LAB EXERCISE #5 Landscape Dynamics: Alternative Scenarios

Transcription

1 LAB EXERCISE #5 Landscape Dynamics: Alternative Scenarios Instructors: K. McGarigal Overview: In this exercise, students will use the results of RMLands to quantify the dynamics in landscape structure for a sample landscape under several alternative management scenarios representing a factorial of climate scenarios, level of fire suppression and intensity of vegetation management. Students will gain a practical understanding of scenario analysis as an aid to forest planning and gain an appreciation for the challenges and limitations associated with the interpretation of results. Specifically, students will learn how to conduct scenario analysis to compare the potential impacts of alternative management strategies aimed at controlling fire and insect/pathogen disturbance regimes. Objectives To develop a practical understanding of the RMLands model framework, as an example of a Landscape Disturbance-Succession Model (LDSM). To gain practical understanding of how to use simulation modeling to evaluate alternative land management scenarios for strategic assessment and planning. To gain practical understanding of the interactions among climate, fire suppression and vegetation management and the implications for ecosystem management. To gain practical experience conducting scenario analysis. Model Background In the lectures associated with this lab exercise, we gained an overall familiarity with HRV concepts, described the use of landscape disturbance-succession models (LDSMs) to quantify HRV, and discussed some of the practical challenges in using HRV to inform land management. In this lab, we are going to use the results of the Rocky Mountain Landscape Simulator (RMLands) to quantify the range of variability in landscape structure for a sample landscape in the Bitterroot Mountains of western Montana. For practical reasons it is beyond the scope of this lab to describe the details of RMLands and its parameterization for the case study landscape. Instead, a brief conceptual overview of RMLands will have to suffice. In addition, because we are using an LDSM to estimate the range of variability for the study landscape, we are going to refer to it as the simulated range of variability or SRV to make it explicit that the result is from a simulation. Lab. 5 Pg. 1

2 Conceptual Overview RMLands is a grid-based, spatially-explicit, stochastic landscape simulation model designed to simulate disturbance and succession processes affecting the structure and dynamics of Rocky Mountain landscapes. RMLands simulates two key processes: succession and disturbance. These processes are fully specified by the user (i.e., via model parameterization) and are implemented sequentially within 10-year time steps for a user-specified period of time. Succession occurs at the beginning of each time step in the simulation and represents the gradual growth and/or development of vegetative communities over time. Succession is implemented using a stochastic state-based transition approach in which vegetation cover types transition probabilistically between discrete states (conditions). Transition pathways and rates of transition between states are defined uniquely for each cover type and are conditional on several attributes of a vegetation patch. These patches, as defined for succession, represent spatially contiguous cells having the same cell attributes (e.g., identical disturbance history and age). Most cover types progress through a series of stand conditions (states) over time as a result of successional processes (albeit at different rates due to the stochastic nature of succession). In some cases, these transitions are affected by the occurrence of certain disturbances (e.g., low-severity fire) or are regulated by management (e.g., silviculture). Other cover types (e.g., meadows, barren, water) are treated as having a single, static condition and are not affected over time by the interplay of disturbance and succession. Lab. 5 Pg. 2

3 Model Characteristics RMLands is stand-alone program written entirely in Visual C++ for use in a Microsoft Windows Operating System environment. RMLands expects input grids in Arc Grid (ESRI) format and requires libraries from either ArcGIS or ArcView Spatial Analyst. RMLands was designed to make full use of required Forest Service data, supported by systems such as NRIS, INFRA, and FACTS. Importantly, it does not require detailed stand inventory data. RMLands can be classified as a hybrid statistical/probabilistic model with the following distinguishing characteristics: Grid-based.--RMLands utilizes a gridbased data model in which the landscape is represented in a regular grid lattice structure. The grid structure allows for efficient and powerful spatial processing. Each grid cell (pixel), representing a fixed geographic area, possesses a number of ecological attributes (e.g., cover type, age). Attributes possess multiple states (i.e., unique values), many of which change over time in response to succession and disturbance. Spatially explicit.--consistent with the grid structure, RMLands is a spatially-explicit model; grid cells are geographically explicit and topological relationships are important in all processes (e.g., disturbance initiation and spread). Process-based. RMLands simulates two key processes: disturbance and succession. Disturbance processes include a variety of both natural and anthropogenic disturbances implemented in a common fashion. Succession is based on a discrete state transition model for each cover type. Stochastic.--RMLands is a stochastic model; that is, there is an element of chance (or uncertainty) associated with the outcome of each process. For example, each cell has a probability of initiation for each disturbance process that is contingent on several cell attributes. A probability less than 1 means that there is only a chance of a disturbance initiating. Thus, given the same cell attributes, some cells will initiate while others will not. There is a stochastic element to nearly all processes in RMLands. Spatial scale.--the grid can be defined at any spatial resolution, although current applications utilize a relatively high resolution (25-30 m cell size) grid that allows for detailed representation of landscape patterns. In addition, while RMLands does not limit the extent of the landscape, it is most applicable to landscapes between 10,000's ha to over 1 million ha. Lab. 5 Pg. 3

