1. Erik Porse
  2. http://www.igert.org/profiles/5243
  3. Graduate Student
  4. Presenter’s IGERT
  5. University of California at Davis
  1. Allison Dedrick
  2. http://www.igert.org/profiles/5445
  3. Graduate Student
  4. Presenter’s IGERT
  5. University of California at Davis
  1. Shahla Farzan
  2. http://www.igert.org/profiles/5446
  3. Graduate Student
  4. Presenter’s IGERT
  5. University of California at Davis
  1. Matthew Hamilton
  2. http://www.igert.org/profiles/4425
  3. Graduate Student
  4. Presenter’s IGERT
  5. University of California at Davis
  1. Gabriel Sampson
  2. http://www.igert.org/profiles/5247
  3. Graduate Student
  4. Presenter’s IGERT
  5. University of California at Davis
  1. Derek Young
  2. http://www.igert.org/profiles/5444
  3. Graduate Student
  4. Presenter’s IGERT
  5. University of California at Davis

Judges’ Queries and Presenter’s Replies

  • May 20, 2013 | 08:53 p.m.

    Please describe the criteria used to develop the potential management options. There seem to be endless possible scenarios depending on the size of the management unit and the size of the managed patch.

  • Icon for: Derek Young

    Derek Young

    Co-Presenter
    May 21, 2013 | 05:52 p.m.

    Dear Dr. Yavitt,

    Thank you for your question. The set of potential treatment combinations is indeed very large. In addition to deciding which sites to treat, it is also necessary to decide on the particular treatment method to use at each site. These decisions depend on numerous social and ecological factors across the landscape. We have conducted extensive semistructured interviews with rangeland and sage grouse managers in the Modoc Plateau region to develop an understanding of the variables that are most important in affecting the costs and outcomes of treatment for different management objectives. Using existing data sources primarily from the Natural Resource Conservation Service (NRCS), we have constructed a high-resolution GIS-based map of the variables that managers and the scientific literature have highlighted as the most likely to affect treatment decisions. These variables include juniper canopy cover, forage production potential, distance from nearest potential sage grouse lek, distance to nearest road, and land ownership. Taken together, these variables allow us to estimate the cost of treatment, the potential increase in forage production, and the potential improvement in sage grouse habitat that would result from treating any give site. We used this map to develop a grid of approximately 3000 “management units” that are 2000 m x 2000 m, representing the median size of juniper projects that are carried out in our study region.

    We are implementing optimization algorithms to select multiple treatment sites from this set of 3000 options in order to maximize a given management objective (forage production, sage grouse, or a combination of the two) while staying within a specified budget. We are using two separate tools for the optimization: mixed integer linear models implemented in CPLEX (IBM, Inc.) and a greedy algorithm implemented in Zonation (University of Helsinki). Mixed integer linear optimization models efficiently search the entire set of potential treatment configurations and identify the most efficient configurations for a specified objective within a specified budget. However, CPLEX is a very expensive system; therefore, we are comparing results from CPLEX with those from Zonation, a free conservation tool that uses a greedy algorithm heuristic and therefore is not guaranteed to identify the most efficient solutions. If results from Zonation are similar to those from CPLEX, this will highlight the potential for managers to use Zonation to help prioritize the allocation of juniper treatment funding across the landscape.

    To assess the compatibility of juniper management for the benefit of sage grouse and for forage production, we will conduct multiple model runs, each giving different relative weight to the two benefits. We will first assess the ideal management approaches for either sage grouse or forage production alone; a comparison will reveal how compatible or incompatible the two objectives are. Then we will explore management for both objectives simultaneously to understand the tradeoffs that are necessary when managing for multiple objectives.

    Our results are not meant to represent a final or complete solution; they will serve as a starting point for more complex and accurate models and will also serve as a template for others wishing to address similar questions in other socio-ecological systems.

    Thank you again for your inquiry.

  • May 21, 2013 | 12:06 p.m.

    Given the preliminary nature of your research, can you provide some examples of management strategies and predicted outcomes that provide insights into the limits of your models?

  • Icon for: Derek Young

    Derek Young

    Co-Presenter
    May 22, 2013 | 11:30 a.m.

    As a simplified representation of reality, our model does not account for all of the socioeconomic and ecological variables that control the costs and benefits of any given management strategy. An example of an outcome that is very difficult to predict is invasion by non-native grasses, such as cheatgrass and medusahead, following management. These grasses can invade the disturbed soil that sometimes results from juniper removal. When they establish on a site, they are very difficult to remove and can dramatically reduce production of palatable forage for livestock. However, invasion is controlled by a number of very difficult-to-model factors, including amount of ground disturbance incurred, proximity of nearest invasive plants, and presence/absence of invasive plant seeds on the treatment equipment. Currently, our model predicts change in forage production assuming no invasion; this represents an upper bound on benefits. Soon, we plan to include a stochastic component that incorporates random variation in the processes that we are not able to pinpoint with high confidence, including invasion probability.

