Content uploaded by Christine M Laney
Author content
All content in this area was uploaded by Christine M Laney on Feb 19, 2014
Content may be subject to copyright.
Agricultural
Research
Service
Technical
Bulletin
Number 1931
September 2013
Long-Term Trends in
Ecological Systems:
A Basis for Understanding
Responses to Global Change
United States Department of Agriculture
V
Contents
Contributors ...........................................................................................................................................VIII
Technical Consultants ................................................................................................................................X
Introduction to Cross-Site Comparisons and History and Organization of the
EcoTrends Project
Chapter 1: Long-Term Trends in Ecological Systems: An Introduction to Cross-Site Comparisons
and Relevance to Global Change Studies .................................................................................1
Chapter 2: History and Organization of the EcoTrends Project ...............................................................21
Cross-Site Comparisons of Ecological Responses to Global Change Drivers
Chapter 3: Cross-Site Comparisons of Ecological Responses to Climate and Climate-Related
Drivers ....................................................................................................................................28
Chapter 4: Cross-Site Comparisons of State-Change Dynamics ..............................................................36
Chapter 5: Patterns of Net Primary Production Across Sites ...................................................................42
Chapter 6: Cross-Site Comparisons of Precipitation and Surface Water Chemistry ................................46
Chapter 7: Cross-Site Comparisons of Ecological Responses to Long-Term Nitrogen Fertilization ......51
Chapter 8: Long-Term Trends in Human Population Growth and Economy Across Sites ......................54
Chapter 9: Disturbance Regimes and Ecological Responses Across Sites ...............................................58
Chapter 10: Cross-Site Studies “By Design”: Experiments and Observations That Provide New
Insights ..................................................................................................................................72
Long-Term Trends in Global Change Drivers and Responses at Site and Continental
Scales
Chapter 11: Long-Term Trends in Climate and Climate-Related Drivers ................................................81
Chapter 12: Long-Term Trends in Precipitation and Surface Water Chemistry .....................................115
Chapter 13: Long-Term Trends in Human Demography and Economy Across Sites ............................162
Chapter 14: Long-Term Trends in Production, Abundance, and Richness of Plants and Animals ........191
Chapter 15: Management and Policy Implications of Cross- and Within-Site Long-Term Studies .......206
Chapter 16: Recommendations for Data Accessibility ...........................................................................216
Chapter 17: Long-Term Research Across Sites, Ecosystems, and Disciplines: Synthesis and
Research Needs ...................................................................................................................226
Appendices
Appendix 1: Site Descriptions ................................................................................................................234
Appendix 2: Average (Standard Error) Maximum, Mean, and Minimum Air Temperature
and Annual Precipitation at Each Site ................................................................................312
VI
Appendix 3: Average (Standard Error) Ice Duration, Sea Level, Streamow, Water Clarity, and Water
Temperature for Sites With Data ....................................................................................... 314
Appendix 4: Regression Coefcients and R2 Values for Nine Climatic Variables for Which Linear
Regression Against Time Is Signicant (p < 0.05) ........................................................... 316
Appendix 5: Annual Average (Standard Error) Nitrogen (as Nitrate) From Various Sources at Sites
With Data ........................................................................................................................... 319
Appendix 6: Regression Coefcients and R2 Values for Nitrogen (as Nitrate) From Various Sources
for Which Linear Regression Against Time Is Signicant (p < 0.05) .............................. 321
Appendix 7: Annual Average (Standard Error) Nitrogen (as Ammonium) From Various Sources at
Sites With Data ................................................................................................................. 323
Appendix 8: Regression Coefcients and R2 Values for Nitrogen (as Ammonium) From Various
Sources for Which Linear Regression Against Time Is Signicant (p < 0.05) ................. 325
Appendix 9: Annual Average (Standard Error) Sulfur (as Sulfate) From Various Sources at Sites
With Data .......................................................................................................................... 326
Appendix 10: Regression Coefcients and R2 Values for Sulfur (Sulfate) From Various Sources for
Which Linear Regression Against Time Is Signicant (p < 0.05) .................................. 328
Appendix 11: Annual Average (Standard Error) Chloride From Various Sources at Sites With Data .. 330
Appendix 12: Regression Coefcients and R2 Values for Chloride From Various Sources for Which
Linear Regression Against Time Is Signicant (p < 0.05) .............................................. 332
Appendix 13: Annual Average (Standard Error) Calcium From Various Sources at Sites With Data ... 334
Appendix 14: Regression Coefcients and R2 Values for Calcium From Various Sources for Which
Linear Regression Against Time Is Signicant (p < 0.05) .............................................. 336
Appendix 15: Human Population and Economy Variables in 2000 for the Focal County of Each Site,
as Grouped by Ecosystem Type ...................................................................................... 338
Appendix 16: Annual Average (Standard Error) Aboveground Net Primary Production (ANPP) at
Sites With Data ................................................................................................................ 341
