Professional Documents
Culture Documents
The Impact of Future Land Use Scenarios PDF
The Impact of Future Land Use Scenarios PDF
DOI 10.1007/s00267-010-9533-z
Bryan C. Pijanowski
Abstract In this article we compared the response of configuration, twenty-five (62.5%) of the subwatersheds
surface water runoff to a storm event for different rates of produced the greatest amount of runoff in 1900, shortly
urbanization, reforestation and riparian buffer setbacks after the entire watershed was clear-cut. One third (14/40)
across forty subwatersheds of the Muskegon River of the subwatersheds contained the minimum amount of
Watershed located in Michigan, USA. We also made these runoff in the 1960s and 1970s, a period when forest
comparisons for several forecasted and one historical land amounts were greatest and urban amounts relatively small.
use scenarios (over 140 years). Future land use scenarios to
2040 for forest regrowth, urbanization rates and stream Keywords Land use change modeling Surface water
setbacks were developed using the Land Transformation runoff Policy Impacts
Model (LTM). Historical land use information, from 1900
at 5-year time step intervals, was created using a Backcast
land use change model configured using artificial neural Introduction
network and driven by agriculture and housing census
information. We show that (1) controlling the rate of The distribution of land use influences the hydrology of
development is the most effective policy option to reduce watersheds across a variety of spatial and temporal scales
runoff; (2) establishing setbacks along the mainstem are (Leopold 1968; Harbor 1994; Mölders and Rühaack 2002;
not as effective as controlling urban growth; (3) refores- Pijanowski and others 2002a; Tang and others 2005a; Tang
tation can abate some of the runoff effects from urban and others 2005b). Agricultural and urban dominated
growth but not all; (4) land use patterns of the 1970s landscapes increase runoff compared to a landscape that is
produced the least amount of runoff in most cases in the in its natural state (Leopold 1994; USEPA 2000; Tong and
Muskegon River Watershed when compared to land use Chen 2002; NRC 2008). The increased runoff carries with
maps from 1900 to 2040; and, (5) future land use patterns it nutrients (e.g., nitrogen and phosphorus) from farm fields
here not always lead to increased (worse) runoff than the (Carpenter and others 1998) and contaminants from urban
past. We found that while ten of the subwatersheds con- activities (USEPA 1983; Wayland and others 2002;
tained futures that were worse than any past land use Wayland and others 2003; Fitzpatrick and others 2007).
Increased runoff can also lead to more frequent or intense
floods (Bhaduri and others 2000; Booth and others 2004)
and alter the apportionment of water to wetlands and
D. K. Ray J. M. Duckles B. C. Pijanowski (&)
groundwater (Mitsch and Gosselink 1993). Nutrient rich
Department of Forestry and Natural Resources, Purdue
University, West Lafayette, IN 47906, USA runoff from agriculture is considered to be the primary
e-mail: bpijanow@purdue.edu cause of eutrophication of surface water bodies and the
expansion of dead zones (Diaz and Rosenberg 2008),
Present Address:
which negatively impacts sensitive biological organisms,
D. K. Ray
Institute on Environment, University of Minnesota-Twin such as fish and aquatic invertebrates (Wang and others
Cities, Saint Paul, MN 55108, USA 1997; Cifaldi and others 2003; Allan 2004).
123
Environmental Management
Several best management practices exist to protect sur- For example, in the Muskegon River Watershed, the mid-
face water bodies, such as rivers and lakes, from urban and portion of the Muskegon River mainstem is highly chan-
agricultural runoff (O’Callaghan 1996; Diebel and others nelized but upper and lower portions of the watershed flow
2008). These best management practices include control- over relatively flat landscapes. It would be important to
ling the rate of development (cf. Paul and Meyer 2001), know the effect of using a fixed size buffer as a policy
protecting highly erodible soils or areas with significant option in reducing runoff volumes.
hydrologic impact from agricultural use (e.g. Bradford and Walter and others (2009) argue that buffer configura-
others 1987) and the creation of setbacks or buffers (Bren tions should be made on the basis of ‘‘the potential for an
1998) that prevents urban development or agriculture area to generate runoff’’. Such an assessment can be made
occurring within the riparian zones of rivers and streams. using simple runoff models (Walter and others 2009) such
Controlling the rate of urban development however is as the Soil Conservation Service (SCS) Curve-Number
challenging. At current rates of urbanization, a doubling in (CN) equation (USDA 1986). The CN equation estimates
the amount of urban area will occur in 20 years for most of runoff volume from a ratio of effective rainfall and maxi-
the United States (Alig and others 2004). In the Upper mum soil storage capacity. It has also been argued (e.g.,
Midwest of the United States urban development has been Tomer and Burkart 2003; Dosskey 2002) that areas that are
rapid over the last 10–20 years (Pijanowski 2006; Wolter hydrologically sensitive, such as locations in watersheds
and others 2006), with many areas in this region increasing where runoff would impact streams and rivers directly
their urban footprints at an average of 4.4% per year through high hydrologic connectivity, should be targeted
(Pijanowski unpub). In Michigan’s Muskegon River for setback regulations.
Watershed, the focus study area of this article, Tang and Given the myriad of management options available to
others (2005a) found that runoff and non-point source control runoff, which are most important? Addressing these
pollution was highly dependent on urbanization rates; a factors in land use planning and natural resource manage-
doubling of urban would increase runoff by 1.5 times over ment has been successfully explored through the use of
a baseline land use distribution (assumed as 1998). Thus, coupled land use-hydrologic models (Harbor 1994). When
one way to control the future runoff rates would be to slow applied in a policy context, these coupled models can be
urban growth rates. used to forecast land use change or best management
Increasing forestland can also reduce runoff rates. In the options into the future using policy scenarios as one
Muskegon River Watershed, nearly the same amount of component of the land change model (e.g., Koomen and
forestland was added to the watershed from 1978 to 1998 Stillwell 2007) and hydrology examined using any number
as was new urban, partially as a result of the implemen- of hydrologic models.
tation of USDA’s Conservation Reserve Program (CRP) in Assessing the relative impacts of different policies or the
1985 to reduce sediment loading to streams by converting implementation of best management practices on future
high erosion cropland to natural vegetation. However, water quality and runoff rates will require a benchmark
while watersheds across the United States continue to with which to compare. These benchmarks could involve
urbanize, the relative effect of increasing forestland that comparing output of coupled models to simulations of
serve as a mitigation strategy towards reducing the runoff current conditions (e.g., Tang and others 2005a) or optimal
from increased urbanization is unknown. configurations of land use in a watershed (Tang and others
The creation of setbacks on rivers and streams has also 2005b; Chaubey and others 2008). Historical trends offer a
been argued to protect surface water from non-point source third means of generating baselines for comparing future
pollution (Bren 2000) by reducing surface water runoff projections of runoff due to changes in land use patterns.