4 Temporal scale. RMLands currently operates on a 10-yr time step and is most applicable to simulating landscape dynamics over 100's to 1000's of years. GIS Database RMLands is designed to work in concert with an ArcGIS grid database, although the relationship consists solely of input/output. On the input side, RMLands requires at a minimum 9 separate Arc grids; 19 additional optional input grids can be used as well depending on the application. On the output side, RMLands produces a large number of optional grids, including 62 specific grids plus an unlimited number of additional grids based on userspecified reclassifications and/or rescalings of the cover-condition grid. A complete description of these grids is included in a separate document (see...\exercises\rmlands\parameters\rmlands_s patial_data.pdf). Lab. 5 Pg. 4

5 Succession RMLands simulates succession using a simple state-based transition approach in which discrete vegetation states are defined for each cover type. Succession involves the probabilistic transition from one state to another over time and it occurs at the beginning of each time step in response to gradual growth and development of vegetation over time. Transition probabilities are typically based on the age of the stand (i.e., the time since the last stand-replacing event), but they can be based on any number of parameters such as the abiotic setting or disturbance history. Each cover type has a separate transition model that uniquely defines its successional stages. In contrast to disturbance, which is implemented at the cell level, succession is entirely patch based. Specifically, each cell belongs to a temporary patch defined as contiguous (touching based on the 8-neighbor rule) cells sharing the same values for each of the attributes used to define succession probabilities. For example, if in a particular cover type transition model, age, time since lowmortality fire and aspect are all used to define transition probabilities, then contiguous cells with the same values for these three attributes will be treated as a patch and undergo succession together. Note, successional patches are not static; they change over time in response to disturbance events, which can act both to break up single patches into several new patches and to coalesce several patches into a single patch by changing the disturbance history at the cell level. This patch-based approach for succession is necessary to avoid the salt-and-pepper effect of cell-based succession given that succession is implemented as a stochastic process. Disturbance Processes RMLands simulates a variety of natural and anthropogenic disturbances. Natural disturbances are simulated with a generic disturbance process that can be parameterized for a wide variety of disturbance agents, includeing wildfire, a variety of insects/pathogens (pinyon decline [pinyon ips beetle and black stain root rot], pine beetle complex, Douglas-fir beetle, spruce beetle, and spruce budworm), and drought. Each natural disturbance process is implemented separately, but effects and is effected by other disturbance processes operating concurrently to produce changes in landscape conditions. For example, the occurrence of beetle-killed trees derived from the spruce beetle disturbance process can affect the local probability of ignition and spread of wildfire. Lab. 5 Pg. 5

6 Each natural disturbance is modeled as a stochastic process; that is, there is an element of chance (or uncertainty) associated with the initiation, spread, and ecological effects of the disturbance. The disturbance algorithm is common among all natural disturbance processes, however, it is parameterized differently for each disturbance agent, and consists of the following key components: # Climate. The climate plays a significant role in determining the temporal and spatial characteristics of the disturbance regime. Climate is specified as a global parameter that optionally can effect initiation, spread, and mortality of all disturbances within a time step. Climate can be specified as constant with a userspecified level of temporal variability, a trend over time (with variability), or as a user-defined trajectory - perhaps reflecting the climate conditions during a specific reference period (e.g., as indexed using the Palmer Drought Severity Index). # Initiation. Disturbance events are initiated at the cell level. Each cell has a probability of initiation in each time step that is a function of its susceptibility to disturbance and, optionally, its proximity to other disturbance events or landscape features (e.g., roads). Susceptibility to wildfire, for example, is a function of cover type, stand condition, time since last fire, time since last insect outbreak, elevation, aspect, slope, and road proximity - factors that influence fuel mass and moisture and risk of humancaused ignition. Lab. 5 Pg. 6

7 # Spread. Once initiated, the disturbance spreads to adjacent cells in a probabilistic fashion. Each cell has a probability of spread that is a function of its susceptibility to disturbance (as above), which is modified by its relative position (e.g., relative elevation or wind direction) and the influence of potential barriers (e.g., roads and streams). In addition, there is an optional provision for the spotting of disturbances during spread so that disturbances are not constrained to contiguous spread only. # Termination. The spread is terminated based on a user-specified disturbance size distribution intended to reflect variable weather conditions associated with the disturbance event that may cause a disturbance to terminate despite otherwise favorable fuel conditions. # Mortality. Following spread, each cell is evaluated to determine the magnitude of ecological effect (i.e., severity) of the disturbance. Each cell can exhibit high or low mortality of the dominant plants. High mortality occurs when all or nearly all (>75%) of the dominant plant individuals are killed. Cells are aggregated into vegetation patches for purposes of determining mortality response, where patches are defined as spatially contiguous cells having the same cell attributes (e.g., identical disturbance history and age). # Transition. Following mortality determination, each mortality vegetation patch is evaluated for potential immediate transition to a new stand condition (state). Transition pathways and rates of transition between states are defined uniquely for each cover type and are conditional on several attributes at the patch level. Note, these disturbance-induced transitions are differentiated from the successional transitions that occur at the beginning of each time step in response to gradual growth and development of vegetation over time. Lab. 5 Pg. 7