Appendix 17: Other Measures of Average (Standard Error) Terrestrial Production at Sites With Data..343
Appendix 18: Average (Standard Error) Aquatic Production at Sites With Data .................................. 344
Appendix 19: Average (Standard Error) Biomass of Primary Producers (Plants, Algae) for Sites With
Data ................................................................................................................................. 345
Appendix 20: Average (Standard Error) Plant Species Richness for Sites With Data ........................... 347
Appendix 21: Average (Standard Error) Animal Abundance for Sites With Data ................................. 349
Appendix 22: Average (Standard Error) Animal Species Richness for Sites With Data ....................... 352
Appendix 23: Regression Coefcients and R2 Values for Plant and Animal Variables for Which Linear
Regression of Each Variable Against Time Is Signicant (p < 0.05) and the Trend
Appears Linear ................................................................................................................ 353
Appendix 24: Lead Principal Investigator(s) (PI), Information Managers (IM), and Administrative
Program of the LTER Programs ...................................................................................... 355
Appendix 25: Researchers Involved in the EcoTrends Project at Non-LTER Sites .............................. 359
VII
Appendix 26: List of Stations and Length of Record for Each Climate Variable by Site ...................... 362
Appendix 27: List of Stations and Length of Record for Each Precipitation or Surface Water
Chemistry Variable by Site ............................................................................................. 367
Appendix 28: List of Stations and Length of Record for Each Plant and Animal Variable by Site, as
Grouped by Ecosystem Type .......................................................................................... 371
Index ........................................................................................................................................................... i
216
Long-Term Trends in Ecological Systems:
Chapter 16
Recommendations for Data
Accessibility
C.M. Laney, K.S. Baker, D.P.C. Peters, and
K.W. Ramsey
The EcoTrends Project was established to aid
researchers and others interested in synthetic studies of
long-term, continental-scale and national-level patterns
in environmental drivers and ecological responses.
Hundreds of standardized, documented datasets from
many sites and scientic elds were synthesized to
meet this goal. Generating comparative data at many
sites across several organizational networks and
nding novel solutions to technical, organizational,
and communication challenges required ongoing
collaborative work with all project participants,
including researchers and information managers.
The lessons learned from this collaborative effort
contributed to our understanding of contemporary
ecological information management (that is, the
management of digital ecological data via multifaceted,
interdependent arrangements and systems). Drawing
on these lessons learned by EcoTrends participants—
project leaders, researchers, and network- and
site-level information managers—we present
10 recommendations for site-level information
management and for future synthesis projects. These
recommendations for supporting synthesis projects are
related to three broad categories:
• Data management and products
• Project design
• Information environments
Challenges
The collection, management, and sharing of ecological
data are rapidly changing because of escalating
advances in technology and in knowledge-sharing.
Advances in automated, continuous collection of data
from sensors are increasing the number of methods
available to observe and measure the environment.
These technologies and methods can generate data
that span a wide range of spatial and temporal scales
(see Porter et al. 2005, Collins et al. 2006, Benson
et al. 2010 as examples). Management of data has
evolved along with statistical software and database
technologies. For example, quality checking of data for
errors in values and formats was previously conducted
manually by researchers or technicians but is now
often performed using automated statistical software
(for example, Michener and Brunt 2000). Data that
were once stored in simple spreadsheets are now often
stored in more complex relational databases. The
sharing of data and knowledge has increased as more
research sites post links to their data on web pages
or make the data available via new web services. To
aid in the sharing of data, data practices, policies, and
documentation standards have been and continue to be
developed among research communities (for example,
Karasti and Baker 2008, Porter 2010, Vanderbilt et al.
2010).
Large synthetic studies of diverse ecological data have
been greatly facilitated in recent years by advances
in data collection, management, and sharing, which
is exciting for the research community, but these
new projects also pose new challenges. Comparing
large amounts of data across diverse ecosystems can
aid in understanding of ecological processes and the
effectiveness of new research methodologies. When
such analyses lead to new understandings about ecology
and ecological data, the lessons learned can inform the
next round of data collection, processing, analysis, and
documentation. Thus, large synthesis projects have
been increasingly popular over the past few decades
(for example, Riera et al. 2006, Moran et al. 2008).
However, new challenges have appeared with each
large-scale project. Here, we describe the primary 10
challenges that the EcoTrends Project faced, grouping
them into three categories.
The rst category addresses data management and
products. Ideally, datasets would be easy to nd online
and to incorporate into a well-dened workow for
databasing and analysis. However, as the EcoTrends
project illustrates, the task of nding and creating
comparable datasets from disparate sources can be
challenging because of several underappreciated
impediments, including—
• difculties in nding data,
• inadequate data and metadata standards,
• inaccurate or incomplete data and metadata
content, and
• complex datasets.