rates. Traditionally, riparian setback regulations establish Historical comparisons could be very useful as many areas
minimum distances that buildings or certain agricultural of the world have undergone significant changes in land
activities can occur from surface waters. Establishing set- use. Via such comparison, we can answer the question
backs or buffers along streams is also known to decrease whether the surface runoff volumes projected for future
bank erosion (Shields and others 1995). However, there is land use changes would be any worse than those that have
considerable debate as to the appropriate size and config- occurred in the past.
uration that these buffers need to be in a watershed This article presents the results of coupled land use-
(Dosskey 2002) to control runoff. It is accepted by many runoff simulations for Michigan’s Muskegon River
researchers that buffer sizes and configurations are likely to Watershed designed to address the following research
be influenced by topography (e.g., slope), soil type and questions:
land use; there is likely to be tremendous variation in slope,
soil type and land use across regional watersheds making 1. What is the impact of afforestation patterns on runoff?
the establishment of standard setback distances unpractical. Does the presence of new forests cancel the effects of
123
Environmental Management
further projected urbanization? What is the effect of Divisions or MCDs). There are 124 MCDs that fall wholly
different urban development rates on total runoff or partially within the watershed. Each is required to
amount? develop Land Use Master Plans and create ordinances and
2. What is the affect on runoff rates by establishing zoning laws that control land use decisions. The city of
different setback distances (none versus 100 m versus Muskegon, the largest city in the Muskegon River Water-
300 m) on the mainstem of the Muskegon River? Are shed with a population of about 40,000 in 2000, is located
fixed setbacks along riparian corridors sufficient or along southern shores of Muskegon Lake.
should setbacks be defined more ecologically using a The main river, the Muskegon River, runs the length of
hydrologically defined setback distance on runoff the watershed from its headwaters at two large inland
amounts? lakes (Houghton and Higgins Lakes) to its outlet in
3. How do future scenarios of land use-runoff differ from Muskegon Lake which is a drowned river mouth. Mus-
runoff patterns generated for past changes in land use? kegon Lake connects to Lake Michigan through a narrow
Are future scenarios of runoff better or worse than the channel. The watershed is composed of 40 subwatersheds.
past? Major subwatersheds connect inland lakes, wetlands and
tributaries to the mainstem of the Muskegon River mak-
We show that (1) controlling the rate of development is
ing this a very hydrologically connected watershed. The
the most effective policy option to control runoff; (2)
peak flows in this system tend to be in the spring asso-
reforestation can abate some of the runoff effects from
ciated with either snowmelt or large rain events. The
urban growth but not all; (3) the size and configuration of
average annual rainfall from 1899 through 2001 was
buffers are not as effective as controlling urban growth; (4)
83 cm (32.7 in).
land use patterns of the 1970s produced the least runoff
The Muskegon River Watershed serves as an important
rates in most cases when land use maps from 1900 to 2040
spawning and nursery site for economically important
are compared; and, (5) future projected land use patterns
Great Lakes fish such Chinook salmon (Oncorhynchus
would not always lead to increased (i.e. worse) runoff rates
tshawytscha), steelhead (Oncorhynchus mykiss) and wall-
than the past.
eye (Sander vitreus). Assessment of the environmental
conditions of the Muskegon River and its watershed has
been made several times over the last 30 years. In the
Background 1980s, a series of reports by the Michigan Department of
Natural Resources (cf. O’Neal 1997) identified runoff,
Study Area thermal pollution, toxic substances from industry, and
aquatic invasive species as the main threats to the health of
This study was conducted in the Muskegon River Water- this river. In the late 1990s, a group of scientists from a
shed, located in the west-central part of the Lower Penin- regional university (Grand Valley State University—
sula of Michigan, USA (Fig. 1). The Muskegon River GVSU), members of natural resource agencies (e.g.,
Watershed (7,057 km2) in 1998 was dominated by forests Michigan Department of Natural Resources), city govern-
in the north, agriculture in the central portion, and urban in ment officials (e.g., planners, drain commissioners), and
the south. The watershed was heavily logged in the late personnel from regional economic groups began meeting to
1880s through the 1890s, mostly in efforts to rebuild discuss the health of the Muskegon River Watershed
Chicago after the 1871 fire (Alexander 2006), and agri- (GVSU 2002a, b). These efforts culminated in a Sec-
cultural expansion occurred in the 1910–30s and then again tion 319 grant (cf. EPA 1972) awarded to GVSU from the
in the 1960s (Ray and Pijanowski 2010). After World War USEPA and Michigan Department of Environmental
II, a lot of abandoned farmland was taken over by the state Quality. This grant funded the development of a Watershed
of Michigan and converted to State Forests. Currently, Management Plan (GVSU 2002a) that identified several
approximately 20% of the watershed is in public owner- areas of water quality concerns, including thermal pollution
ship, either as State Forest (a majority), Federal Forests or and excessive nutrient loading via runoff from agricultural
Federal Wildlife Reserve. Between 1978 and 1998, the fields. Several best management practices were recom-
amount of urban land use nearly doubled, growing from mended (GVSU 2002c) as part of the 319 study. These
4.1% of the watershed to 7.8%. The amount of forests also included adopting best management practices for (1) con-
increased during this twenty-year period, rising from 46% struction sites; (2) reducing chemical inputs and stream-
of the watershed to over 49% in 1998. bank erosion (e.g., reducing fertilizer application); (3)
Twelve counties fall within the Muskegon River reducing runoff conveyance and outlets; (4) sedimentation
Watershed. Land use planning occurs at the township, city control (e.g., buffer/filter strips); (5) vegetation establish-
or village level of government (also called Minor Civil ments (i.e., to reduce runoff and sedimentation); and (6)
123
Environmental Management
improving agricultural management (e.g., fencing, manure streams would exclude a large portion of the watershed
control, planting strategies). from development, which was not realistic and (2) buffers
on the mainstem would exclude less than 10% of the
watershed if they were 300 m or smaller. Thus, we
Planning and Management examined setbacks of 100 m and 300 m and on the river
mainstem only.