    Further, some components of our model are based on assumptions about how land managers choose among a range of treatment methods given certain biophysical characteristics (for example, felling trees using chainsaws when canopy cover is below a threshold). From our own observations in the field, and from speaking with a number of land managers, we have come to recognize that there’s considerable diversity in decision-making heuristics, which in turn inform management strategies. In this regard, another current limit to our model is that we have not included the effects of decision-making heterogeneity. However, in our ongoing interviews with land managers, we are collecting the sort of quantitative data that may allow us to explore the effects of different treatment method strategies.

    Another limitation is that we are modeling a static, one-shot decision. There are no time dynamics in our model. Because sites may be re-invaded by juniper following treatment, thus requiring subsequent re-treatment, the management of juniper is actually a repeated decision-making process. Potential for re-invasion in subsequent periods is something that should influence management decisions in the present. Although incorporating future ecological dynamics is currently beyond the scope of our project, the time-scale of reinvasion is such that re-treatment would most likely not be necessary during the careers of the current decisionmakers.

    Another challenge we faced involved developing an approach for comparing multiple responses (cattle grazing and sage grouse habitat) that are not measured on the same scale. It is a bit of an apples-and-oranges problem: which should be worth more, one cow or one sage grouse? To get around this problem, we are using an approach known as the “constraint method” in optimization to help us understand trade-offs in objectives. This approach involves optimizing first for just one management objective, then again for only the other management objective, then multiple times for different relative weightings of both management objectives simultaneously. This approach allows us to explore the tradeoffs among objectives (e.g., you must give up two sage grouse to get one more cow) without requiring a common metric to compare the two.

  • May 21, 2013 | 07:53 p.m.

    Can you tell me more about how the IGERT team prepared for and conducted interviews with any of the key decision-making groups?

  • Icon for: Allison Dedrick

    Allison Dedrick

    Co-Presenter
    May 22, 2013 | 05:02 p.m.

    Dr. Anderson,

    Thanks for your question. After visiting the study region and speaking with a number of key stakeholders, it became apparent that semi-structured interviews would help us gain a deeper qualitative understanding of the perspectives and decision-making strategies of land managers. Additionally, from our review of literature on ecological and economic dimensions of juniper encroachment, we recognized that interviews could serve to address information gaps and provide quantitative data to inform model parameters.

    Several members of our group have been trained in social science research methods, including interview protocols and survey methodologies. To prepare for our interviews, we identified key stakeholders and managers involved in regional juniper treatment, drawing upon existing literature and reports, contacts from our visit to the region, and individuals identified by other key stakeholders. Interview subjects included representatives from federal agencies, non-profit organizations, the private sector, and a representative of the tribal organizations in the area.

    We developed a set of questions to address the information gaps identified from our literature review as well as our need for a better understanding of decision-making heuristics and land management strategies. After getting approval from our campus Institutional Review Board, we scheduled a series of interviews with as many of our potential respondents as possible. Interviews were conducted over the phone or in person with at least two members of our group present to independently record and validate responses. We have compiled a database of interview responses and are using the data to estimate key model parameters.

  • May 22, 2013 | 08:56 p.m.

    How nice to meet you in cyberspace, Allison! Thank you for explaining the processes which are as integral to collaborative research as the findings because they help us build on a body of knowledge. I liked that you ncluded interacting with your Institutional Review Board first and had at least two members to collect data.

  • May 21, 2013 | 09:53 p.m.

    What sort of algorithm are you using to weight the importance/influence of different data sources/inputs? For example, do the structured interviews have as much of an influence in your optimization as physical habitat characteristics? Trying to integrate “hard” data with socio-economic factors is always tricky and I’m curious how you did this because it’s really the most novel part of this project.

  • Icon for: Allison Dedrick

    Allison Dedrick

    Co-Presenter
    May 22, 2013 | 08:27 p.m.

    Dear Dr. McGarvey,

    Thank you very much for your question. Integrating data from multiple sources is always difficult, especially when they cut across biophysical and socioeconomic factors. For this study, we are working on a topic that has a large volume of data available but in many forms. Some data are specific to the Modoc region. Other data are more generalized for a large, multi-state region. Finally, some data are very localized to areas other than our study area, such as Oregon. Thus, the challenge is integrating data of different types and scales.