217
A Basis for Understanding Responses to Global Change
Similar issues have been identied in other
environmental science synthesis projects (for example,
Benson et al. 2005, Jones et al. 2006, Michener et al.
2007, Baker and Chandler 2008).
The second category addresses synthesis project design.
There are many ways to start, design, and implement a
synthesis project, and it is important to begin with well-
dened goals, knowledgeable and enthusiastic partners,
and a well-informed sense of the challenges that may
be faced throughout the project. Challenges in this
category include—
• data heterogeneity and scaling issues,
• planning exibility into project design, and
• making decisions on how to best design and
implement a project and its requisite information
infrastructure.
Finally, the third category addresses information
environments to support synthesis. Challenges
include—
• working with and developing environments in which
information is effectively shared among participants,
• nding motivation to continue the project over time,
and
• encouraging involvement of a large number of
research sites.
Over the course of the EcoTrends project, participants
accumulated a rich body of experience with data
processes and collaborative data practices. While
large datastreams and technology congurations have
prompted a variety of large-scale program endeavors,
the EcoTrends project is unique as a multisite,
multinetwork activity involving ecological data that
span biological, chemical, and physical realms. The
project simultaneously informed development while
coordinating site- and network-level information
environments.
In the next section, we provide recommendations
related to the challenges listed above. For each
recommendation, we rst provide specic examples of
the challenges that EcoTrends faced, then the lessons
that we learned, and then explain the recommendation
that may help address the challenge in future projects.
These recommendations are expected to resonate with
researchers and information managers, who work
together as a cohesive, integrated team at both research
sites and in multisite comparative studies of ecological
data.
Recommendations for Data,
Metadata, and Derived Data Products
1. Make data easily accessible online to
researchers.
Locating data for the EcoTrends Project was a time-
intensive exercise. A small, but signicant, portion of
datasets were not stored online, but were submitted
via email by individual researchers or information
managers. Moreover, online long-term datasets
were often difcult to nd within extensive catalogs
of datasets on the webpage for each research site.
Occasionally, when a research site updated its webpage,
the link to a dataset changed, and the dataset would
have to be relocated by EcoTrends personnel. These
challenges were met by contacting researchers and
information managers at each research site in order to
solicit data that were not online, locate data that were
online but difcult to nd, and nd datasets when they
had been moved.
We recommend that research sites be supported in
developing practices and procedures to make high-
quality, well-documented datasets publicly available
online as soon as possible. For example, the Long
Term Ecological Research (LTER) program data
policy, based on guidelines from the National Science
Foundation, states that data should be posted within
2 years of being collected, with a few exceptions. In
addition, we recommend that each dataset be assigned
locally a unique identier code, or accession number,
that does not change over time. This identier would
make it easier for a synthesis project to more easily nd
a dataset that has been moved. Dataset titles are often
used as identiers, but these titles are subject to change
when datasets are reorganized or displayed at different
Internet locations.
218
Long-Term Trends in Ecological Systems:
2. Implement and develop metadata
standards at the site and community
levels.
The metadata documentation format was highly
variable between research sites. At some research
sites, each researcher documented datasets in a format
unique to his or her personal standards of completeness.
Other sites maintained site-level standards, such as
lling out specic elds in a text document. Data
downloaded from national repositories usually adhered
to the standards created or adopted by that particular
repository. For example, metadata from the Climate
and Hydrology Databases Project reports metadata for
each dataset via a standardized form, the completeness
of which varies between participating sites. The LTER
sites (approximately half of the participating research
sites), however, recently adopted a standard metadata
protocol, the Ecological Metadata Language (EML).
This specication documents datasets with information
such as study location, data collection methods, data
policies, and descriptions of data table elements. It
also includes community-dened lists of terms, or
ontologies, to aid standardization. With EML only
recently adopted by the LTER community, many LTER
datasets were not yet fully documented and many
documentation best practices are still in development.
As a result, the metadata documents that EcoTrends
personnel worked with were highly variable between
datasets and were error-prone, such that time was spent
trying to understand the data. In metadata documents,
the locations where data collection took place were
often missing. We found that a lack of variable naming
conventions (for example, primary productivity
may be labeled “primprod” in one table, and “PP”
in another table—even within the same study) made
data processing difcult. Species names were often
recorded as codes in data tables, yet in many cases,
the codes contained typographical errors or were not
adequately documented in the metadata. In other cases,
a lack of detail in the methods led to misinterpretations
of how the data were collected. Discussions between
the EcoTrends Project Ofce (EPO) and the lead
researcher of the study became a necessary component
in processing the data correctly.