In summer of 2002, the Great Lakes Fishery Trust (Wiley
and others 2008, 2010) identified the Muskegon River as a
new ‘‘River Initiative’’ and in response to this initiative, a Methods
two-day retreat (Wiley and others 2008) was held to
determine how simulation models could be used to address Data Preparation
what-if questions related to land use, water quality and
runoff, and, in particular, which of the several dozen rec- To prepare data for analysis and modeling, the two land use
ommended best management practices could be examined vector datasets (1978 and 1998) were rasterized (i.e., a grid
more thoroughly with coupled land use-hydrologic models. of cells were created from polygons of land use/cover
Participants were asked to discuss the various best man- patches) at 26.5 meters to match the digital elevation model
agement practices that were identified by the 319 study and (DEM) available for the region. Along with the two time
determine scenarios for coupled land use-hydrology mod- steps of land use for the region, we obtained vector data for
els. Workshop participants were interested in having roads, rivers and lakes in order to parameterize drivers for
models address: (1) the impact of riparian setbacks, the LTM (Pijanowski and others 2007). The vector datasets
especially size, on hydrologic dynamics of the watershed; were obtained from the Michigan Geographic Framework,
and (2) the impact of controlling urban development on version 4 (www.michigan.gov/cgi).
hydrologic dynamics. Surface water runoff modeling required a map of
Following the focus group meeting, an assessment of the hydrologic soil groupings for the region which was obtained
size of setbacks was conducted using GIS to determine from the State Soil Geographic (STATSGO) database. We
reasonable setback sizes and configurations. Analysis of then subsetted the soil grouping polygons in the study-region
setback sizes and configurations across the mainstem and and rasterized and recoded soils based upon hydrologic soil
tributaries resulted in the following conclusions: (1) type. The hydrologic soil groups were assumed to remain
establishing setbacks (no development) on all rivers and fixed for the time period investigated.
123
Environmental Management
Modeling Land Use Changes simulated the expected runoff in response to a 5 inch storm
event for each of the subwatersheds to investigate how
Future urban growth was projected using the LTM fol- past, current and future land use patterns affects runoff
lowing Pijanowski and others (2002a, b, 2005). The LTM during such an identical storm event. A five-inch storm
uses drivers of land use changes as inputs; then using an event was selected since it represented a large enough
artificial neural network the LTM ‘‘learns’’ the relationship storm event that would be sensitive to a variety of land use
between the drivers of land use change and observed patterns across this very heterogeneous watershed and its
changes. Once trained, the LTM is used to project the impact could have economic effects as well and thus be
future urban growth using calculations of future population important for policy. Based on historical rainfall events at
growth estimates and per capita urban use requirements. three weather station locations in the Muskegon River
Stakeholder derived scenarios were articulated within the Watershed such a storm is classified as a one in 30 year
LTM by excluding urbanization in a 100 m development storm event.
setback region, 300 m development setback region and In an attempt to determine which future land use sce-
1 year ground water travel time (GWTT) setback (cf. nario was most effective at controlling runoff, we also
Pijanowski and others 2007) in the watershed area. The examined the relative contribution of each policy option on
100 m and 300 m development setbacks were made using runoff amounts compared to a baseline or reference year.
a vector line dataset of mainstem river segments; these To accomplish this, the proportional changes in runoff for
setbacks represented 3% and 6% of the watershed area all years and simulation treatments were standardized
respectively. The stream setback method was applied based across all subwatersheds using the following equation:
on a GWTT model described further in Pijanowski and Ryr;i;tmt R1975;i
others (2007) and based on MODFLOW-2000 (Boutt and Pyr;i;tmt ¼ ; yr [ 1975 ð1Þ
R1975;i
others 2001; Wayland and others 2002, 2003). We created
a policy buffer in GIS which prevented future urbanization where Pyr;i;tmt is the proportional runoff difference for year
from occurring within areas corresponding to GWTT of (yr), subwatershed (i), and treatment (tmt); Ryr;i;tmt is the
1 year or less. This setback covered 22% of the watershed runoff for that year, subwatershed and treatment, and
area. The arrangement of these three setbacks is shown in R1975;i is the simulated runoff for the reference year of 1975
Fig. 2. GWTT is an important factor in the flow of of each subwatershed. Note that prior to 1975 there is no
groundwater pollutants into rivers and streams. In a similar treatment as these are historical years and thus we did not
fashion (but obviously using a different set of driver and calculate Pyr;i;tmt for these years.
regrowth rates) forests were projected for the future. The The final coupled model and associated tools are illus-
details of this procedure are given in Appendix 1. A trated in Fig. 3. The main two models, the land change
summary of the future land use scenarios examined in this models and the surface runoff models use the 5 inch pre-
article are contained in Table 1. The historical land use cipitation event to generate average area weighted surface
changes were generated using a modified version of the runoff per subwatershed. GIS data processing (e.g., soils
LTM called the Backcast model (Ray and Pijanowski and moisture conditions) are also necessary to create the
2010). Details of the Backcast modeling are given in CN runoff model and the processing steps are shown in
Appendix 1. Fig. 3.
In order to evaluate the hydrologic impacts of land use Comparison of Land Use Distributions Across
changes in the Muskegon River Watershed, we chose to the Simulated Scenarios
use the curve number approach (Ponce and others 1996;
Harbor 1994; Garen and Moore 2005) to estimate runoff. A business as usual (Urbbase) urban growth in the Mus-
The curve number method was developed to aid watershed kegon River Watershed was projected to cover 19.9% of
managers by calculating peak storm runoff based on the the watershed with urban land cover by the year 2040,
land use and soil types at a location. It is a frequently used whereas the slow growth mode (Urbslow) was projected to
approach to model surface water runoff. We used this have 13.7% total urban cover by 2040 (Fig. 4; Duckles
method to calculate the peak runoff response to a hypo- 2008). In contrast, the 1998 map of land use for the
thetical storm event, across each land use scenario, treating watershed derived from aerial photography contained 7.4%
each of the 40 sub-watersheds individually. Details of the urban. The six percent difference between urban cover in
specific calculations made are given in Appendix 2. We these two scenarios results in different allocations of
123
Environmental Management
Table 1 Experimental design showing matrix of simulation forecast than doubles; as forest is added a large proportion of it is
scenarios (H = simulations were produced for this scenario) with eventually urbanized. This is further highlighted in the
simulation names
Forslow scenario, where forest cover increases from 58%
Transitions Buffers Urban rates of growth to over 61%; in this scenario, urban growth is controlled. In
Fast Slow the fast urban growth scenario with no addition of forest
results in a decrease of forests from 58% in 1998 to around
New forests and 100 m H (for100 m) 51% by the year 2040.
