: Jack Hayes
New Orleans Mosquito Control Board: Edgar Bordes
The Environmental Protection Agency
The objectives of this project are to reduce the uncertainty in risk assessment
of the impact of climate change on key public health endpoints, and to characterize and
communicate this information to support policy development and analysis. To achieve these
objectives we will: 1) Analyze key climate-sensitive diseases that have the potential to
expand or contract, intensify, or shift in spatial distribution; 2) Develop an
interdisciplinary, integrated approach that addresses the complexity of anticipated
disease responses to climate and ecological change; utilize quantitative methods to
further improve the risk assessment and, for one case study, illustrate the costs of
alternative options for reducing health risk; and 3) Adopt risk communication strategies
to insure that our findings can effectively inform policy makers on the public health
risks associated with climate change. Attainment of our objectives will result in an
expanded knowledge base upon which sound decisions can be made regarding climate change
policy.
Our methodological tools include: down-scaling climate analysis, geographic
information systems, remote sensing, hydrologic modeling, disease modeling and spatial
statistical analysis. Our approach is multidisciplinary and collaborative, involving
participation from twelve academic institutions. It involves case study analyses and
includes climate change simulation analysis, followed by hydrological and ecological
modeling, which in combination constitute the driving forces for the public health
outcomes. The selected climate-sensitive and region-based case studies are divided into
two project stages. Stage one beginning year one involves water-borne Crytptosporidiosis
and Cholera. Stage two beginning year two involves vector-borne diseases, specifically
Lyme and other tick-borne diseases, and Dengue and Dengue Hemorrhagic Fever. Risk
communication models will be used as part of our overall effort to insure that our
findings will effectively support policy decision making regarding the public health risks
associated with climate change.
EPA Grant Progress Reports
Abstract
Objectives of Research: 1)to assess the potential impact of climate change on
important regional public health endpoints; and 2) to appropriately characterize and
communicate this information to support policy development and analysis.
Approach: To achieve these objectives we will: 1) Analyze the altered risk of key
climate-sensitive diseases (cryptosporidiosis, cholera, dengue fever, lyme disease,
hantavirus) that have the potential to expand or contract, intensify, or shift in spatial
distribution; 2) Develop an interdisciplinary, integrated approach that addresses the
complexity of anticipated disease responses to climate and ecological change; utilize
quantitative methods to further improve the risk assessment and, for one case study,
illustrate the costs of alternative options for reducing health risk; and, 3) Adopt risk
communication strategies to insure that our findings can effectively inform policy makers
on the public health risks associated with climate change.
Preliminary Findings
Regarding the cryptosporidiosis-related work, we found a temporal and spatial
correlation between extreme rainfall events and water-borne disease outbreaks for the
United States based on analysis of EPA time series data from 1972-1994. We have
successfully calibrated a rainfall-runoff model (see figure) for a watershed in Lancaster
County (a critical step in assessing increased cryptosporidiosis risk from altered
precipitation patterns producing more flooding of agricultural disease reservoirs for
which we have developed a GIS). A survey of veterinarian records showed upwards of 95% of
agricultural operations to be positive for cryptosporidiosis. The costs from an increase
in the probability of cryptosporidiosis were analyzed using data from Lancaster County (in
the Susquehanna River Basin) as a case study. With respect to the cholera analyses,
cholera-positive algae samples have been detected off the coast of Aeraquipa, Peru where
human cholera cases occurred in 1997. The dengue transmission simulation model has been
set up for Brownsville Texas, New Orleans and two sites in Puerto Rico. At the Brownsville
site, the dengue simulation model has closely predicted disease outbreaks using data from
1979 onward. Using Landsat satellite images to determine habitat characteristics favorable
for deer mice (i.e., soil moisture and vegetation) Hantavirus case households have been
successfully hindmost for the 1993 Four Corners outbreak. Based on this pilot study, field
validation is currently being conducted.