EML was developed for a large, diverse community
that intended to share data using standards that support
consistent data packaging and routine update of datasets
over time. The EcoTrends Project found that source
datasets with EML documentation were often easier
to understand and process than those without such
documentation, thus the Project used EML to document
every derived dataset that the project generated. These
metadata documents contain information about the
source dataset (including ownership and a link to the
original metadata) and about the EcoTrends Project as
well as denitions of the associated data table.
However, while the EcoTrends Project attempted to
support the existing EML standards as thoroughly as
possible, the resulting documents were incomplete.
For example, the methods used to calculate the derived
data from the source data are not included in the EML
because a standard does not exist for this information.
Derived datasets on the EcoTrends website may thus be
misinterpreted, and the source data should be examined
before proceeding with further analysis.
EcoTrends work brought the concept of derived data
to the foreground. The issue of data misinterpretation
was discussed with the broader community, prompting
discussions about how to best accommodate this level
of information within future EML schemas.
EML content standards are still in development, which
means that a number of data comparability issues
remain undened. LTER information managers have
been prominent advocates for improvement of EML,
thereby beneting the ecological research community.
EcoTrends contributed to the development of site-level
conventions and to the enactment of metadata standards
by reporting documentation errors to site personnel.
Specically, benets included prompting sites either
to create EML for their historical data or to improve
on what was available; to standardize attribute, unit,
and taxonomic codes and names; to esh out methods
sections; and to provide stable Internet addresses
(preferably with dataset accession numbers) for each
dataset over time.
We recommend that research sites implement
community-wide metadata standards, such as EML,
and become involved in the process of rening existing
standards and developing new local standards when
community standards are not adequate for local
research. Implementing local procedures with reference
to community standards helps maintain data integrity at
both the site and project levels. Standards that guide the
documentation of a scientic study, its methodology,
and the resulting data tables, can promote responsible
sharing and use among researchers by clearly
representing dataset origin and can make data more
discoverable via online searches.
219
A Basis for Understanding Responses to Global Change
3. Develop and use standard data practices
to create “clean” data.
Data lose their integrity if there are errors. We
consider “clean” or quality-controlled data to be
free of typographical or value errors and to be easily
importable into a spreadsheet, a statistical program, or
a database. In practice, there were frequent errors found
in the source data that signicantly hindered analysis
and synthesis. For example, time-series data often had
unexplained gaps. Occasionally, incorrect values, such
as outliers or incorrectly labeled data (for example,
mean temperature labeled as maximum temperature)
were found by the EPO during the data processing or
during data checking by site personnel. Outliers often
existed in the data early in the study when techniques
were new and the collection process had not been
thoroughly tested. Where data and metadata gave no
indication of poor quality or missing value assignment,
problem data were inadvertently used in the initial
analyses and corrected in the nal analyses and graphs.
There are several plausible reasons for a lack of data
integrity. Long-term data, assumed to be “clean” due to
the long period of time that they have been maintained
and their availability on the Internet, may actually
suffer from neglect. Legacy data practices such as
short and nondescriptive variable names or inadequate
software tools for checking are often an issue.
Alternatively, when delivery of data from site changes
(for example, becomes updated, semiautomated, or
automated), quality control, and other site-level analysis
work may not be carried out or may not be adequately
incorporated into the dataset.
By presenting source data in a recast form on a website,
EcoTrends focused the attention of site participants on
quality-checking of those datasets. Frequently during
the site data checking process in 2008-2009, site
personnel noticed erroneous data points in the annual
summaries of their datasets, attributable to poor-quality
primary data or to erroneous summarization of the data.
Many source datasets and EcoTrends-derived datasets
were corrected following discussions about data
practices that occurred with individual researchers and
at larger meetings.
While good data practices goes beyond the scope of
this chapter, we recommend that sites act upon the
developing resources available in the literature at the
community level (Michener and Brunt 2000, Cook et al.
2001, Baca 2008, Borer et al. 2009) and the national or
international level (NISO 2004, Van den Eynden et al.
2009). Data processing is an iterative exercise involving
multiple facets, from sample analysis and measurement
calibration to data analysis, quality control, statistical
analysis, comparative study, and visualization. All of
these activities can occur at both the site level, driven
by scientic inquiry for a specic use of the data,
and at the multisite or network level, driven by new,
often synthetic uses of the data. Site-based analyses
to scrutinize the data are needed before data can be
used effectively by others. Development of good
information-management practices must include ways
to prevent misuse and/or misinterpretation of data.
4. Provide well-documented derived data
for use by local and remote researchers.