urban growth 300 m H (for300 m) When riparian development and GWTT setbacks were
1 yr GWTT H (for1 yr) applied to the Muskegon River Watershed, the setback
No buffer H (forbase) H (forslow) restrictions pushed the spatial distribution of new urban
Urban growth only 100 m H (urb100 m) cells outside the setback regions in each of the three sce-
300 m H (urb300 m) narios. Not surprisingly, beginning in 2005, the four
1 yr GWTT H (urb1 yr) urbanization scenarios had 93% cell by cell placement
No buffer H (urbbase) H (urbslow) agreement (i.e., 93% of the urban areas were the same
location for each scenario); by 2040 this map agreement
agriculture and forest land across the watershed, and a was reduced to 80% across the four urbanization scenarios.
small difference in amounts of shrub land use category. Afforestation scenarios had the same urban extent as their
The decline in the amount of agriculture over the 10 sce- respective urban scenarios.
narios is significant, with a majority of the forest scenarios Results of the backcast land use modeling (Ray and
resulting in slightly more than 11% agriculture, down from Pijanowski 2010) are shown in Fig. 5. Note that in 1900,
about 16.5% in 2005. The least decrease in agriculture more than half of the watershed is agriculture, mostly as a
occurred in the Urbslow scenario, which is not surprising result of the massive deforestation in the late 1800s. Over
as urban growth is controlled and reforestation is not added time, forests regrow, starting around 1930, making up more
as a transition to the forecast model. The amount of forest than half of the watershed’s land cover by 1960. Urban was
left in 2040 across the 10 scenarios differs as well. Adding not evident until 1940, continuing a steady increase
forests at historical rates of forest regrowth results in a through 1970. The simulation produced shrubland in about
stable amount of forest cover percentage over the 35 year a constant proportion throughout the 1900 to 1970 time
forecast period (2005–2040). This is because urban more frame.
123
Environmental Management
Runoff Amounts and Land Use Forecasts subwatershed 1 has very little runoff over this 140 year
timeline although there is a steady increase over time.
The amount of runoff generated over time and within each Subwatersheds 9 and 31 on the other hand had moderately
subwatershed is highly variable. Figure 6 shows the runoff high runoff in the past, and then from about 1980 through
plotted from 1900 through 1975 for the historical land 2040, runoff volume is reduced. Several subwatersheds
use and then from 2005 through 2040 for the urbbase also contained sharp drops in runoff starting around 1940,
scenario only. Several trends are evident. For example, which is soon after a lot of agricultural land was abandoned
123
Environmental Management
in the Great Depression and the state of Michigan con- observed in last 20 years) as fixed effects. Interactions of
verted these lands to state forests. transition type with buffer type and transition type with
A summary of the proportional changes (Eq. 1) in runoff urban expansion rate were explored as well. ANOVAs
volumes by future scenario and year and plotted by sub- were executed for each year, 2005 through 2040. A sum-
watershed is given in Fig. 7. The Box plots illustrate that mary of the F-statistics for the ANOVAs for each year is
for all scenarios, runoff amounts increase (i.e., are greater given in Table 2. Note that the urban growth rate (i.e. Rate)
than 0.0, which represents no change compared to 1975) effect contains the largest differences in all years; years
over time. Outliers (designated with open circles and stars, 2025 and 2030 are significant at the P \ 0.05 level and
these are values that exceed 2 SE of the mean) are most years 2035 and 2040 are significant at the P \ 0.01 level.
common later during the forecast period, reaching 2 to 3 The presence of additional forest cover was slightly sig-
times the original runoff amount of 1975. Note that even in nificant (P \ 0.10) in years 2025 and beyond. In all years,
the first year of the simulation, 2005, a few subwatersheds buffer configuration did not alter runoff. Interactions were
experience a doubling (near 1.0) of runoff compared to also not significant. Typically, multiple statistical com-
1975. Forslow and Urbslow produce runoff changes that parisons require a Bonferoni adjustment (Sokal and Rholf
are all less than 2.0. 1995) of the individual a (false positive error) rates; such
The relative effect of each policy scenario on propor- an adjustment produces a family-wise alpha (a/N) of
tional changes in runoff values was examined across policy 0.00625 (0.05/8). If this is used to judge significance, then
scenarios, as treatment effects (4 buffer x 2 rates x 2 only the rate effect for 2035 and 2040 are significant at the
afforestation scenarios), and across years, using a Mixed- P \ 0.05. However, the increasing values of the F-statistic
Model ANOVA in SYSTAT 12. The Basic Statistic over time for both the Rate and Affor effect suggests that as
command in SYSTAT 12 (2008) was used to determine urbanization increases and the amount of new forests are
whether the distributions of values were normal. Skewness added to non-urban areas, the effects of these two policy
and kurtosis tests showed, however, that all distributions options on runoff increases.
were heavily skewed to the left with a long right tail (all The potential effects of buffer size on runoff were
P-values were \ 0.01). As a result, proportional runoff explored further using the following approach. Each buffer
values were transformed using Box-Cox transformation was extracted by subwatershed and the total area of the
with k = 0.2. A Mixed-Model ANOVA was performed in buffer expressed as a proportion of the subwatershed was
SYSTAT 12 (Systat Inc 2008) using buffer (no buffer, calculated using ArcGIS; simulations without buffers were
100 m buffer, 300 m buffer and 1 year GWTT buffer), excluded from this analysis. The proportion of subwater-
urbanization rate (slow and historical average) and affor- shed occupied by the buffer was then transformed using a
estation (no new forests versus forests regenerated at rate natural log. A Pearson correlation between natural log of
123
Environmental Management
prop_change
prop_change
prop_change
2 2 2 2
prop_change
1 1 1 1
0 0 0 0
-1 -1 -1 -1
0
0
00
01
02
03
04
05
00
01
02
03
04
05
00
01
02
03
04
05
00
01
02
03
04
05
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
year year year year
prop_change
prop_change
2 2 2
prop_change
2
1 1 1 1
0 0 0 0
-1 -1 -1 -1
0
0
00
01
02
03
04
05
00
01
02
03
04
05
00
01
02
03
04
05
00
01
02
03
04
05
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
year year year year
urbbase urbslow
3 3
prop_change
prop_change
2 2
1 1
0 0
-1 -1
0
0
00
01
02
03
04
05
00
01
02
03
04
05
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
2,
year year
Fig. 7 Relative proportional increases of surface runoff by scenario and across years. The Box Plot represents the mean, 1SE, 2SE and 3SE from
mean for proportional increases in the 40 subwatershed
Table 2 F-statistics of Mixed Model ANOVAs for tests of effects of buffer size (none, 100 m, 300 m, groundwater delineated), rate of
urbanization (fast and slow) and whether forests are added to watershed at recent (1978–1998) historical rate and pattern
Fixed-effect df 2005 2010 2015 2020 2025 2030 2035 2040
buffer size and natural log of runoff differential (using Comparing Runoff Variations in the Past with Future
1975 as the base simulation) was calculated for each Scenarios
forecast year (2005 through 2040 in 5 year time steps). We
found that all Pearson R2 values were small, ranging from Table 3 lists the year in which maximum and minimum
0.088 and 0.207. In other words, even in subwatersheds runoff occurred for each of the subwatersheds. Out of the
that had large areas, in terms of their proportion that fell 40 subwatersheds in the Muskegon River Watershed, 25
within a buffer, runoff was still not reduced. were at their maximum runoff condition in 1900. While the
123
Environmental Management
123
Environmental Management
land use changes (Feddema and others 2005; Ray and scenarios within a matrix, as conducted here, has been
others 2006, 2009, 2010). proposed as a means to achieve the most robust decision
We found that the amount of surface water runoff varied making in order to achieve sustainability (Groves and
across the subwatersheds and given plausible futures, in Lempert 2007). Challenges still exist, as we attempt to
each of them the slow urban growth scenario, with affor- examine not just trends that are apparent on decadal time
estation, resulted in the least amount of runoff. This was steps (e.g., land use change) but those (e.g., unusual storm
because this scenario had the minimum increases in urban, events) that present themselves as surprises (Lempert and
while continuing to increase forest area, each resulting in others 2002). Other futures, such as those that could be
positive impacts (i.e. reduced runoff rates) on the overall posed by land use changes as a result of biofuels
area-weighted curve number for these subwatersheds. (Searchinger and others 2008) may need further attention.