8 RMLands simulates a variety of vegetation treatments. Treatments are implemented via management regimes defined by the user. Management regimes are uniquely specified within management zones, or user-defined geographic units (e.g., urban-wildland interface verus interior). Management zones are further divided into one or more management types based on cover type. Each cover type can be treated separately or it can be combined with other cover types to form aggregate management types. Each management type is then given a unique management regime, which consists of one or more treatment types and associated spatial and temporal constraints. Specifically, each management regime is defined by the following components: # Management zones. Management zones are simply optional user-defined geographic units (e.g., urban-wildland interface verus interior) within which to specify unique management regimes. The geographic units can represent anything and there can be an unlimited number of zones. Lab. 5 Pg. 8

9 # Management types. Management zones can be subdivided into management types based on aggregations of cover types. By default each cover type within a management zone is considered separately and optionally assigned a treatment regime. However, cover types can be aggregated into management types if they are to receive the same treatment regime and you want to essentially ignore the cover type boundaries during treatment unit layout. # Treatment types and allocation. Treatment intensity within a management regime is controlled by treatment area; i.e., within each time step treatments are implemented subject to the availability of suitable area and a maximum treatment area constraint. In addition, treatment intensity is also optionally subject to restriction based on user-specified watershed constraints. Specifically, once a watershed exceeds a specified disturbance threshold, all further treatments can be prohibited in that watershed. The watershed disturbance threshold is defined in terms of clearcut equivalence and is designed to reflect the impact of disturbances (both natural and anthropogenic) on water resources. Each combination of cover type and disturbance type is given a clearcutequivalent coefficient and recovery trajectory over time. Treatments available for inclusion in a management regime include a variety of silvicultural systems associated with commercial timber harvest and fuels treatment. Treatments included in a management regime are implemented according to an allocation scheme in which a specified proportion of the total treatment area is allocated to each treatment type. Lab. 5 Pg. 9

10 Treatment types currently include the following (images not included):! clearcut single entry, regeneration cut removing >90% of the canopy and returning the stand to early seral.! shelterwood three-stage shelterwood; first entry is prep cut (no change in vegetation state); second entry is seed cut in which the bulk of the overstory is removed (change to opencanopy state) and regeneration is initiated; third entry is final cut in which the remaining overstory is removed and the state transitions to early seral. The user-specified treatment interval determines the number of years between entries.! group cut similar to clearcut, except implemented in randomly-sized patches within a userspecified percentage of the treatment unit (i.e., treatment intensity) and harvested over time according to the user-specified treatment intensity. For example, if the treatment period is 100 years and the treatment interval is 20 years and the treatment intensity is 10%, then 10% of the unit will get harvested in randomly sized patches every 20 years over a period of 100 years.! thinning single-tree selection designed to maintain an uneven-age stand structure and promote continuous regeneration. Thinning involves both overstory and understory removal. A user-specified percentage of the treatment unit is thinned at the specified treatment interval over the treatment period.! mastication mechanical mastication of understory vegetation (small diameter stems), primarily for the purpose of compacting and distributing the understory fuels to reduce the likelihood of severe fire, within a user-specified percentage of the treatment unit (i.e., treatment intensity). This has no affect on the overstory of woodlands and forests (i.e., it does not change the vegetation state from closed to open canopy), but it can be used to remove shrubland vegetation.! prescribed burning prescribed fire designed to be predominantly non-lethal surface fire to reduce fine fuels, but allowing for some lethal surface or crown fire in small patches designed to open up the stand.! matrix thin and group cut combined group cuts within a user-specified percentage of the treatment unit (i.e., treatment intensity) and overstory thinning of the rest of the unit.! thin and burn combined overstory thinning within a user-specified percentage of the treatment unit (i.e., treatment intensity) followed by prescribed fire over the entire unit aimed at reducing fuels and maintaining an open-canopy condition over time.! masticate and burn combined understory mastication of small diameter material within a user-specified proportion of the unit (i.e., treatment intensity) followed by prescribed fire over the entire unit.! hand cut, pile and burn combined hand cutting and piling of small diameter material within a user-specified percentage of the unit (i.e., treatment intensity) followed by prescribed fire Lab. 5 Pg. 10