In many cases, the source data were complex and
difcult to process correctly due to unique collection
and analysis methods. A goal of the EcoTrends Project
is to create derived data products whose format is
much simpler than the way the data were originally
collected in order to ensure that a broad range of users
can understand the data. The EPO, in consultation with
the science advisory committee, aggregated data using
methods commonly used by ecologists. Most of the
time, these methods worked well. However, in some
cases no matter how well documented and how cleanly
represented in data tables, the complexity of the dataset
was the main barrier to synthesis. Biotic datasets were
particularly challenging, with numerous species and
different kinds of measures. In many cases, the Project
Ofce needed to discuss with the lead researcher the
suitability of a dataset for a particular aggregation
effort.
We recommend that research sites create and post
online derived data products as long-term, signature
datasets. These types of derived data products are not
typically posted online, though they are often created
and used for in-house analysis. There are two main
reasons for our recommendation.
First, creating derived datasets provides a mechanism
for performing regular checks on the integrity of the
data, a procedure that helps ensure “clean data” (see
recommendation 3). If the data are kept up-to-date
in a standard format, then statistical programs can be
written to periodically recheck the format of the data
tables themselves, check the data table contents against
what is recorded in the metadata, check for errors in
220
Long-Term Trends in Ecological Systems:
the data, and produce visualizations of the data that
an experienced researcher could quickly check for
anomalies. This recommendation would increase the
integrity of the data and increase the stature of the
dataset as other researchers use the data over time.
Second, posting in-house, high-quality derived data
could have great benets for collaborative research by
assuring the use of appropriate and accurate derivation
methods. Moreover, when routinely available, derived
data become a shared product that may prompt dialogue
among researchers. Several discussions were initiated
between the Project Ofce and sites when datasets were
complex and the data aggregation or summarization
approach was unclear. For example, while implicitly
known as being important at the site level, month-long
oceanographic cruises carried out three times a year are
rarely integrated to give annual estimates. In general,
a check on the regularity and frequency of sampling
is required before annual estimates are calculated.
Researchers used to working with terrestrial data may
inadvertently create annual summaries of the data, not
being aware of the issues associated with the logistics
of cruises and oceanographic sampling. However, if
derived data were made available, along with links
to the source data from which they were created and
the methods with which they were derived, including
algorithms and scripts, they would provide a standard in
data quality and use and would increase the integrity of
the dataset in its entirety.
Recommendations for Project
Design
5. Plan for data heterogeneity and
“complexities of scale.”
Data are collected, quality-checked, and organized in
various ways depending on the phenomena sampled
(such as bird counts or wind measurements), the
spatial distribution (for example, single vs. multiple
locations), frequency of sampling (for example, daily
vs. quarterly), regularity of sampling (missing days
in a daily record, for example), and methods of data
collection (for instance, an observer vs. an instrument).
Heterogeneity in data management methods adds
to the challenge of producing comparable data. For
the EcoTrends Project, we focused on time-series
data of specic variables which mitigated some
effects of incoming data heterogeneity. However, no
single programming solution could be developed to
automate data handling; programming solutions were
developed for single datasets or clusters of similar
datasets. To share standardized derived data on a
website, data summarization and organization were
optimized for display of single variables over specic
time aggregations (for example annual bird counts
or monthly wind speed). Decisions made to simplify
website development, such as only graphing variables
through time in the EcoTrends Project, resulted in
limitations in the current underlying data structure.
Data are also collected and aggregated at different
temporal and spatial units under a variety of
circumstances. Scaling from small to large regions and
from short to long time periods can involve complex
processes. For example, sites collect weather data using
a varying number of stations distributed across the
land. The EcoTrends Project asked each site to identify
“representative” weather datasets from their site. For
some sites, particularly those that have relatively at
surfaces, choosing data from site headquarters was
sufcient because differences between stations were
relatively small. At other sites, however, particularly
those with major elevation differences within a small
area, choosing a “representative” dataset was difcult.
If the EcoTrends Project was expanded to use long-term
data from all weather stations at each site, this quandary
would be side-stepped only to introduce scaling issues
due to an increase in the number of datasets to be
handled.
The multiple options for presentation of data also
introduce complexities of scale. The initial plan—for
a website with static content containing data shown
graphically in this book—changed to planning for
dynamic data delivery and visualization. The Technical
Committee recommended structuring the data and
database to support automated metadata generation
for derived datasets using existing tools that were
under develoment (EML for documenting derived
datasets and Metacat for cataloging the resulting
EML documents) and tracking data provenance and
versioning. This proved to be a signicant increase in
project scope and requirements for information system
design and infrastructure building.
We recommend that, before a multisite synthesis
project is completely planned and started, the project
leaders recognize and consider carefully the project
scope, accounting for the variety and complexity of
the source data as well as the constraints associated
221
A Basis for Understanding Responses to Global Change
with their management. Such advance planning is key
to adequate and appropriate information management
for such synthesis projects. We also recommend that
project leaders consider how to best present their
data before implementing information management
solutions. For example, will the data be presented, as
in EcoTrends, as time series? Or will it be expected
that different variables will be compared against one
another or against non-time-series data? Planning for
additional functionality after the project has begun
may require changes in how datasets are organized.