We found that slowing urban growth was the most The modeling of the coupled interaction between forest
effective means of reducing surface water runoff compared regrowth, urban expansion and setback scenarios could be
to any other policy option—such as allowing forests to improved further by incorporating more process based
grow or adding riparian buffers to the mainstem. This work models, especially hydrologic models, that quantify in
supports Walter and others’s (2009) contention that the greater details water balance components like infiltration,
amount of key land uses, such as urban or forest, are the evapotranspiration, stream flow on a more temporally
most important policy levers compared to any urban refined time scale (examining days or seasonal trends). For
development setback scenario. We found that the imple- example, Fohrer and others (2005) used a set of hydrologic
mentation of stream setback scenarios could result in sur- models coupled to land use change models to determine
face water runoff impacts on subwatersheds where there how urbanization impacted stream flow dynamics. They
are no explicit setback regions. The hydrologically defined found that the model predicted annual trends for the
riparian setbacks did not significantly reduce runoff either. watershed and that only peak flow rates were impacted
Using land use forecasting models to examine the significantly by land use changes. Hundecha and Bardossy
impacts of alternate land use scenarios is becoming more (2003) also examined land use change impacts on runoff
common as a land use planning activity (Landis 1995). patterns in Germany and found that increased urbanization
Indeed, with the advent of more powerful computers and impacted peak runoffs during the summer but not during
readily available models, using models to explore policy the winter; afforestation on the other hand, reduced peak
options is likely to become even more prevalent in the near and total runoff in both the summer and winter.
future (cf. Bankes 1992). Models are helpful in character- The historical assessment provides useful context to
izing spatial, temporal and ecosystem variation in respon- gauge the impacts of future changes. The curve number
ses to human perturbation (Rodrguez and others 2006). analysis of past land covers show that for many of the
Recently, more attention has been paid to historical land- subwatersheds the worst runoff condition in the history of
scapes (Foster 1992; Kim and others 2002; Baker and the subwatershed was during the time that the watershed
others 2004) such as in this assessment. Past actions put was largely deforested at the turn of the 20th century
into place structural and functional aspects of ecosystem (Whitney 1987). Furthermore, the best hydrologic condi-
dynamics that need to be considered as we explore sus- tion for many of these subwatersheds occurred during the
tainable planning solutions and management of landscapes 1970s and 1980s. The damage done to the runoff hydrology
for the future (Antrop 2005). Using such models, the of the watershed by clear cutting forests, in many cases, far
effects of future variations in sub-categories of urban land out-weighs the combined impact of steady abandonment of
use and their role on runoff could also be addressed, i.e. agricultural land back to forest and increasing urbanization.
besides setbacks and GWTT, smart growth urban policies The drastic impacts of deforestation are indicated in the
are also alternate policy scenarios. For example, future poor (i.e. large) runoff rates of many of the subwatersheds
studies can evaluate how residential, commercial and for the year 1900. As the century progressed and agricul-
industrial urban landscape is constructed to minimize the tural lands were abandoned, there was steady improvement
amount of surface runoff on the landscape where runoff is in overall runoff condition for nearly every subwatershed.
important. The exceptions tend to be heavily urbanized subwatershed
Engagement of stakeholders in research and modeling areas, such as subwatershed 4 which has significant urban
per se is also becoming more common where stakeholders and open water, and subwatershed 38 which contains much
help to identify scenarios, solutions and provide the context of the city of Muskegon. With these exceptions, most of the
for ‘‘how we got to where we are’’ (Walker and Salt 2003). watershed areas improved their surface water runoff con-
Impacts of research results are shown to be much greater dition until the late 1970s. The steady subsequent increase
when stakeholders are engaged early and for a long term in urbanization was found to increase runoff across many
(Johnson and others 2003). Examining an array of possible of the subwatersheds since the late 1970s.
123
Environmental Management
Despite showing that establishing no-development set- distance to unlimited access to highways (state roads). A
backs along the buffers of the mainstem have less impact separate GIS layer of non-transitioning locations of areas in
on controlling surface runoff, riparian buffers do have state and federal forests, all urban categories in 1978, and
many other positive benefits to maintaining water quality areas that are in open water, were excluded from the neural
of rivers and streams. Riparian buffers composed of veg- network training exercise. Urban expansion rates were
etation will trap sediment and other pollutants (cf. Correll based upon research performed by the Michigan Land
1997; Lowrance 1998; Lowrance and others 2000) that are Resource Project, which modeled the entire state of
generated from the land surface and flows towards a Michigan at coarse spatial resolution for the years 2020 and
stream. Trees in riparian zones also help to reduce stream 2040 developing growth rate factors of slow, moderate and
temperature (e.g. Macdonald and others 2003). Additional fast growth based upon observed past changes (Levy
benefits of riparian buffers composed of vegetation include 2001).
the support of movement corridors for birds (Machtans and Urban growth rate represents the ratio of percentage
others 1996). growth of urban area to that of population. The urban area
The coupled model that we present here is one of several percentage change U, is calculated as:
that are being developed by a team of researchers studying u1 u0
U¼ ð2Þ
the dynamics of the Muskegon River Watershed. More u0
complex hydrologic and hydraulic models are being cou-
pled to the Land Transformation Model and the Backcast where u1 is urban in 1998 and u0 is the total urban area in
Model (Wiley and others 2008, 2010). These models 1978. Population percentage change is calculated similarly
incorporate complex processes of overland flow, evapo- as:
transpiration, recharge, stream flow, etc. Impacts to the p1 p0
P¼ ð3Þ
biological community are also being studied (Stevenson p0
and others 2008). We are currently engaging stakeholders where p1 is population in 1998 and p0 is total population in
in various ways across the Muskegon River Watershed in 1978. Thus, the urban growth rate index is the ratio of the
order to enhance their understanding of how the watershed two percentages U and P:
works. The development of Research Bulletins (www.