    The scientific data, such as soil type and precipitation amount, inform the biophysical modeling aspects of our project, such as estimating forage production from treated areas. The interview data inform the socioeconomic factors, including treatment costs, how land cover characteristics influence treatment type, and thresholds for treatment type. Some of the interview data serve as background information and inform our understanding of how both public and private land managers currently make decisions about juniper management, which affects the way we structure the model. For example, interviews have provided information on general relationships between percentage juniper cover and treatment approach. We have integrated this information with very specific, imagery-based data for juniper cover in the region to set parameters for treatment decisions in our modeling. In other instances, the interview data inform specific values that are otherwise unavailable for the Modoc region, such treatment cost per acre. To get Modoc-specific parameters, we asked a variety of individuals heavily involved in juniper management to estimate such values. Thus, we are working to effectively integrate data sources while remaining cognizant of the limitations of each data source.

  • May 23, 2013 | 09:51 a.m.

    Fair enough. Thanks Allison.

  • May 21, 2013 | 10:55 p.m.

    Dear Erik et al.,

    Huge issue! Will be very interesting to see what you find!

    What I’m wondering though is where these landscapes are predicted to ‘go’ given climate change? Here in the Midwest, honey locust is invading many prairie sites. Restoration ecologists are working hard to get rid of it. However, climate predictions would suggest that it ‘should’ establish here in the future. So when is honey locust no longer an invader? Or more specifically in the context of your study, is juniper woodlands where these landscape are likely to transition towards given climate change?

    Best,
    Volker

  • Icon for: Shahla Farzan

    Shahla Farzan

    Co-Presenter
    May 22, 2013 | 04:41 p.m.

    Dear Volker,
    Thank you for your question and for your interest in our research! We feel that one of the most interesting aspects of our project is the fact that “native invasions” are occurring across a multitude of habitats and involve a broad diversity of plant species. It’s great to hear from someone who is witnessing a similar phenomenon in another part of the country.

    Like other native species experiencing rapid range expansions, Western juniper encroachment appears to be highly dependent on climatic conditions. In particular, changes in annual precipitation will likely determine how future climatic conditions favor range expansions in Northern California. Western juniper seems to prefer conditions that are a bit wetter than those in the Great Basin today. During the period of most rapid Western juniper expansion in the Great Basin (1850-1916), winters were considerably milder and wetter. Predicting future Western juniper habitat suitability is difficult because climate models are split about 50/50 as to whether Northern California will get wetter or drier. Therefore, at this point it is unclear whether conditions conducive to further Western juniper expansion will exist again in the future.

    Climatic projections aside, your question raises the interesting issue of human subjectivity and managing the range expansion of native plant species. In our interviews with land managers, private contractors, and government agencies, we have encountered a broad range of juniper control philosophies. Though most managers have taken the view that juniper is not an appropriate part of the landscapes they are treating, most federal programs have stipulations that old-growth juniper should not be removed. Even more, some disagree with the premise of juniper removal more broadly because juniper expansion may be a natural phenomenon. As a result, federal and state policies for managing juniper often reflect the difficulty of balancing these conflicting views.

  • Further posting is closed as the competition has ended.

Presentation Discussion
  • Icon for: Brian Drayton

    Brian Drayton

    Faculty
    May 20, 2013 | 10:41 a.m.

    Very interesting project. How does the tree disperse/establish? What were the changes that enabled it to become invasive? What role does fire suppression play?

  • Icon for: Erik Porse

    Erik Porse

    Presenter
    May 22, 2013 | 11:12 p.m.

    Hi Dr. Drayton,
    Apologies for not directly replying to your comment previously (it was listed as a separate comment). Thank you very much for your interest and questions regarding our project. Western juniper disperses by seed in a few ways. Berries with seeds either fall onto or wash to fertile soil where they can take root. Birds also spread seeds to new areas. Young junipers become established through strong taproots, which allow them to better access soil moisture. Young junipers are also less palatable to browsing herbivores due to their sharp needles.
    A number of factors are linked to the expansion of Western juniper. A general consensus exists among experts that severe overgrazing in the late 1800s reduced grasses and other herbaceous plants, which gave young juniper plants a competitive advantage. Historically, juniper densities were kept in check, both in scale and location, by periodic fires. Many decades of fire suppression also helped juniper out-compete herbaceous plants. There have been some recent efforts to carry out large-scale controlled burns to address juniper encroachment.

  • Icon for: Noam Ross

    Noam Ross

    Trainee
    May 20, 2013 | 05:36 p.m.

    Sounds promising! I’m curious about the track record of such optimization in socio-ecological systems. What are some examples of how “mixed-integer linear programming and genetic algorithm-based optimization” have been used in related problems?

  • Icon for: Erik Porse

    Erik Porse

    Presenter
    May 20, 2013 | 05:45 p.m.

    Hi Bryan,

    Thank you very much for your interest and questions regarding our project. Western juniper disperses by seed in a few ways. Berries with seeds either fall onto or wash to fertile soil where they can take root. Birds also spread seeds to new areas. Young junipers become established through strong taproots, which allow them to better access soil moisture. Young junipers are also less palatable to browsing herbivores due to their sharp needles.