Significance of Findings
The occurrence of certain water-borne diseases has long been recognized as
highly seasonal. Our preliminary findings show a relationship between heavy rainfall and
disease incidence. Moreover, we are developing the capability to identify high risk
watersheds by observing clustering in historical outbreaks. Retrospective analyses
involving mosquito-borne dengue fever are similarly beginning to show a relationship
between disease incidence and temperature. We now believe it possible to predict high risk
conditions for Hantavirus on the basis of ecological characteristics, moreover these
predisposing conditions can be detected well in advance thus providing a potentially very
useful early warning capability.
Next Steps
We now are positioned to develop down-scaled climate scenarios that will be used
to drive the regional scale modeling efforts reported above. For example, down-scaled
climate scenarios for the Susquehanna River Basin will be used as input to the calibrated
rainfall runoff model, which in turn will produce altered flooding patterns. These
patterns will then be integrated with our existing GIS-based data on agricultural disease
reservoirs, finally yielding estimates of altered risk for cryptosporidiosis. The
historical analyses will be expanded and refined for water-borne outbreaks (1940 onward)
and dengue fever (1920 onward). We will continue to analyze human epidemiologic data and
historical algae samples for cholera in South and Central America, and the Chesapeake Bay
as well as integrating remotely sensed data obtained from the SeaWifs satellite. An
ecological model will be linked to a Lyme disease transmission model for the Mid-Atlantic
region. We are continuing the development of the interactive web site/data
repository/model analysis suite.
Hantavirus
Both stages of the Hantavirus/El Niño analysis have been completed and first stage
submitted for publication. We were able to accurately predict high risk areas for
Hantavirus Pulmonary Syndrome based on satellite generated risk maps
of land cover. Predicted risk was also concordant with vegetative growth determined by the
normalized difference vegetative index (NDVI), supporting the hypothesis that heavy
rainfall from El Niño in 1992 were associated with higher rodent populations that
triggered the Hantavirus outbreak in 1993.
Dengue Fever
The dengue fever transmission model (DENSiM) has been improved to better
account for water container temperature and water height, important to larval mosquito
development rates (Chen et al. 1998). Full model parameterization has been completed at
all study sites (Brownsville, New Orleans, and Puerto Rico), and model runs are being made
using incremental temperature changes to assess potential thresholds key to any future
climate analysis. Matrices of the downscaled climate runs have recently been completed for
input into the DENSiM model.
Lyme disease
We have classified landcover for the mid-Atlantic region by use of remote
satellite sensing. We have tuned the USDA Lyme disease simulation model to better account
for humidity and temperature and we will apply our recently developed spatial modeling of
tick abundance using generalized linear mixed models for more accurate risk prediction,
given a GIS of landcover.
Cryptosporidiosis
Much progress has been made in all aspects of this highly integrated
project. Climate downscaling from the Max Planck GCM is complete for the mid-Atlantic
region, and we have successfully calibrated a hydrological run-off model for a
watershed in Lancaster County. The historical analysis of extreme precipitation and
water-borne disease outbreaks in the US since 1940 is almost complete. A positive temporal
and spatial correlation was found between heavy precipitation and disease occurrence, with
strongest correlation in the Fall season. In preparation for modeling cryptosporidiosis
risk under climate change, livestock operations within the 100 year floodplain of a major
creek in Lancaster County were sampled for cryptosporidium oocysts; in field manure more
than half of the samples tested positive. We have now obtained drinking water treatment
facility data and will finalize the full risk model (from rainfall and run-off, to
cryptosporidium oocyst loads at the drinking water facilities). Run-off modeling to
determine changes in salinity of the Chesapeake Bay is progressing. Economic analysis for
baseline costs of cryptosporidiosis cases has been completed.
Cholera
In Peru, 11 patients have been identified as probably the first indicator
cases, occurring from late October, 1990 to January, 1991. In addition to SST satellite
data from 1991, we have obtained SST data for 1997 and SeaWifs data for algal blooms in
1998. In the Chesapeake Bay, preliminary analysis shows a strong correlation between
warmer water temperature and the presence of Vibrio cholerae.