U
mwrp.net), a web site that is managed by researchers and rate ¼ ð4Þ
P
watershed organizations, helps to promote the results of our
research (Wiley and others 2008). Urban growth rate based upon observed past growth
used a factor of 4.2 for the base (i.e., business as usual)
Acknowledgments We would like to acknowledge funding from scenario and 2.1 for the slow urban growth scenario
the NSF Water Cycle Program, (Grant #WCR 0233648), the NASA
(Table 1).
Land Cover/Use Change and Hydrology Program, NSF III-XT Pro-
gram (Grant #IIS 0705836), the Great Lakes Fishery Trust and the
Wege Foundation. Dave Hyndman and Anthony Kendall provided the Forest Regrowth Modeling for the Future
output from the groundwater travel time model. Kimberly Robinson
read an earlier version of the manuscript and provided useful input.
Forest regrowth was modeled separately and projected
based upon training using 12 spatially-explicit drivers for
forest regeneration in the LTM: distance to forests in 1978;
Appendix 1: Modeling Land Use Changes density of agriculture at 10 km, 25 km and 100 km; den-
sity of forest in 1978 at 10 km, 25 km and 100 km; dis-
Urban Growth into the Future tance from urban in 1978; distance from agriculture;
distance from road; size of non-forest patch; and distance
Five types of urban land classes (residential, commercial from state parks. In the projections, new forests were
services and institutional, industrial, transportation, allowed to grow where the 1998 land cover was either
extractive and recreation and cemeteries) in the underlying agricultural land or shrubs. The resultant map represented
land use vector datasets were collapsed into a single urban the propensity for a given cell to undergo forest regrowth.
land use category for modeling urban changes. Urban Using the output of the forest LTM we transition cells into
growth was projected using six spatially explicit drivers the forest land use class at the rate of forest regrowth that
processed using ArcGIS 9.3 and then trained using the was observed between 1978 and 1998 in the land use
LTM following Pijanowski and others (2002a, b, 2005, datasets. The forest regrowth scenario captured the steady
2007). The inputs were: distance to county roads, distance abandonment of agricultural lands and their return to forest.
to lakes, distance to rivers, distance to existing urban, The rate was forward projected to determine new forest
distance to limited access highways (interstates), and cells at 5 year intervals. Since wetlands and urban areas are
123
Environmental Management
ðP I a Þ2
Q ¼ ð5Þ
ðP I a Þ þ S Table 4 Curve number (CN) values for seven land use classes, across
four hydrologic soil types for the MRW, based on TR-55
where Q is the runoff in inches, P is the rainfall in inches,
Ia is the initial abstraction, capturing all loses before runoff Land use Soil group
begins and S is the maximum retention after runoff begins. A B C D
Substituting:
Urban 87 87 87 87
Ia ¼ 0:2S ð6Þ Agriculture 67 78 85 89
gives: Shrub 55 56 70 77
Forest 30 55 70 77
ðP 0:2SÞ2 Water 0 0 0 0
Q ¼ ð7Þ
ðP þ 0:8SÞ Wetlands 0 0 0 0
where: Barren 77 86 91 94
123
Environmental Management
D for runoff modeling). For each of the 40 subwatersheds Diaz RJ, Rosenberg R (2008) Spreading dead zones and consequences
in the Muskegon River Watershed an area weighted aver- for marine ecosystems. Science 321:926–929
Diebel MW, Maxted JT, Nowak PJ, Vander Zanden MJ (2008)
age curve number value was computed. The area weighted Landscape planning for agricultural nonpoint source pollution
curve number for each of the 40 subwatersheds was cal- reduction I: a geographical allocation framework. Environmental
culated across 100 policy maps, resulting in 400 curve Management 42(5):789–802
number summarizations for the forward projections and Dosskey M (2002) Toward quantifying water pollution abatement
response to installing buffers on crop land. Environmental
600 curve number summarizations for the 15 backcast Management 28(5):577–598
maps across 40 subwatersheds. Area weighted curve Duckles JM (2008) Impacts of land use change on runoff in the
numbers were input into Eq. 7 with a rainfall amount of Muskegon River watershed of Michigan, past, present and
five inches to simulate the runoff for such a storm event. future: a stakeholder driven, scenario-based approach. MS
Thesis, Department of Forestry and Natural Resources, Purdue
University, West Lafayette, Indiana, p 109
EPA (1972) Clean Water Act, Section 319
Feddema JJ, Oleson KW, Bonan GB, Mearns LO, Buja LE, Meehl
GA, Washington WM (2005) The importance of land-cover
References change in simulating future climates. Science 310:1674–1678
Fitzpatrick M, Long D, Pijanowski B (2007) Characterizing biogeo-
Alexander J (2006) The Muskegon: the majesty and tragedy of chemical fingerprints of land use on surface water quality across
Michigan’s rarest river. Michigan State University Press a regional watershed in the Upper Midwest, USA. Applied
Alig RJ, Kline JD, Klein M (2004) Urbanization on the US landscape: Geochemistry 22(8):1825–1840
looking ahead in the 21st century. Landscape and Urban Fohrer N, Haverkamp S, Frede H (2005) Assessment of the effects of
Planning 69:219–234 land use patterns on hydrologic landscape functions: develop-
Allan JD (2004) Landscape and riverscapes: the influence of land use ment of sustainable land use concepts for low mountain range
on river ecosystems. Annual Reviews of Ecology, Evolution and areas. Hydrologic Processes 19:659–672
Systematics 35:257–284 Foster D (1992) Land-use history (1730–1990) and vegetation
Antrop M (2005) Why landscapes of the past are important for the dynamics in central New England, USA. Journal of Ecology
future. Landscape and Urban Planning 70:21–34 80(4):753–771
Bankes S (1992) Exploratory modeling for policy analysis. Opera- Garen DC, Moore DS (2005) Curve number hydrology in water
tions Research 41(3):435–449 quality modeling: uses, abuses and future directions. Journal of
Baker JP, Hulse D, Gregory S, White D, Sickle J, Berger P, Dole D, the American Water Resources Association 41(2):377–388