11 over the entire unit.! thin and masticate combined understory mastication of small diameter material within a userspecified percentage of the unit (i.e., treatment intensity) and overstory thinning of the entire unit.! matrix thin, group cut and burn combined group cuts within a user-specified percentage of the unit (i.e., treatment intensity) and overstory thinning of the rest of the unit, followed by prescribed fire over the entire unit.! thin, masticate and burn combined mastication of small diameter material within a userspecified percentage of the unit (i.e., treatment intensity) and overstory thinning of the entire unit, followed by prescribed fire over the entire unit.! thin, hand cut, pile and burn combined hand cutting and piling of small diameter material within a user-specified percentage of the unit (i.e., treatment intensity) and overstory thinning of the entire unit, followed by prescribed fire over the entire unit. # Static spatial constraints and priorities. For each treatment type in a management regime, static spatial constraints limit where treatment is allowed. Spatial constraints can be defined on the basis of several factors, including ownership, timber suitability, roadless areas, riparian buffer zones, fire management zones, road proximity, and slope. Each of these factors alone or in combination can restrict the potential treatment area. In addition, the initiation of treatment units within the potential treatment area can be prioritized based on these factors. These spatial constraints and priorities are static; that is, they do not change over the course of the simulation. # Dynamic suitability constraints. For each treatment type in a management regime, dynamic constraints limit where treatment is allowed in any particular time step based on vegetation characteristics that change over time. Dynamic constraints can be defined on the basis of stand condition class (i.e., seral stage), stand age, or any other agerelated attribute (e.g., age since last low Lab. 5 Pg. 11

12 mortality fire). These constraints are dynamic because they vary spatially over the course of the simulation. The static and dynamic constraints and priorities layers are combined in each timestep using a geometric mean to determine the actual constraints and priorities within each timestep. In this manner, the static constraints and priorities have an overarching and constant effect on the constraints and priorities and the dynamic constraints and priorities modify them based on the current condition of the vegetation. The combined constraints and priorities determine the probability of treatment initiation and spread of the treatment units as they are created from the points of initiation. Importantly, the combined constraints and priorities layer changes over time due to changing vegetation conditions, and it differs among treatment types within the same timestep depending on the userspecifications. Lab. 5 Pg. 12

13 # Treatment units. Treatment units are derived stochastically within each management zone and management type according to the targeted treated area and allocation among treatment types, subject to the constraints and priorities above, and the following parameters:! Initiation and spread individual units are created from a point of initiation within the corresponding management zone and management type according to the corresponding constraints and priorities, and then the unit spreads outward from the point of initiation based on the resistance conferred by the constraints and priorities layer such that the unit will tend to grow through areas of higher priority and where constraints represent absolute barriers to spread. There is also an optional boundaries layer that can impose additional constraints on spread. The unit stops growing when there is no further suitable area to spread into (i.e., a constraint or boundary) or a user-specified maximum unit size is achieved. The final unit must also exceed a user-specified minimum unit size as well.! Minimum canopy cover constraint this is an optional user-specified minimum canopy cover constraint that must be maintained or the unit is not implemented. Specifically, the potential unit is created and treated to determine the resulting average canopy cover, and if this falls below the average canopy cover as specified in a user-defined minimum canopy cover layer, then the unit is discarded (i.e., not treated). Lab. 5 Pg. 13

14 ! Unit dispersion this is the manner in which treatment units should be spatially dispersed across the landscape. There are four options: (1) random - units are randomly distributed, (2) aggregated within watersheds - units are randomly distributed within user-specified watersheds, (3) aggregated within compartments - units are randomly distributed within user-specified compartments, and (4), aggregated by distance units are aggregated within a user-specified distance of the first unit.! Unit buffers and fallow periods this is an optional spatial buffer zone (in meters) around a treatment unit in which no other treatments are allowed for the duration of the designated fallow period. The fallow period is only meaningful if a buffer zone is designated.! Treatment regime each treatment type within a management zone and management type is implemented according to a treatment regime (or prescription) that includes a treatment period, treatment interval, and treatment intensity. The treatment period is the period (in years) under which the treatment unit stays under management and remains inviolate by other treatments. At the end of the treatment period the unit is released and the cells are free to be selected and incorporated into a new unit. The treatment interval is the interval (in years) between individual treatments (cutting or burning) or stand entries for all treatment types that involve periodic entries (see above). Lastly, treatment intensity is the percentage of the treatment unit to be treated with one of the treatments included in the treatment type (see above). For example, for group cut it represents the percentage of the unit to be included in group cuts. Lab. 5 Pg. 14

15 ! Vegetation transitions the last step of the treatment unit implementation is the vegetation transitions. Based on user-specified disturbance transition rules, each cell potentially undergoes a state transition (i.e., change to a different seral stage). In summary, the vegetation treatment module in RMLands is quite complex, which provides great flexibility for specifying management scenarios. A single management scenario involves specifying one or more management zones (i.e., geographic units), which are further subdivided into one or more management types based on aggregations of cover types. Within each management type, a total treatment intensity (i.e., maximum total treatment area) is specified and allocated among one or more treatment types. A treatment regime is specified for each treatment type. The treatment regime has many components, but includes specifying static spatial constraints and priorities (e.g., based on timberland suitability, road proximity, ownership, etc.) dynamic suitability constraints (e.g., based on seral stage, age and disturbance history), constraints on treatment unit size, dispersion (i.e., random, aggregated or dispersed), adjacency (i.e., buffer width and fallow period), and temporal attributes associated with rotation period and treatment interval. Lab. 5 Pg. 15