Therefore, accounting for data heterogeneity and
scaling complexity, both in the source data and the
resulting data, before the project begins is important.
Information specialists trained in both economies
of scale and complexities of scale can add insight to
project planning (Baker and Chandler 2008).
6. Iteratively design and assess project
processes and systems.
Interdependent information environments existed at
research sites EPO and LNO. Work at the interfaces of
these environments involved an unanticipated amount
of coordination and design work as well as mediation,
negotiation, and decisionmaking.
The EcoTrends Project started with a linear workow
(traditional for many data management processes),
but the workow rapidly evolved into a cyclical set
of processes using feedback from participants to
inform further development. Just as the scientic
process often does not proceed linearly, there was
value in envisioning the data processes as a complex
set of interdependent systems, sometimes operating
on differing time scales. In the case of the EcoTrends
Project, feedback from discussions among various
groups subsequently informed further development.
Similarly, data handling cannot be solved by a single
technical solution, but rather requires ongoing redesign.
Our recommendation for improving data handling and
information management is to plan for modications,
whether in the short term or the long term, according
to insights gained and lessons learned throughout the
process. For example, when initial assumptions about
the readiness and easy access of long-term data and
metadata from site web pages proved to be incorrect,
the science advisory committee was formed to inform
the process of identifying the variables and datasets of
interest and the common aggregations to be performed.
The project coordinator position was developed to
work directly with site personnel to obtain, correct,
and understand their data in preparation for inclusion
as derived data products and to ensure that committee
decisions were followed. As the volume and complexity
of the data increased, new communication systems
evolved, including ways to share derived data with site
contributors. The project coordinator position expanded
into an interactive role in both assembling data and
creating the derived products needed for the EcoTrends
Project and in providing feedback to site personnel
on the quality of their data and metadata. Iterative
modication of a project may include striving to rene
conceptual models of how data are stored and related,
continuing design of information systems, working
iteratively in phases, and incorporating inquiry-based
collaborative learning.
7. Involve advisors from elds who reect
the breadth of the project and who are
experienced with information management.
Science-driven ecological synthesis projects may be
either narrow, focusing on a single variable over space
or over time, or broad with respect to space, time, and/
or variables. In either case, advice from experts in
the elds that the project embraces is highly useful.
The breadth of the EcoTrends Project mandated the
collaboration of experts in different elds without
which EcoTrends would have fallen short of its goals.
When EcoTrends was rst started, communications
regarding project development were principally
between two scientists and site principal investigators
because it was thought that the data of interest would
be easily accessible online. When it was discovered
that the data were difcult or impossible to nd, the
project was formulated more formally. The science
advisory committee was formed to widen the breadth of
scientic knowledge and the technology committee was
formed to inform technological development (chapter
2). Communications were then expanded to rst
include researchers from each site, then information
managers. The LNO formally became involved when
supplemental funding from the National Science
Foundation became available.
The combined advice from a wide range of expert
contributors had a profound effect on the success of the
project. We recommend for a new synthesis project that
the project leader(s) recruit experts whose knowledge
spans the breadth of the anticipated project and that
222
Long-Term Trends in Ecological Systems:
they be involved at the start of project planning. This
expansion should include not just experts in the focal
science but also experts in roles necessary for the
implementation of the project, such as information
systems designers, information managers, and
statisticians.
Recommendations for Improved
Information Environments To
Support Synthesis Products
8. Focus on development of both local and
network information environments.
An “information environment” is a collection of
scientists, information managers, and analysts and
of the technology needed to manage and share
data. Effective information environments involve
development of shared language, conventions,
and practices for communication among people
from different backgrounds. These environments
exist at both site and network levels. They include
development and use of technical, organizational, and
social work processes to manage multiple types of
data and the translation of science. Comparing data
from multiple sites can stimulate new information
management activities and approaches; however, work
on collaborative data activities must be constantly
balanced with the need to meet site requirements.
The EcoTrends Project needed an effective information
environment to successfully manage data and
communications. The environment established
included a technological system to track, process, and
manage data and a communications system to support
collaboration and decisionmaking among participating
scientists, information managers, and developers.
These systems had to develop iteratively with lessons
learned from one iteration informing the development
of the next. Specically, these systems promoted
understanding of technical and cultural issues regarding
data; informed decisions on how data should be
selected, processed, and shared; and provided feedback
on data handling. Time invested in identifying,
developing, and using coordination mechanisms
accounted for a large amount of unplanned time that
was ultimately recognized as well spent.
We recommend that sites that already have information
environments continue to invest in their multifaceted
growth and ongoing redesign and that sites without
a formal environment dedicate time to developing
strategies for creating one, even if resources are scarce.