Risk Communication and Integrated Assessment
An update of the multiple components under this grant can be found on a
dedicated website (www.jhu.edu/~climate) -featured recently on CNN. An integrated modeling
framework under development, will incorporate the disease-specific climate models now
being finalized. Also a survey of water resource managers has been conducted to assess
their level of public health knowledge and the potential risks posed by climate
variability.
Future Activities
Hantavirus: satellite images from non El Niño "control years" are
being finalized prior to submission of the field validation manuscript.
Dengue Fever: Matrices of downscaled climate data for the Southeast will be used integrate
the dengue transmission model with climate change scenarios.
Lyme Disease: We will run the Lyme-Sim model and include the vegetative projections under
climate change scenarios to assess differences in key landscape features, such as forest
edge, that are risk factors for Lyme disease.
Cryptosporidiosis: Point process model of historical water-borne disease outbreaks will be
completed. The last component of the cryptosporidiosis risk model that links rainfall and
run-off to reported oocyst counts in Lancaster County will be completed. More
comprehensive cost/benefit analysis is being planned.
Cholera: Temporal & spatial analysis of Peruvian cholera cases, SST, and
plankton samples positive for cholera bacteria in 1991 and 1997 is soon to begin.
Integrated Assessment and Risk Communication: Analysis of surveys to water resource
managers will be completed. While all project-specific analyses are being completed,
overall integrated assessment will continue to incorporate completed models into an IA
framework, and if feasible, a decision analysis tool.
ABSTRACT
Objectives and Approach:
The objectives of this research are to assess the potential impact of climate change on
important regional public health endpoints, including vector-borne diseases (Hantavirus,
Dengue and Lyme disease) and water-borne diseases (Cryptosporidiosis and Cholera) and to
appropriately characterize and communicate this information to support policy development
and analysis. Our approach varies by disease outcome being studied. For example, satellite
remote sensing is being used to assess climate and landuse contribution to risk for
hantavirus, Lyme disease, and sea surface temperatures in the case of cholera.
Process-based models of disease risk and general circulation models (GCMs) are utilized
for dengue fever, while multiple methods including GIS, time series analysis, hydrological
modeling and downscaled GCMs are employed in the cryptosporidiosis project.
Preliminary Findings
Since the analyses of the five diseases under investigation are at varying levels of
completion results are listed separately at this time:
Hantavirus: The Hantavirus/El Niño analysis showed that high risk areas for
Hantavirus Pulmonary Syndrome can be predicted based on satellite generated risk maps of
land cover over six months in advance. Predicted risk paralleled vegetative growth,
supporting the hypothesis that heavy rainfall from El Niño in 1992 were associated with
higher rodent populations that triggered the Hantavirus outbreak in 1993. Satellite images
from 1995, a non El Niño "control" year, showed low risk in the region, whereas
the images from the 1998 strong El Niño again showed high risk areas (see figure).
Trapping mice in the field (collectors blinded to risk category), validated these
satellite generated risk maps with mouse populations directly related to risk level.
Dengue: The dengue fever transmission model (DENSiM) has been parameterized for
the study sites (Brownsville, New Orleans, and Puerto Rico), and model runs have been
completed for Brownsville, Texas using the Hadley Center HADCM2 and VEMAP interpolated
climate projections for the years 2030, 2060 and 2100. Transmission risk shows very high
sensitivity to relative humidity and temperature (e.g., excessive dryness limits mosquito
survival). Also, while warm temperatures generally increase transmission dynamics, the
extreme high temperatures predicted by HADCM2 for Texas diminish the risk of dengue in
this location. It is expected that the study sites on the island of Puerto Rico will show
opposite trends, however, model runs are pending CDC human case data.
Lyme disease: Tick survey data has been analyzed for the mid-Atlantic region and
validates landuse predictions for Lyme disease risk. Extent of forest edge remains the
strongest predictor for tick abundance. We have classified landcover for the mid-Atlantic
region by use of satellite Landsat imaging. We have improved a USDA Lyme disease
simulation model to better account for humidity and temperature and we will apply spatial
modeling of tick abundance using generalized linear mixed models for more accurate risk
prediction, given a GIS of landcover.