16 Detailed Instructions Step 1. Establish the objective of the analysis The first step (as always) is to establish the objective of the analysis. Our overall objective is to evaluate the impact of alternative land management strategies on landscape dynamics and, more specifically, to assess the interaction of climate, fire suppression and vegetation management. Here, we will focus on the following specific question: How do climate, fire suppression and vegetation management potentially interact to effect landscape dynamics, and what are the potential ecological and socio-economic consequences of these interactions? Step 2. Define the digital landscape The next step is to define the landscape. First, let s get familiar with the sample landscape and the associated GIS data. The sample landscape is the 47,058 ha Prospect Creek watershed in the Bitterroot Mountains of western Montana. The landscape is mostly public land managed by the Lolo National Forest, although there is some private land holdings in the lower elevation valleys. This landscape supports a wide range of environmental gradients producing a forest diverse in vegetation and disturbance processes, and is part of a much larger case study landscape encompassing the Bitterroot Mountains. Open up in GoogleEarth the project file...\exercises\scenarios\scenarios.kml Take some time to survey the study area in GoogleEarth. Next, let s define the digital landscape. To meet the objective above (and for purposes of this lab exercise), we defined the landscape as follows: We selected a single sample landscape from the broader case study landscape in the Bitterroot Mountains based on the following criteria: 1) landscape extent large enough to incorporate meaningful landscape dynamics given the scale of the major disturbance processes, yet small enough to be computationally efficient for lab use, 2) a heterogeneous mixture of land use practices, including developed lands with a wildland-urban interface, a mixture of public and private lands dominated by the former, and an adequate road network to facilitate future vegetation treatments, and 3) a logical ecological unit, in this case, a watershed, meeting the above criteria. We classified the sample landscape into land cover classes based on the LANDFIRE project. Specifically, land cover classes represent unique biophysical settings (BpS) or potential vegetation types (PVT). See...\exercises\scenarios\Z10_BpS_Model_Descriptions.pdf for the descriptions of each BpS class. The only significant change we made to this classification scheme was to combine three separate BpS classes corresponding to riparian settings into a single riparian class. Note, it is beyond the scope of this exercise to describe how these BpS s classes were derived, but full documentation is available at the LANDFIRE website. Note, not all BpS classes found in the case study landscape are present in the sample landscape. See...\exercises\scenarios\composition.xls for the land cover composition of Prospect Creek Lab. 5 Pg. 16

17 basin in comparison to the case study landscape. As noted above, the spatial scale of the sample landscape was established in part to meet computational efficiencies for the lab, thus it is smaller than we might otherwise prefer. The spatial grain (or resolution) of the landscape was set at 30 m, consistent with the spatial resolution of the data sources used in the LANDFIRE project. The spatial extent of the landscape was based on the hydrological watershed of Prospect Creek, a tributary of Clark Fork River; however, for simulation purposes we included a 2-km wide buffer zone around the basin, bringing the total extent of the simulation landscape to 69,293 ha. A wealth of GIS data is available for this landscape. Open up in ArcMap the project file...\exercises\scenarios\scenarios.mxd Take some time to review each of layer for the purpose of familiarizing yourself with the landscape the instructor will guide you through this process. Most of these layers are used by RMLands in the simulation. See the document entitled...\exercises\scenarios\ RMLands_spatial_data.pdf for a detailed description of each layer. The single most important layer is the cover grid, since it establishes a static vegetation template within which vegetation seral stages change over time in response to disturbance and succession processes. This is a land cover GRID at 30 m spatial resolution classified into biophysical settings (BpS) from a variety of remotely sensed data sources, including LANDSAT images, terrain variables and meteorological variables. Note, this data layer is based principally on the BpS layer produced by the LANDFIRE project, although it has been modified slightly here to also include current development (e.g., agriculture and urban), large roads (i.e., highways), and large rivers. In addition, three separate BpS classes corresponding to riparian settings have been combined into a single riparian class. See the document entitled...\exercises\scenarios\z10_bps_model_descriptions.pdf for the descriptions of each BpS class. Questions to ponder: 2.1 What are the tradeoffs and/or limitations imposed by the chosen landscape definition? Is the thematic content and resolution of the landscape appropriate for the stated objective? How would you modify the thematic content and resolution and why? Is the spatial extent and resolution appropriate for the stated objective? How would you modify the spatial scale of the analysis to better meet the stated objective and why? Step 3. Design the simulation experiment (i.e., alternative scenarios) The next step is to set up an experimental design to evaluate the interaction among climate, fire suppression and vegetation management. While there are numerous possibilities, for the purpose of this exercise we established an incomplete 2x2x2 factorial of climate, fire suppression and vegetation management, as follows: Climate Scenario Fire Suppression Vegetation Management Lab. 5 Pg. 17