    A number of factors are linked to the expansion of Western juniper. A general consensus exists among experts that severe overgrazing in the late 1800s reduced grasses and other herbaceous plants, which gave young juniper plants a competitive advantage. Historically, juniper densities were kept in check, both in scale and location, by periodic fires. Many decades of fire suppression also helped juniper out-compete herbaceous plants. There have been some recent efforts to carry out large-scale controlled burns to address juniper encroachment.

  • Icon for: Erik Porse

    Erik Porse

    Presenter
    May 20, 2013 | 06:02 p.m.

    Hi Noam,

    There is a good collection of literature that has looked into this problem over the past few decades. In the conservation and reserve siting realm, back in the 1990’s and 2000’s a number of researchers were using mixed integer and heuristic search methods to identify optimal configurations of reserve sites based on cost and habitat. Marxan and Zonation are products of this literature and have a lot of publications behind them for land and marine reserve sites (these programs use different types of search algorithms, including simulated annealing and greedy algorithm). In addition, there are a number of papers from Williams, ReVelle, and Levin- a 2005 paper from these authors is the best review I have found on the MIP applications (“Spatial attributes and reserve design models: A review”). Among neat applications, Williams worked on flyway reserve sites.

    Plenty of studies in the water realm would qualify as well, depending on how you consider a social-ecological system.

    Regarding track record, that is a good question. I tend to think that the level of effective implementation of model solutions is related to how savvy the model developers are in translating the findings in the context of complex policy and ecological factors.

  • May 22, 2013 | 01:57 p.m.

    Hello !! I like the video a lot, very informative. But for my experience, trying to remove an invader by mechanical methods could be costly and not really effective. Are you trying to determine other possibilities? Maybe trying to study the biology of the species and find the most important life stage to attack. Cattle or any other species can forage on juniper saplings? Are there some landscape attribute that you found that avoid or prevent juniper establishment? Just thinking loud !!! GOOD JOB !!

  • Icon for: Shahla Farzan

    Shahla Farzan

    Co-Presenter
    May 22, 2013 | 10:02 p.m.

    Hi Maria,

    You make an excellent point. Mechanical removal of juniper can be costly, depending on a number of site-specific characteristics. For instance, the distance of the site from roads can have a substantial impact on the cost of treatment per acre. Based on our interviews, it appears that most treatment focuses on sites fairly early in the encroachment phase. Once a site becomes too dense, it is not only more difficult and costly to treat, but it is less likely to support threatened sage grouse even after treatment (the birds move on to less degraded sites as the juniper stand becomes denser and are unlikely to reestablish after the juniper are removed).

    At this point, mechanical removal is essentially the only treatment option that is widely used. Prescribed fire, while undoubtedly effective in removing juniper, can be difficult to control. As a result, land managers appear to avoid using fire to control juniper encroachment. Herbicide application has met with mixed results, but it tends to be more expensive and can have undesirable ecological consequences. Because it’s a tough woody conifer species, it is difficult to train cattle to eat Western juniper. However, after seeing Kathy Voth train cattle to forage on seemingly unpalatable invasive plants such as yellow star thistle, perhaps anything is possible! (see www.livestockforlandscapes.com)

  • May 22, 2013 | 11:19 p.m.

    Good work, and great discussion here so far. I’m curious how the expanding juniper might interact with other large-scale vegetative changes in the West, such as cheatgrass expansion and pine-bark beetle outbreaks. Do either of these dynamics (or others) play into how the juniper is able to expand so well?

  • Icon for: Derek Young

    Derek Young

    Co-Presenter
    May 23, 2013 | 06:45 p.m.

    Michael,

    Glad you brought this up; indeed, disturbances never occur in a vacuum, and interactions among disturbances are arguably the most important drivers of ecological change. Juniper in our region has not been affected by bark beetles to nearly the degree that many stands of pine in other parts of the West have been, but it is certainly possible this will change. Although bark beetle influence is expected to continue increasing throughout the West, we aren’t aware of specific projections for our area.

    Cheatgrass invasion could also interact significantly with juniper invasion. There have been no studies to our knowledge examining how the two invasions will interact, but one could imagine it going either way. On the one hand, cheatgrass, once established, is very effective at carrying fire, and a regime of more frequent, larger fires could help prevent juniper seedlings from establishing. On the other hand, fire suppression is still a top priority of management agencies and private landowners in the West, and these groups may focus suppression on the more fire-prone, cheatgrass-invaded areas, allowing juniper seedlings to establish. How these interactions will actually play out is still very poorly understood, however. Regarding the other direction of this interaction, there is some anecdotal evidence that removing juniper can facilitate cheatgrass invasion if steps aren’t taken to restore the native shrub and herb cover post-treatment.

  • Further posting is closed as the competition has ended.