Cryptosporidiosis: The historical analysis of extreme precipitation and
water-borne disease outbreaks in the US from 1946 to 1994 shows spatial clustering of
outbreaks indicating high-risk geographical regions by water basin. The fall season
contained the highest proportion (40%) of surface water-related outbreaks preceded by
heavy rainfall events; winter, spring and summer percentages were 25%, 35% and 32%
respectively. Climate downscaling from the Max Planck GCM is complete for the mid-Atlantic
region, and we have successfully calibrated a hydrological run-off model for a watershed
in Lancaster County. Random sampling of 50 farms within the 100 year floodplain of a major
creek in Lancaster County and over 60% tested positive for cryptosporidium oocysts in
field manure samples. Economic analysis for baseline costs of cryptosporidiosis cases has
been completed.
Cholera: In the Chesapeake Bay, preliminary analysis shows a strong correlation
between warmer water temperature and the presence of Vibrio cholerae. In 1998 a
cholera epidemic occurred in Lima, with more than 1,000 reported cases. Surveillance of
sewage water for cholera was strongly associated with ambient temperature and peak number
of cases lagged 3 weeks behind a peak in ambient temperature. Regarding the 1991 epidemic
in Peru, 11 patients have been identified as probably the first indicator cases, occurring
from late October, 1990 to January, 1991.
Significance of Preliminary Findings
Many of the diseases under investigation are highly seasonal, however, the contribution
of weather factors or habitat is not fully understood. For hantavirus, the ability to use
satellite images of habitat features to predict disease risk far in advance is a valuable
tool for disease prevention. For Lyme disease, preliminary results confirm the importance
of landuse features in determining tick abundance. The dengue fever analysis demonstrates
how sensitive transmission risk is to specific weather parameters, and future risk
projections will be highly site specific. The historical analysis of water-borne diseases
shows preliminarily that precipitation likely contributes to the occurrence of these
outbreaks. Hydrological modeling of future risks, however, is necessary considering the
balancing between heavier rainfall projections versus warmer temperature (thus, more
evapotranspiration). For cholera, while the role of algal blooms preceding the 1991
epidemic remains undetermined, we now show evidence of a link between ambient
temperatures, Vibrio cholera in sewage and the marine environment, and cholera
outbreaks.
Next Steps
We plan to finalize all incomplete analyses. For example, matrices of downscaled
climate data for the Southeast (per Dr. Linda Mearns) will be applied to the Puerto Rico
study sites, and human immunity data will be obtained from the CDC dengue lab in San Juan.
For Cryptosporidiosis, precipitation/outbreak analysis is being conducted at the
watershed level to better determine contribution of rainfall in predicting risk of
outbreaks. The last component of the cryptosporidiosis risk model that links rainfall and
run-off to reported oocyst counts at treatment facilities in Lancaster County can now be
completed with recently acquired data. For cholera, we have obtained sea surface data for
1997 and SeaWifs satellite data for algal blooms in 1998. Also, water samples of algae are
still pending from our Peruvian collaborators. While all project-specific analyses
are being completed, the overall integrated assessment will incorporate the completed
models into an integrated framework communicated though our website.
Hantavirus
The Hantavirus/El Niño analysis showed that high risk areas for Hantavirus
Pulmonary Syndrome can be predicted based on satellite generated risk maps of land cover
over six months in advance (Glass et al. 2000). Predicted risk paralleled vegetative
growth, supporting the hypothesis that heavy rainfall from El Niño in 1992 were
associated with higher rodent populations that triggered the Hantavirus outbreak in 1993.
Satellite images from 1995, a non El Niño "control" year, showed low risk in
the region, whereas the images from the 1998 strong El Niño again showed high risk areas.