18 No Treatment Historic No hc-letburn-notreat (session.id=5) Yes hc-suppress-notreat (session.id=6) Future No fc-letburn-notreat (session.id=7) Yes fc-suppress-notreat (session.id=8) Vegetation Treatments none none none fc-suppress-vegtreat (session.id=1) We created these scenarios (models) in advance based on the criteria below; unfortunately, the detailed model parameterization for each scenario can only be examined by having RMLands installed on your computer and loading the models into RMLands (see...\exercises\scenarios\models\model.name from the table above), which is likely to be beyond the scope of this exercise (but see the instructor if you want to try to install the software). Climate Scenarios We created two climate scenarios representing a contrast between historic and potential future climate conditions, as follows: Historic Climate. In this scenario, we parameterized the climate modifier in RMLands for the various disturbance processes as follows: Wildfire based on the historic record as represented by the mean Palmer Drought Severity Index, averaged over 5 sample locations in the vicinity of the case study landscape for each 10-year interval for the period (note, this sequence was repeated to create the 1,000 year time series needed for the simulation). Pine beetle based on the historic record as represented by the cumulative threshold Palmer Drought Severity Index, averaged over 5 sample locations in the vicinity of the case study landscape for each 10-year interval for the period , as above. Note, the cumulative threshold PDSI is based on the maximum cumulative consecutive years of drought within each 10-year interval, but is thresholded so that timesteps with an index < 1 are set to 0, preventing any disturbances from occurring. This results in periodic or episodic outbreaks (or epidemics) against a background of endemic levels of disturbance. Future Climate. In this scenario, we parameterized the climate modifier as above except that the mean climate value was increased by 10% (from 1 to 1.1). Note, there are many possible alternative future climate scenarios. This particular scenario represents the case in which the frequency and severity of drought conditions conducive to burning and bark beetle outbreaks is increased by 10%. Lab. 5 Pg. 18

19 Fire Suppression We created two fire suppression scenarios representing a contrast between a let burn policy and an aggressive fire suppression policy, as follows: Let burn. In this scenario, we employed a let burn policy and simulated no fire suppression. In this case, no suppression means that the frequency, size and severity of wildfires is based on the HRV disturbance regime. Suppress. In this scenario, we employed an aggressive fire suppression policy. Unfortunately, the best way to emulate fire suppression effects is not entirely clear; there are many possibilities. For the purpose of this exercise, we assumed that fire suppression per se does not change the frequency of ignitions or the severity of fires, although indirectly it will likely increase both over the long term if the vegetation becomes more susceptible with age. Instead, we assumed that fire suppression directly effects the probability of fire spread, and thus directly influences the distribution of fire sizes. To emulate this effect, we modified the size distribution of fires in the spread parameters for wildfire as follows: Size (ha) Letburn Percent of Fires Suppress , , , Note, the changes here are designed to emulate a fire suppression policy that is reasonably but not perfectly effective in preventing the spread of fires. Consequently, while the distribution is considerably compressed to the left (more smaller fires), large fires (including the maximum fire size) are still simulated under the suppression scenario, but with a much reduced probability. One could certainly imagine other possible scenarios. Vegetation Management We created two vegetation management scenarios representing a contrast between no treatment and aggressive vegetation treatment, as follows: No treatment. In this scenario, we employed a do nothing policy and simulated no active vegetation management. Vegetation treatment. In this scenario, we employed an aggressive vegetation management Lab. 5 Pg. 19

20 policy. There are myriad options for vegetation treatments in RMLands; however, in the context of this lab it is impractical to consider more than one vegetation treatment scenario. Thus, we attempted to emulate the current National Forest management focus on: (1) fuels reduction in the wildland-urban interface (WUI) and (2) salvage of timber following large-scale disturbance events. Other current management objectives, such as ecosystem restoration and timber stand improvement, were considered as secondary objectives and were addressed only indirectly as byproducts of the vegetation treatments aimed at the primary objectives. The parameterization of the vegetation treatments is quite complex and requires considerable understanding of RMLands. Rather than try to describe the detailed parameterization, a summary of the important distinctions are give below: (1) Wildland-urban interface (WUI) zone [MZ-1 in the output]: Objective. WUI treatments were designed primarily to reduce fuels, and thus reduce the risk of high severity fire and improve the likelihood of effective fire suppression, and secondarily to restore ecosystem structure and function to something akin to the historic reference condition. Spatial constraints and priorities. Treatments were excluded from private lands, unsuitable timberland (as designated), riparian zones, and roadless areas. All other lands were considered eligible for treatments. Intensity. The goal was to treat all eligible lands with the WUI, which amounts to approximately 15,000 ha. Recognizing that even under the best circumstances, it is highly unlikely or even undesirable to get 100% of the eligible land under treatment, we instead targeted 12,000 ha, with a target of 3,000 ha of land treated per decade on a 40-year treatment interval. Thus, the goal was to get up to 12,000 ha of land treated every 40 years. However, there are numerous factors conspiring against meeting this target. For example, if we target closed-canopy forest conditions for treatment (which we did), it is likely that following extensive low-mortality wildfire (which by and large converts closed-canopy forest to open-canopy forest), there may considerably less area suitable for treatment or in need of treatment no need to treat lands that mother nature treats for us. In addition, if we constrain treatments to sufficiently large contiguous areas of eligible lands containing suitable forest conditions for logistical and economic reasons, there will be locations and times when patches of eligible forest are simply too small and too scattered for efficient treatment. Thus, the targeted maximum treatment area per timestep should not be interpreted as a hard target to be met at all costs, but rather as a flexible target that varies depending on the vegetation conditions and which is probably never met due to numerous other constraints. Treatment regime. Two silvicultural treatments were employed: (1) restoration treatments, which involves the combination of individual tree removal (i.e., basal area reduction) and prescribed underburning (i.e., low mortality fire); and (2) individual tree selection, which involves individual tree removal without under burning. Treatment units were distributed in aggregated fashion in units of ha. Lab. 5 Pg. 20