Schumaker N (2004) Alternative futures for the Willamette Groves DG, Lempert RJ (2007) A new analytic method for finding
River Basin, Oregon. Ecological Applications 14(2):313–324 policy-relevant scenarios. Global Environmental Change
Bhaduri B, Harbor J, Engel B, Grove M (2000) Assessing watershed- 17:73–85
scale, long-term hydrologic impacts of land-use change using a GVSU (2002a) Muskegon River Watershed project: watershed plan,
GIS-NPS Model 2000. Environmental Management 26(6):643– vol I. Annis Water Search Institute, Muskegon, Michigan, p 191
658 GVSU (2002b) Muskegon River Watershed project: management
Booth D, Karr J, Schauman S, Konrad C, Morley S, Larson M, Burges plan appendices, vol II. Annis Water Search Institute, Muskegon,
S (2004) Revising urban streams: land use, hydrology, biology Michigan, p 31
and human behavior. Journal of the American Water Resources GVSU (2002c) Muskegon River Watershed project: education plan
Association 40(5):1351 III. Annis Water Search Institute, Muskegon, Michigan, p 42
Boutt DF, Hyndman DW, Pijanowski BC, Long DT (2001) Identi- Harbor JM (1994) A practical method for estimating the impact of
fying potential land use-derived solute sources to stream land-use change on surface runoff, groundwater. Journal of the
baseflow using ground water models and gis. Ground Water American Planning Association 60(1):95
39(1):24–34 Hundecha Y, Bardossy A (2003) Modeling of the effect of land use
Bren LJ (1998) The geometry of a constant buffer-loading design change on the runoff generation of a river basin through
method for humid watersheds. Forest Ecology and Management parameter regionalization of a watershed model. Journal of
110(1–3):113–125 Hydrology 292(1–4):281–295
Bren LJ (2000) A case study in the use of threshold measure of Johnson NL, Lilja N, Ashby JA (2003) Measuring the impact of user
hydrologic loading in the design of stream buffer strips. Forest participation in agricultural and natural resource management
Ecology and Management 132:243–257 research. Agricultural Systems 78:287–306
Carpenter S, Caraco N, Correll D, Howarth R, Sharpley A, Smith V Koomen E, Stillwell J (2007) Modelling land-use change: theories
(1998) Nonpoint source pollution of surface waters with and methods. GeoJournal Library 90:1–22
phosphorus and nitrogen. Ecological Applications 8(3):559–568 Kim Y, Engel B, Lim K, Larson V, Duncan B (2002) Runoff impacts
Chaubey I, Maringanti C, Schaffer B, Popp J (2008) Targeting versus of land-use change in Indian River Lagoon Watershed. Journal
optimization: critical evaluation of BMP implementation for of Hydrologic Engineering 7(3):245–251
watershed management. In: World Environmental and Water Landis J (1995) Imagining land use futures: applying the California
Resources Congress 2008 urban futures model. Journal of the American Planning Asso-
Cifaldi R, Allan JD, Duh J-D, Brown DB (2003) Spatial patterns in ciation 61:438–457
land cover in exurbanizing watersheds of southeastern Michigan. Lempert R, Popper S, Bankes S (2002) Confronting surprise. Social
Landscape and Urban Planning 66:107–123 Science Computer Review 20(4):420–440
Correll DL (1997) Buffer zones and water quality protection: general Leopold L (1968) Hydrology for urban land planning—a guidebook
principles. In: Haycock N, Burt T, Goulding K, Pinary G (eds) on the hydrologic effects of urban land use. USGS Circular 554
Buffer zones: their processes, potential in water protection. Leopold L (1994) A view of the river. Harvard University Press,
Quest Environmental, Harpenden, UK Cambridge, MA, p 298
123
Environmental Management
Levy D (2001) Michigan land resource project. Technical report. Mesoamerican biological corridor. Global and Planetary Change
Public Sector Consultants, Inc 54:150–162. doi:10.1016/j.gloplacha.2005.09.004
Lowrance R, Altier L, Williams R, Inamdar S, Sheridan J, Bosch D, Ray DK, Pielke RA Sr, Nair US, Welch RM, Lawton RO (2009)
Hubbard R, Thomas D (2000) REMM: the riparian ecosystem Importance of land use versus atmospheric information verified
management model. Journal of Soil and Water Conservation from cloud simulations from a frontier region in Costa Rica.
55(1):27–34 Journal of Geophysical Research. doi:10.1029/2007JD009565
Lowrance R (1998) Riparian forest ecosystems as filters for non-point Ray DK, Pijanowski BC (2010) A backcast land use change model to
source pollution. In: Pace M, Groffmann P (eds) Chapter 5 in generate past land use maps for the Muskegon River Watershed
success, limitations, and frontiers in ecosystem science. Springer of Michigan, USA. Journal of Land Use Science 5:1, 1–29. doi:
Publishers, New York, p 499 10.1080/17474230903150799
Macdonald JS, MacIssac E, Herunter H (2003) The effect of variable Ray DK, Pielke R A Sr, Nair US, Niyogi D (2010) Roles of
retention riparian buffer zones on water temperatures in small atmospheric and land surface data in dynamic regional down-
headwater streams in sub-boreal forest ecosystems of British scaling. Journal of Geophysical Research 115:D05102. doi:
Columbia. Canadian Journal of Forest Research 33:1371–1382 10.1029/2009JD012218
Machtans C, Villard M, Hannon S (1996) Use of riparian buffer strips Rodrguez JP, Beard JTD, Bennett EM, Cumming GS, Cork S, Agard
as movement corridors by forest birds. Conservation Biology J, Dobson AP, Peterson GD (2006) Trade-offs across space,
10(5):1366–1379 time, and ecosystem services. Ecology and Society 11(1), art28
Mitsch WJ, Gosselink JG (1993) Wetlands. Van Nostrand Reinhold, Searchinger T, Heimlich R, Houghton R, Dong F, Elobeid A, Fabiosa
New York J, Tokgoz S, Hayes D, Yu T (2008) Use of US croplands for
Mölders N, Rühaack W (2002) On the impact of explicitly predicted biofuels increases greenhouse gases through emissions from
runoff on the simulated atmospheric response to small-scale land-use change. Science 319:1238–1240
land-use changes—an integrated modeling approach. Atmo- Shields FD, Bowie AJ, Cooper CM (1995) Control of streambank
spheric Research 63:3–38 erosion due to bed degradation with vegetation and structure.