Trapping mice in the field (collectors blinded to risk category), validated these
satellite generated risk maps with mouse populations directly related to risk level. Our
methods, developed in partnership with CDC and the Indian Health Service are already being
implemented for disease prevention by the US DHHS.
Dengue
Model runs have been completed for Brownsville, Texas using the Hadley Center
HADCM2 and VEMAP interpolated climate projections for the years 2030, 2060 and 2100.
Dengue transmission simulation shows very high sensitivity to relative humidity and
temperature (e.g., excessive dryness limits mosquito survival). Also, while warm
temperatures generally increase transmission dynamics, the extreme high temperatures
predicted by HADCM2 for Texas diminish the risk of dengue in this location. It is expected
that the study sites on the island of Puerto Rico may show different results.
Lyme disease
Tick survey data has been analyzed for 15 locales the mid-Atlantic region.
Multi-Resolution Land Characteristics (MRLC) data provide data on landcover based on
satellite imagery, and along with ten year retrospective climate values (1984-93) were
entered into the Lyme simulation model to determine baseline Lyme disease risk via the
vertebrate hosts involved in maintaining its life cycle, white-tailed deer, and
white-footed mice. With the exception of a few sites (which were only lightly forested),
deer carriage of the Ixodes tick approaches 100% and mouse carriage of the Lyme spirochete
varied between 60-80% for the baseline model runs. Under climate change scenarios from
GCMs generated from GFDL and Max Plank transient 2 (2020) and transient 3 (2050), the risk
of infection increases for mice in most locales, and tick carriage remains high for deer.
Transient 2 models yielded slightly higher infection rates than transient 3 models.
Cryptosporidiosis
The historical analysis of extreme precipitation and water-borne disease outbreaks
by watershed in the US from 1948 to 1994 shows spatial clustering of outbreaks indicating
high-risk geographical regions by water basin. The fall season contained the highest
proportion (40%) of surface water-related outbreaks preceded by heavy rainfall events;
winter, spring and summer percentages were 25%, 35% and 32% respectively. For surface
water-identified outbreaks, there is a statistically significant association between
occurrence of an extreme precipitation event and the disease outbreaks. Climate
downscaling from the Max Planck GCM is complete for the mid-Atlantic region (Easterling
1999), and we have successfully calibrated a hydrological run-off model for a watershed in
Lancaster County (Szilagyi &Parlange 1999; Najjar 1999). Random sampling of 50 farms
within the 100-year floodplain of a major creek in Lancaster County found over 60% tested
positive for cryptosporidium oocysts in field manure samples (Graczyk at al. 2000), and
calculations for the amount of infected manure in the region are competed. For 76
watersheds in Pennsylvania, we have meteorological, geological, agricultural, and
microbiological (Giardia cysts and Cryptosporidium oocysts) variables entered into an
Arc/Info database. Economic analysis for baseline costs of cryptosporidiosis cases has
been completed (Kocagil et al. 1998).
Cholera
In the Chesapeake Bay, preliminary analysis shows a strong correlation between
warmer water temperature and the presence of Vibrio cholerae. We have also examined
long-term (~ 50 year) trends in salinity in the Chesapeake Bay. While salinity is a
function of stream flow, a strong trend in the residual has been observed, which may be a
signal of sea level rise. At our Peruvian study site, in 1998 a cholera epidemic occurred
in Lima, with more than 1,000 reported cases. Surveillance of sewage water for cholera was
strongly associated with ambient temperature and peak number of cases lagged 3 weeks
behind a peak in ambient temperature. Also, time-series analysis for Lima, Peru from
1993-98 showed that during the warm winter temperatures of the 1997/8 El Niño, the number
of children hospitalized for severe diarrhea was 2-fold over expected numbers (Checkley et
al 2000). Regarding the 1991 epidemic in Peru, 11 patients have been identified as
probably the first indicator cases, occurring from late October, 1990 to January, 1991.
For a list of papers and presentations resulting from the grant,
please see the bibliographic database or news
and events.