The rewards of a smoothly operating set of practices
and systems more than compensate for the cost.
9. Combine long-term data handling with
short-term scientic products and data
checking procedures.
Throughout the several years that the EcoTrends
Project needed to produce its intended products—this
book and a complementary website—it was important
to keep participants engaged with the project and to
share preliminary products. EcoTrends generated both
short-term scientic products and periodic data checks
requested by the participating sites. The scientic
products included papers written by the 2009 scientic
working groups. These prompted review of the website
content and accessibility, fostered new ideas for future
website features and content, and motivated supporters
of the project. EcoTrends also developed a data quality
report when requesting sites to check their derived
data. Created as a spreadsheet and distributed easily
by email, this le provided a much needed feedback
mechanism for sites and provided a useful, albeit
improvised, approach to recordkeeping. Each round of
responses from the sites after a data-checking session
generated improvements to the report. In the long term,
however, a more sophisticated online solution may be
more robust, transparent, and user-friendly.
Balancing long-term goals with short-term actions is
central to development of a contemporary information
environment. Juxtaposing the fulllment of immediate
tasks within a well-dened long-term project creates an
environment in which design can be proactive planning
for the future while meeting immediate needs. Short-
term scientic products, such as papers that examine the
data, can justify the usefulness of the project, motivate
participants to continue with further development,
and inform future development. Data-checking events
can validate data processing, elicit feedback from the
supporting community, and generate enthusiasm for the
project. However, short-term products may require the
development of new methods or work-arounds to create
them, potentially involving new analysis procedures,
communication mechanisms, or types of collaborative
223
A Basis for Understanding Responses to Global Change
activities. These methods or work-arounds can be
very useful, but they should inform long-term project
development.
10. Develop and maintain transparency by
fostering communication and feedback.
Project transparency refers to making participation,
processes, and systems accessible and clear for both
those closely involved and those casually connected to
the project. Transparency requires constant attention
to ensure availability of information and openness
of the decisionmaking process. While the original
intent of the EcoTrends Project was to be open and
inclusive, identifying and developing mechanisms
for collaboration and documentation took time.
Initially, the existing LTER community networking
infrastructure—from listservs to use of regular LTER
community meetings and monthly information
management video conferences—served the project
well. However, there was a persistent push to create
and continue collaborative activities that would open up
discussions concerning data by EcoTrends committees
or individual research sites to a public arena that could
engage a full spectrum of data providers and users.
The EcoTrends Project Ofce communication systems
evolved in response to projects’ and participants’
needs. For example, an initial group email request
for data submission was followed by individual site
communications; committee work with individual
hardcopies of graphs evolved to presentation of graphs
on an internal website. Presentations at community
events improved multisite awareness and engagement.
Initial contact with principal investigators and selected
members of committees eventually broadened to
include information managers and eventually the LTER
information management community. The development
of a site-specic spreadsheet summarizing dataset
submissions created much needed feedback to sites and
a coordination mechanism for joint recordkeeping, both
within a site and between sites and the Project Ofce.
Graphical representations were referenced online to
allow sites to check their contributions.
Attention to project transparency improved both
quality and quantity of data submitted, inuenced the
practice of collaborative science, and promoted buy-
in to the EcoTrends Project by participants at all sites.
We recommend that future projects assess the needs of
their stakeholders as involved and engaged participants
and plan accordingly for project transparency.
Research into existing communications systems and
online networking tools may help. In addition, we
recommend that the project be poised to evolve their
communication systems as further needs are perceived.
Conclusions
In this chapter, we presented key lessons learned and
recommendations for future synthesis projects from the
perspective of a distributed information management
team tasked to support network-level ecological
research. Alternatively, a site-based research scientist
using the data from such a project might have further
recommendations on how to best expand analysis teams
and develop software routines to statistically explore
the data. A software or database developer might have
further insights in framing unique, iterative design
situations for use in dynamic synthesis environments.
Successful planning of any large data synthesis project
can be signicantly enhanced by the perspectives and
knowledge of people from diverse backgrounds and
experience.
The EcoTrends Project can be considered a success for
the following reasons:
• First, this book, with a diverse array of summarized
long-term data collected from 50 sites, and an
associated website with some searching and data
exploration functionalities fulll the initial goals of
the project.
• Second, EcoTrends contributed signicantly to
both individual- and community-level understanding
of multilevel information management by providing
hands-on experience with multisite data integration.
• Third, the EcoTrends Project was unique in carrying
out a data production process in a collaborative,
interdisciplinary setting with a well-established
information management community and in having
the information system work distributed between
two geographically distinct, but communicating
centers (EcoTrends Project Ofce in Las Cruces,
NM, and LTER Network Ofce in Albuquerque,
NM). These arrangements reveal a number of
underappreciated dimensions of the work involved
in creating comparable data.