21 (2) Non-wildland-urban interface (non-wui) zone [MZ-2 in the ouput]: Objective. Non-WUI treatments were designed primarily to salvage timber following major wildfires and insect outbreaks. Spatial constraints and priorities. Treatments were excluded from private lands, unsuitable timberland (as designated), riparian zones, and roadless areas. All other lands were considered eligible for treatments. Intensity. Given spatial constraints above, approximately 17,000 ha of land are eligible for treatments in the non-wui zone. The goal was to salvage up to a maximum of 2,000 ha in any decade experiencing extensive wildfires and/or insect outbreaks. Treatment regime. The basic silvicultural treatments included a combination of clearcut and individual tree selection in RMLands. Both treatments were single entry treatments without follow-up. Treatment units were distributed in aggregated fashion in units of 4-40 ha for clearcut and ha for individual tree selection. Questions to ponder: 3.1 Given your current understanding of RMLands, and Landscape Disturbance-Succession Models in general, what would be an alternative strategy for simulating climate change that is significantly different than the one implemented in this experiment? 3.2 Similarly, what would be an alternative strategy for simulating fire suppression that is significantly different than the one implemented in this experiment? 3.3 Similarly, what would be an alternative strategy for simulating vegetation treatments that is significantly different than the one implemented in this experiment? Step 4. Run the simulations and quantify landscape structure The next step is to run the simulations and quantify the structure of the simulated landscape trajectories. Given the limitation of this lab, it is not feasible or that important for you to learn how to parameterize and run RMLands. Moreover, it is time-consuming and somewhat tedious to run these eight simulations and analyze the output using Fragstast. Thus, we have already done this for you in advance, as follows: RMLands simulations. Each scenario was run for 2,000 years. The cover-condition (covcond) output grids plus a suite of grids associated with wildfires and pine beetles were saved in a separate folder for each scenario (...\exercises\scenarios\results\scenario). FRAGSTATS analysis. The cover-condition (covcond) grids output from each scenario were analyzed with FRAGSTATS. For the purposes of this exercise, we chose a broad suite of metrics encompassing a variety aspects of landscape pattern, including both structural and functional metrics at the class and landscape level. For the functional metrics, we applied the edgedepth and edge contrast weights constructed based on expert opinion. All the files used to complete the FRAGSTATS analysis are included in...\exercises\scenarios\fragstats. At the class level, we enabled the classes corresponding to two cover types: (1) mixed-conifer forestponderosa pine-douglas-fir and (2) mesic-wet spruce-fir forest and woodland. The Lab. 5 Pg. 21

22 FRAGSTATS output files have been saved and are included in...\exercises\scenarios\results. The landscape and class metrics are included in fragout.land and fragout.class, respectively. Questions to ponder: 4.1 Given your current understanding of RMLands and the disturbance scenarios implemented in this experiment, what is your prediction for the relative effect of future climate on landscape structure? Specifically, which aspects of landscape structure will be effected and how will they be effected? 4.2 Similarly, what is your prediction for the effect of fire suppression on landscape structure? Specifically, which aspects of landscape structure will be effected and how will they be effected? 4.3 Similarly, what is your prediction for the relative effect of your vegetation treatment scenario on landscape structure? Specifically, which aspects of landscape structure will be effected and how will they be effected? Step 5. Examine the results The final step is to examine the results. There are numerous ways to examine the results, graphically and statistically, as follows. 5.1 View simulation grids The first thing we can do is visually examine the simulation output grids. This can be a challenging and time-consuming task due to the many different output grids available. Here, we will focus on examining only a small subset of the spatial data layers produced by the simulation. (1) Perhaps the easiest way to view the simulation is as a slide show. Here, we will view a slide show of the simulation based on the cover-condition, wildfire mortality, and mountain pine beetle mortality grids in which each slide represents a 10-year time step. To view a slide show of any of these output grids for any of the scenarios, simply navigate to the desired folder containing images (e.g,....\exercises\scenarios\results\movies\fc-letburn-notreat\covcond\ and right click on the first image and choose preview. With any luck, it will open with Windows Picture and Fax Viewer and you can advance through the slides at your own pace: (2) A second more tedious option is to view the output grids themselves in ArcMap. While not necessary, if you want to view the grids in this way, open up the ArcMap project (...\exercises\scenarios\scenarios.mxd) and load any of the stored output grids in the corresponding results folders. 5.2 Summarize results using R The second thing we can do is examine the results statistically. For this purpose, we will use the R language and statistical computing environment (note, it is assumed that you have pre-installed the R software). Note, detailed instruction on the use of R is beyond the scope of this exercise, so the instructor will guide you through the following steps. Open the R interface: Start Programs R R (or higher) For the series of R functions used in this exercise, you can execute these functions from the script provided. To open the provided R script, do the following: Lab. 5 Pg. 22