Waterborne Cryptosporidiosis Case Studies
The Parasitology Laboratory at the School of Hygiene and Public Health of the Johns
Hopkins University has been accredited for analyses of human pathogens since 1986. The
project activities related to waterborne cryptosporidiosis will be directed by Dr.
Thaddeus Graczyk and Dr. Clive Shiff. Dr. Shiff has conducted several national programs
focused on control of human pathogen transmission. Project activities related to water
processing and recovery of Cryptosporidium oocysts will be carried out in a specially
designed room isolated from the rest of our lab/office complex. The room is approved for
the work with human pathogen (JHMI Safety Office Approval B9003060206), is properly
labeled, and inspected every 3 months by a Safety and Environmental Health Officer. The
room was equipped by Memtec America Corp. (Baltimore, MD) and includes two additional
systems (Filterite 10-#m-pore-size yarn-wound cartridge filter, and an activated carbon
filter) to process (or store) water for any experiments with water-transmitted pathogens.
We have a security lock system, so that only project collaborators (Graczyk and Shiff) or
personnel authorized by them have access to this room. Analysis of water recovered
material will be performed in other parts of our laboratory complex. This part of the
laboratory complex has an isolated dark compartment with fluorescent microscope (Olympus
BH2 BH50/BH40) for immunofluorescent analysis. Our lab/office complex is maintained by a
North Star phone system, which allows for intercom contact without picking up a handset.
We are familiar with the method for recovery of waterborne oocysts by the American Society
for Testing and Materials (ASTM) - 1, the Standard Methods for the Examination of Water
and Wastewater (SMEWW) - 2, and the Alternate Method - 3 which utilizes a yarn-wound
cartridge filter, or a membrane filter. Only these standard methods will be used for
recovery and the identification of Cryptosporidium oocysts. We have considerable
experience in the identification of Cryptosporidium oocysts from water recovered material
using immunofluorescent antibody (IFA) recommended by the above mentioned methods used by
us before (Graczyk et al. 1996 a,b,c). The material recovered from raw water and
positively diagnosed for Cryptosporidium oocysts will be filed. The materials in the file
will be accessible upon request. The equipment present in our laboratory/office complex is
fully capable of carrying out the activities outlined in the research proposal. In case of
temporary failure of any equipment in our lab (which we do not anticipate), we have full
access to the Common Facility Equipment of the Department of Molecular Microbiology and
Immunology, School of Hygiene and Public Health, Johns Hopkins University.
Cholera Case Studies
Three Case Study sites are included in the proposed work: 1) Chesapeake Bay; 2)
Gulf of Mexico; and 3) the Peruvian Coast. Data on the reported cases of cholera occurring
in the USA, Peru and Mexico will be obtained from CDC, USA, Public Health Department of
Peru and from the Mexican Health Department (Secretaria de Salud) respectively. The data
maintained at the above mentioned institutes represents the best sources available and are
also reliable, having been used in numerous studies worldwide. Environmental sampling for
the detection of V. cholerae in water and plankton, and collection of plankton samples
will be conducted following existing methods (Huq et al. 1990, Huq et al 1994, Islam et al
1995). For direct detection, the direct-fluorescent antibody (DFA) test (Hasan et al.
1999) and the standard culture method where TCBS (thiosulphate citrate bile-sall-sucrose
agar) plates with enrichment in Alkaline Peptone water will be used. All the tests will be
performed in duplicate and representative of the samples will be retested in the main
laboratory of Dr. Colwell at the Center for Marine Biotechnology (COMB) for quality check.
Samples and specimen will be hand carried from the field to the laboratories, in properly
secured containers under optimal conditions. All the samples will be saved and stored
under appropriate condition for reference check as necessary.
Field sampling will be conducted by masters level scientists or graduate students under
direct-supervision by our investigators in the project or a person with experience
equivalent to the investigator to be assigned by the investigator. Prior to commencing the
work a standardized protocol will be prepared and if needed, a 1-2 day workshop will be
conducted so that all the collaborating laboratories follow identical methods.