National Research Council (2008) Urban stormwater management in Water Research Bulletin 31:475–489
the United States. The National Academies Press, Washington, Sokal RR, Rohlf FJ (1995) Biometry, 3rd edn. WH Freeman and
DC, 529 pp Company, 887 p
NRCS (2008) US General Soil Map (STATSGO2) Description. Stevenson RJ, Wiley M, Gage S, Lougheed V, Riseng C, Bonnell P,
http://www.soils.usda.gov/survey/geography/statsgo/description. Burton T, Hough R, Hyndman D, Koches J, Long D, Pijanowski
html B, Qi J, Steiman A, Uzarski D (2008) Watershed science:
O’Neal R (1997) Muskegon River Watershed assessment. Fisheries essential, complex, multidisciplinary and collaborative. In:
Division Special Report No. 19. State of Michigan Department Wetland and water resource modeling and assessment: a
of Natural Resources watershed perspective. Taylor and Francis Publishers, Baca
O’Callaghan JO (1996) Land use: the interaction of economics, Raton, Florida
ecology and hydrology. Chapman and Hall Publishers, London Systat Inc (2008) Systat 12, Introduction of linear mixed models,
Paul M, Meyer J (2001) Streams in the urban landscape. Annual chap. 5. Systat Software Inc., San Jose, CA
Review of Ecology and Systematics 32:333–365 Tang Z, Engel B, Lim K, Pijanowski B, Harbor J (2005a) Minimizing
Pijanowski B, Ray D, Kendall A, Duckles J, Hyndman D (2007) the impact of urbanization on long-term runoff. Journal of the
Using backcast land use change and groundwater travel-time Water Resources Association 41(6):1347–1359
models to generate land use legacy maps for watershed Tang Z, Engel BA, Pijanowski BC, Lim KJ (2005b) Forecasting land
management. Ecology and Society 12(2):25 use change and its environmental impact at a watershed scale.
Pijanowski BC (2006) Afforestation patterns in the Upper Midwest, Journal of Environmental Management 76:35–45
USA. In: Proceedings of the IUFRO landscape ecology confer- Tomer M, Burkart M (2003) Long-term effects of nitrogen fertilizer
ence, Locorotondo, Bari, Italy, 26–29 Sept 2006 user on ground water nitrate in two small watersheds. Journal of
Pijanowski BC, Pithadia S, Shellito BA, Alexandridis K (2005) Environmental Quality 32:1931–1938
Calibrating a neural network-based urban change model for two Tong S, Chen W (2002) Modeling the relationship between land use
metropolitan areas of the Upper Midwest of the United States. and surface water quality. Journal of Environmental Manage-
International Journal of Geographical Information Science ment 66(4):377–393
19(2):197–215 United States Department of Agriculture (1986) Tr-55: urban
Pijanowski BC, Shellito B, Pithadia S (2002a) Using artificial neural hydrology for small watersheds. Technical report. Natural
networks, geographic information systems and remote sensing to Resource Conservation Service Conservation Engineering
model urban sprawl in coastal watersheds along eastern Lake Division
Michigan. Lakes and Reservoirs 7:271–285 USEPA (1983) Final peport of the nationwide urban runoff program.
Pijanowski BC, Brown DG, Shellito BA, Manik GA (2002b) Using US Environmental Protection Agency, Water Planning Division,
neural networks and GIS to forecast land use changes: a land Washington, DC
transformation model. Computers, Environment and Urban USEPA (2000) Nutrient criteria technical guidance manual: lakes and
Systems 26(6):553–575 reservoirs, 1st edn. EPA 8222-B00-001. Washington, DC
Ponce VM, Hawkins RH, ASCE M (1996) Runoff curve number: Has Walker B, Salt D (2006) Resilience thinking. Island Press, Washing-
it reached maturity? Journal of Hydrologic Engineering 1:11–19 ton, DC
Pontius R G Jr, Boersma W, Castella J-C, Clarke K, de Nijs T, Dietzel Walter MT, Archibald J, Buchanan B, Dahlke H, Easton Z,
C, Duan Z, Fotsing E, Goldstein N, Kok K, Koomen E, Lippitt Marjerison R, Sharma A, Shaw S (2009) New paradigm for
CD, McConnell W, Sood AM, Pijanowski BC, Pithadia S, sizing riparian buffers to reduce risks of polluted storm water:
Sweeney S, Trung TN, Veldkamp AT, Verburg PH (2008) practical synthesis. Journal of Irrigation and Drainage Engineer-
Comparing the input, output, and validation maps for several ing 135(2):200–2009
models of land change. Annals of Regional Science 42(1):11–47 Wang L, Lyons J, Kanehl P, Gatti R (1997) Influences of watershed
Ray DK, Welch RM, Lawton RO, Nair US (2006) Dry season clouds land use on habitat quality and biotic integrity in Wisconsin
and rainfall in northern central America: implications for the streams. Fisheries 22(6):6–12
123
Environmental Management
Wayland KG, Hyndman DW, Boutt D, Pijanowski BC, Long DT the Muskegon River: tools for ecological risk assessment in a
(2002) Modelling the impact of historical land uses on surface- Great Lakes Watershed. In: Ji W (ed) Wetland and water
water quality using groundwater flow and solute-transport resource modeling and assessment: a watershed perspective,
models. Lakes and Reservoirs 7:189–199 chap 20. Taylor & Francis, London
Wayland KG, Long DT, Hyndman DW, Pijanowski BC, Woodhams Wiley M, Hyndman D, Pijanowski BC, Kendall A, Riseng C,
SM, Haack SK (2003) Identifying relationships between base- Rutherford E, Cheng S, Carlson M, Tyler J, Stevenson R, Steen
flow geochemistry and land use with synoptic sampling and R- P, Richards P, Seelbach P, Koches J (2010) A multi-modeling
mode factor analysis. Journal of Environmental Quality approach to evaluate impacts of global change on river
32(1):180–190 ecosystems. Hydrobiologia. doi:10.1007/s10750-010-0239-2
Whitney GG (1987) An ecological history of the great lakes forest of Wolter PT, Johnston CA, Niemi GJ (2006) Land use and cover
Michigan. The Journal of Ecology 75(3):667–684 change in the U.S. Great Lakes basin 1992–2001. Journal of
Wiley M, Pijanowski B, Jan Stevenson R, Seelbach P, Richards P, Great Lakes Research 32:607–628
Riseng C, Hynman D, Koches J (2008) Integrated modeling of
123