23 From the R console: Select open script from the file drop-down menu: File Open script Navigate to and open the following file:...\exercises\scenarios\scripts\rmlstats.calls.r To execute a line of the script, do either of the following: With the cursor on the desired line, enter control R from the keyboard With the cursor on the desired line, right click mouse and select run line or selection Now let s look at the results of the simulation. To begin, we need to load the RMLands R stats library (rmlstats.r): source('.../exercises/scenarios/scripts/rmlstats.r') Important: Note the use of forward slashes / instead of the customary backslash. The rmlstats.r library includes a number of scripts that can be used to create pre-formatted tables and graphs. For this exercise we will focus on a several different tables and graphs useful for comparing scenarios. Each table and/or graph is produced by calling one of the pre-existing rmlstats functions, as described below. In using the rmlstats functions below, note the following: Most of the functions contain the same set of arguments, e.g., session=, start.step=, stop.step=, and var=, so once you get used to using these arguments to modify the query, it is basically the same for all functions. All of the functions contain a session= or scenario= argument that allows you to specify single or multiple scenarios. In all cases, the single scenario query produces a different table and/or plot than the multiple scenario query. In general, it is useful to first produce a single scenario query in order to familiarize yourself with the nature of the output and then produce a multiple scenario query to compare scenarios. Ultimately, the multiple scenario queries will be most useful for this exercise. The multiple scenario queries can be done using all of the existing scenarios (dy default) or any specific set of scenarios. Given that we have 5 scenarios, the plots may be too cluttered to visualize effectively. It may be more useful to focus on specific comparisons. For example, to examine the effect of climate on the simulated range of variability (SRV) in landscape structure, it may be useful to compare scenarios that differ only in climate, such as 5 vs 7 or 6 vs 8. Most of the functions have an outfile= argument, which is a logical; i.e., if outfile=true, then output files will be generated and saved to disk using a default naming convention. The default is always outfile=false. 1. Area disturbed summary. Examine the tabular and graphical summary of the total area Lab. 5 Pg. 23

24 disturbed by timestep for each natural disturbance process and scenario, as follows: darea('.../exercises/scenarios/results/',session=5,start.step=1) A separate table and graph is produced for each disturbance process. To advance through the disturbance processes, simply hit the return key (you may have to click the cursor in the console window before hitting enter). To save the graph as a bitmap for use in a presentation, right click on the graph and select copy as bitmap and then paste directly into your presentation. To change the first timestep to be displayed, change the start.step= argument. To change the scenario to be displayed, change the session number using the table in step 3. Alternatively, it may be more effective to compare particular scenarios in single plot, by specifying a vector of session ids, as follows: darea('.../exercises/scenarios/results/',session=c(7,8),start.step=1) Or you can omit the session= argument altogether and all the unique session id s in the darea.csv file will be compared. darea('.../exercises/scenarios/results/',start.step=1) Note, when multiple sessions are specified, this function produces a clustered bar chart, with each cluster representing a different scenario (session id) and the bars representing the mean area disturbed per timestep (by default) for low, high and any mortality disturbances. To view the minimum, maximum, or median area disturbed per timestep, instead of the default mean, specify the var= argument, as follows: darea('.../exercises/scenarios/results/', start.step=1,var= median ) 2. Treated area summary. Examine the tabular and graphical summary of the total area treated per timestep for each management type and management zone, as follows: tarea('.../exercises/scenarios/results/',session=1,start.step=1) Note a couple of things. First, the treated area represents the acreage actually treated during a timestep, not the total area under management at any point in time. Second, the tarea.csv table that this function calls only contains records for the vegetation treatment scenarios (session 1), so any calls to the function with other sessions will produce an error. Lastly, the codes reported in the output correspond to the codes used to parameterize the model, as follows: MZ1 = Wildland-urban interface (WUI) zone MZ1-15 = includes individual tree selection treatment of the cool-moist forest cover types (e.g., western hemlock-western red cedar, dry-mesic spruce-fir forest and Lab. 5 Pg. 24