Raw data will be entered in the computer at the field laboratory and subsequently
forwarded to the central laboratory at the Center for Marine Biotechnology of the
University of Maryland Biotechnology Institute for final analyses.
Climate Modeling, Hydrology
Projections of changes in water-borne Cryptosporidiosis and Cholera depend
primarily on potential changes in flood frequency and intensity (as moderated by
mitigating actions taken by water suppliers and individual water users and
recreationists). Thus, the size of the population exposed, and the extent of their
exposure, are key inputs for projecting the number of illness cases. The steps in the
analysis will rely primarily on existing data.
1) Work at Penn State has been validating down-scaled results for projecting changes in
temperature and precipitation within the Susquehanna River Basin (SRB) using GENESIS-2, a
modified version of the NCAR Community Climate Model. The climate data used (e.g., gridded
sea level pressure and 200 mb data from the National Meteorological Center, temperature
and precipitation data from SRB stations) are publicly available and already have met
standard tests for precision, accuracy, representativeness, completeness, and
comparability. For regions as small as 3 x 3 degrees longitude/latitude, tests show that
GENESIS-2 provides much better regional projections than the General Circulation Models
commonly used in climate change research.
2) Another Penn State project is refining the hydrologic predictions of what will happen
to increased precipitation in the SRB. This research accounts for variables such as soil
moisture, vegetation, snow melt and surface runoff, and uses existing data (e.g., from
USGS or Soil Conservation Service) on elevation, temperature, area, riparian area,
latitude, and precipitation to project stream flow and evapotranspiration. Much of the
data for the hydrologic analysis already has been entered in Geographic Information
Systems (GIS), with known characteristics related to precision, accuracy,
representativeness, completeness, and comparability. The data are being compiled in GIS
format so that stream discharge from small tributary watersheds can be isolated for
analysis; tributary watersheds easily can be aggregated for larger geographic regions as
needed.
3) Output from (2) will be combined with (existing) GIS data on land use and water uses,
topography, and infrastructure to project changes in flood frequency and intensity, and
the impacts on various types and sizes of municipal and private water suppliers. There are
a very large number of water suppliers in the SRB, most of which are small. Data will be
compiled from existing sources on as many of these as possible; much of this is available
(in GIS format) through the Pennsylvania Department of Environmental Protection (PA DEP).
Variables include number of customers, amount of water used by largest customer, frequency
of flooding, actions taken to prevent contamination, etc. After preliminary analysis, a
sampling strategy will be developed and documented, so that more detailed analysis can be
extrapolated for the entire SRB.
4) If needed, additional information will be gathered for a subset of the sample
identified in (3), regarding how they cope with flooding. Variables include how long
contaminated water might reach their customers (so that dose and number of people exposed
can be estimated), costs of cleanup before water service can be resumed, investment costs
for preventing future flooding, duration of beach closings, etc. (so that
cost-effectiveness of alternative prevention/treatment strategies can be compared with
illness costs). This subset will be selected so that results could be extrapolated for the
entire region. A data collection protocol will be developed (if available data are
inadequate), along with a verification strategy consistent with standard procedures for
collecting socioeconomic data. The data will be entered into a computer file for analysis
using statistical packages such as SPSS and SAS.
Data sources will be documented and statistical results will be reported using accepted
formats (so that, if desired, others could reproduce the analysis).
Risk Characterization and Communication
Four small data sets will be collected for the Risk Characterization and
Communication component of the project: (1) interviews with water quality experts, public
health officers, and other experts on cryptosporidium and drinking water quality at CDC;
(2) email surveys of drinking water managers, to be contacted via the AWWA, (3) written
(fax) surveys of key congressional and CRS staff, and (4) telephone or, if possible
face-to-face interviews with former members of congress (very small convenience samples
from both the House and Senate).