224
Long-Term Trends in Ecological Systems:
In addition to the highlighted successes, the EcoTrends
Project demonstrates the importance of addressing and
supporting knowledge production, data production,
and infrastructure growth within a single framework.
The project also highlights the importance of
broadening participants’ perspectives over time via
transparent processes and communication. Specically,
the perspectives of EcoTrends Project participants
broadened from simply dening digital products and a
single companion workow to eventually envisioning
multiple interdependent data processes and information
environments. These processes and environments
included not only a technical infrastructure but an array
of organizational and social arrangements. Besides
just considering the data and the individual work
arenas, participants learned to consider the variety of
participant roles and activities that tied them together.
Iterative, collaborative learning throughout a project
and planned exibility to react to new ideas were
important elements of the EcoTrends Project and may
well serve any new multisite synthesis project.
References
Baca, M. 2008. Introduction to Metadata. 2nd ed. Getty
Research Institute, Los Angeles, CA.
Baker, K.S., and C. Chandler. 2008. Enabling long-
term oceanographic research: changing data practices,
information management strategies and informatics.
Presentation at Fall American Geophysical Union
Conference, Dec 15-19, 2008, San Francisco, CA.
Benson, B.J., P. Hanson, J. Chipman, et al. 2005.
Breaking the data barrier: research facilitation through
information management. In JJ Magnuson, TK Kratz,
and BJ Benson, eds., Long-Term Dynamics of Lakes in
the Landscape, pp. 259-279. Oxford University Press,
Oxford, United Kingdom.
Benson, B.J., B.J. Bond, M.P. Hamilton, et al. 2010.
Perspectives on next-generation technology for
environmental sensor networks. Frontiers in Ecology
and the Environment 8:193-200.
Borer, E.T., E.W. Seabloom, M.B. Jones, et al. 2009.
Some simple guidelines for effective data management.
Bulletin of the Ecological Society of America 90:205-
214.
Collins, S.L., L.M. Bettencourt, A. Hagberg, et al.
2006. New opportunities in ecological sensing using
wireless sensor networks. Frontiers in Ecology and the
Environment 4:402-407.
Cook, R.B., R.J. Olson, R. Kanciruk, et al. 2001. Best
Practices for Preparing Ecological Data Sets to Share
and Archive. Bulletin of the Ecological Society of
America 82:138-141.
Jones, M., M. Schildhauer, O.J. Reichman, et al. 2006.
The new bioinformatics: integrating ecological data
from the gene to the biosphere. Annual Review of
Ecology, Evolution, and Systematics 37:519–544.
Karasti, H., and K.S. Baker. 2008. Digital data practices
and the Global Long Term Ecological Research
Program. International Journal of Digital Curation 3:42-
58.
Michener, W.K., and J.W. Brunt. 2000. Ecological
Data: Design, Management and Processing. Blackwell
Science, London.
Michener, W.K., J.H. Beach, M.B. Jones, et al. 2007.
A knowledge environment for the biodiversity and
ecological sciences. Journal of Intelligent Information
Systems 29:111-126.
Moran, M.S., D.P.C. Peters, M.P. McClaran, et al. 2008.
Long-term data collection at USDA experimental sites
for studies of ecohydrology. Ecohydrology 1:377-393.
NISO [National Information Standards
Organization]. 2004. Understanding Metadata.
NISO Press. http://www.niso.org/publications/press/
UnderstandingMetadata.pdf. Bethesda, MD.
Porter, J., P. Arzberger, H. Braun, et al. 2005. Wireless
sensor networks for ecology. BioScience 55:561-572.
Porter, J.H. 2010. A brief history of data sharing in the
U.S. Long Term Ecological Research Network. Bulletin
of the Ecological Society of America 91:14-20.
Riera, J.L., T.K. Kratz, and J.J. Magnuson. 2006.
Generalization from intersite research. In J.J.
Magnuson, T.K. Kratz, and B.J. Benson, eds., Long-
Term Dynamics of Lakes in the Landscape: Long-Term
Ecological Research on North Temperate Lakes, pp.
107-120. Oxford University Press, New York, NY.
225
A Basis for Understanding Responses to Global Change
Van den Eynden, V., L. Corti, M. Woollard, et al. 2009.
Managing and Sharing Data: Best Practice Guide for
Researchers. www.data-archive.ac.uk/media/2894/
managingsharing.pdf. United Kingdom Data Archive,
University of Essex, Wivenhoe Park, Essex, United
Kingdom.
Vanderbilt, K.L., D. Blankman, X. Guo, et al.
A multilingual metadata catalog for the ILTER: issues
and approaches. Ecological Informatics 5(3):187-193.
vi