The interviews will be ethnographic in nature and similar to those interviews collected by
cognitive anthropologists and risk communication researchers in previous research (e.g.,
Kempton, 1991; Bostrom, Fischhoff and Morgan, 1992; Maharik and Fischhoff, 1992). The
primary purpose of the both the interviews and surveys will be to provide some sense of
how experts and policy makers think about the domain, what they think the key issues are
and how they conceptualize the associated hazardous processes. The first set of interview
data will be used to guide survey question formulation and, subsequently, communication
design.
Standard survey design and interviewing techniques will be used (see, e.g., Fowler and
Mangione, 1990, Sudman and Bradburn, 1982; Tanur, 1992), to minimize interviewer error and
bias. All interviews will be based on standardized protocols, to be reviewed by other
members of the research team and made available on request. Use of standardized protocols
enables comparability across interviews and helps assure complete coverage of the topic.
Interviews will be taped, whenever permitted by the interviewees, to permit corroboration
of notes taken by the investigator. A small random sample of the tapes will be abstracted
by a research assistant for comparison, to check the reliability of the investigator's
notes. Tapes will be archived either at Johns Hopkins or at Georgia Institute of
Technology. All tapes will be labeled with information that preserves the interviewee's
confidentiality, but enables identification of the specific interview by date, time, and
the expertise or authority of the interviewee.
Data entry for all survey data will be checked for accuracy, by having a second person
enter at least some of the data for comparison. If data entry is not at 95% or above
accuracy, the data will be re-entered. Copies of all surveys will be provided for project
management on request, with identifying information removed to preserve confidentiality.
For the collection of the first set of interviews from CDC, an exhaustive search will be
made to identify all qualified interviewees (as specified above) at CDC's headquarters in
Atlanta. All of these will be contacted and asked to participate, either in interviews or
surveys. Responses will be recorded, to allow assessment of response rates. If more than
two dozen are available for interviews, a subset will be chosen randomly (because of time
and resource constraints on the investigator). Repeated contacts will be made until at
least 50% of those identified are contacted.
The email surveys of water quality managers will also be designed using best available
practice, to encourage high response rates and elicit as much information as possible from
the participants. The survey instrument will be reviewed by multiple investigators and
pilot tested before use. AWWA will be contacted for their help in compiling an appropriate
email distribution list for the survey.
CRS and congressional staff will be surveyed in the third data collection effort. CRS
staff will be asked to review the list of congressional staff identified, to assure as
complete coverage as possible of relevant staff (e.g., staff working on water quality or
global climate change issues). All surveys will be faxed at least twice, to increase
response rates. Within budget constraints, as much of an effort as possible will be made
to obtain high response rates. This survey instrument will build on the findings from the
first two data collection efforts.
The last set of interviews will be of a convenience sample of former members of congress,
with every attempt made to contact some with a strong interest in the research topic, and
some without, to provide some measure of how strongly similar or dissimilar their
perceptions are, and how they compare with those of staff and experts.
See above for specification of handling, identification and preservation. Once interview
abstracts and notes have been checked by the investigator (they will be entered on the
computer at the time of the interview) and survey data entered in spreadsheets, both will
be kept archived on diskette, and copies made available to the project management.
Interview notes will be kept in either Word or WordPerfect, survey data in Excel or in
ASCII, comma-delimited format.
Standard quality assurance and quality control procedures will be followed. However,
transcriptions of the tapes will not be made, due to the high costs of doing so. Notes
from interviews will, however, be checked for reliability using the tapes, as described
above.
For the purposes of the project, most of the data will be reduced using comparisons with
expert knowledge structures, though those comparisons will be less formal than equivalent
comparisons carried out in previous research (e.g., Bostrom et al. 1992). Survey data will
be summarized using standard statistical approaches. Frequency distributions and measures
of central tendency will be reported, as will comparative statistics (e.g., of water
quality managers' responses, congressional staff, and other, previous samples who have
received the same or similar questions). Standard analysis of variance techniques (and
general linear models) will be used for comparisons, with appropriate use of other
statistical analyses as warranted by sample sizes and multiple comparisons, for example.
Graphical representation of data will be used whenever possible, guided by available
literature on appropriate representations (e.g., Henry, 1995).
| |
|
