DRAFT December 4th, 2008 Applying Intrinsic Potential models – Workshop Session Notes Lead(s): Mindi Sheer (NOAA) and Erin Gilbert (ODFW) Note takers: Jacque Schei and Shallin Busch Below is a brief summary of the session, followed by notes from the “IP Applications” workshop session at the “State of the IP 2008 Workshop” held in Portland on November 19th and 20th. The initial summary (draft) and notes were compiled from discussions between workshop participants. If you attended, and would like to provide edits or corrections, please do – this is a working document. Background document for this session is: IP Workshop - Applying Intrinsic Potential Models Strawman SUMMARY(DRAFT): There was some interest in tracking IP-related projects and data sets. NetMap is one method of doing this. Sharing results across regions was important. There was consensus that there should be a long-term goal of a standardized DEM-derived stream layer, and a short-term need to document parameters that researchers are using. Some of this documentation might come out of this workshop. Group desired some kind of documentation of technical considerations related to generating streams from DEMs and for calculating IP scores by reach. There is also a need to document software (and version) used to generate streams (because settings may change depending on version). Group discussion indicated that (depending on interpretation and application), IP tends to overpredict the amount of habitat, so that should be considered for interpreting results. Group discussed a list of ways IP analyses are could be used by managers (eg, density dependence, areas of contiguous habitat, sensitivity analyse, criteria for historical size), and also discussed caveats for certain applications. Emphasized the need to consider historic changes in geomorphology in IP models. All recognized the need for groundtruthing actual usability of reaches prior to field work (such as restoration), if using IP as a method to identify restoration areas. Most management actions are done after relatively careful checking and comparing against other data sets. IP can be a piece in this – if used in conjunction with other data sets, it can provide a good backdrop. The group discussed the need for care when presenting IP scores and results to managers, and tailoring visual representation of results to the questions/needs of the managers. The subject of visual representation needs more exploration – The IP scoring system generates reach-based estimates, but there seems to be consensus that the accuracy level of smaller reaches is less so than a generalized, scaled up version. Discussed the need to “fuzzy” or generalize represents appropriately. One method of doing this is clustering results using the core area function – a NetMap tool function. The temporal aspects of climate and the affects of climate on flow were discussed, as was the change in peak flow timing. It is understood, in terms of IP models, that discharge may be similar year to year, but peak flows are changing, which could introduce another element into developing an optimal, simple IP model for certain species. There was a desire to keep IP simple (in fact the group expressed that was one of the major strengths of IP) and not add in too many other components, even though they could be useful at later steps or for some regions. IP APPLICATIONS SESSION Notes SESSION NOTES The purpose of this session is to provide an overview of ongoing applications of an IP approach (traditional or broad-scale habitat potential analyses), to open up a discussion on caveats to consider when using IP scores, and to collect information on ongoing projects from participants. Supporting doc: IP Workshop - Applying Intrinsic Potential Models Strawman Uses of IP Analyses 1. How are IP analyses currently used by managers (see Appendix A – what else)? Question 1: How do you use IP models to identify population structure? Answer: Use IP data in addition to other data sets, such as life history, productivity, or spatial distances. Question 2: How do you address historic changes in geomorphology? Answer: Consider known changes and apply professional judgment. Question 3. Are IP models biased in their predictions of suitable habitat? Answer: Probably, but more work should be done on this subject. Question 4. Can IP models be used to help screen restoration potential? Answer: Yes, depending on the level at which you are screening Question 5. Have we considered enough factors in IP models? Answer: IP focuses on intrinsic factors. Extrinsic factors may be added in addition to IP models, to better understand data variance and better pinpoint areas for restoration/watershed planning. For viability analysis o Determining potential spawning locales in geographic areas o Determine historical population size o Define populations boundaries o Set Minimum Viability Thresholds (targets and recommendations) o Set quasi-extinction threshold o Define areas of contiguous habitat Used to differentiate major spawning areas vs minor spawning areas o Used in conjunction with other data when addressing density dependence and population structure o Need for sensitivity analyses of models Restoration Scenarios And Watershed Planning o IP models used as a planning tool Identify reaches with high potential in relation to current condition of locale Overlay maps of habitat quality relative to intrinsic potential o Determine geomorphic types of habitat Habitat specific information o Assess reasons for differences in fish use What is due to extrinsic factors What is due to intrinsic factors o Consider the possibility of changing underlying habitat or geomorphology See supporting documents for information on: 2 IP APPLICATIONS SESSION Notes o Forestry /Land use related Analysis o Habitat Diversity and Spatial Structure o General Fishery Management 2. What are appropriate applications of data from IP models? There isn’t much risk that restoration actions would be planning and implemented without site visits. It’s ok to use IP score to give general idea of habitat potential and /or condition. This information can help direct site visits Using IP to rank possibilities for dam removal – use IP to prioritize –ok use (obviously, not actions again, would be taken without some kind of site visits). Determining recovery potential - What is the response of high, moderate, low disturbances with respect to species? Lay IP out as a hypothesis then test it - Develop parameters (GIS scenario) and select which you want to use based on the habitat you are looking at. Probability of species use and spatial distribution (can be improved with ground-truthing) Classification of habitat by geomorphology Classification of habitat by habitat features Use at the sub watershed scale to compare potential among areas Standardized approach for identification of population segments based on contiguous habitats Useful to organize assessments of fish use and habitat potential Determine juvenile rearing potential, filtered with spawning potential 3. What are inappropriate applications of data from IP models? Inappropriate use as opposed to what? Managers need to make a decision whether the data is there or not. Is there something besides IP that will improve their decision-making capabilities? If so, that can be used but sometimes there isn’t. Assessment of low gradient habitat to evaluate current conditions Estimating fish density/productivity Estimating historic abundance (proceed with caution because not able to validate results) Applying models developed for one locale or species to a different locale or species Make conclusions on specific reaches based on broad scale assessments o Esp for low gradient habitat Cautioned use at the reach scale due to lack of geomorphological data Cautioned use of extrinsic attributes at the reach scale for alteration of geomorphological features over time DATA AND MODEL ISSUES Need for ground truthing of GIS data Need to snap IP data to 1:24k hydro or appropriate hydro layers. Validation of drainage enforcement. Consider adding wood recruitment models to IP models Validation of IP models are coarse o Tend to line up with predictions for fish o But validation also evaluates habitat (?) o Models don’t deal well with ephemeral streams o Some models don’t deal well with wide streams 3 IP APPLICATIONS SESSION Notes Greater than 50 meter wide streams for Steelhead APPLICATION CONSIDERATIONS IP used to predict amount of habitat – it DOES over-predict habitat though The major benefit of IP is that it is simple Need to clearly identify species and life stage before applying a particular model. What life stage is the limiting factor life stage? Which habitat metric measures that stage? What remotely-sensed variables can be used to measure that metric? At what point does an IP model become simply a fish distributon model. If you consider spawners in different habitats in different years and summarize this somehow – does this move away from IP and become more of a steelhead distribution model? IP users and managers using IP need to consider and incorporate the temporal domain as well as the spatial – IP models don’t typically do this, and this can be confusing or limited when presenting results Is it appropriate to think the past can be in the future (ie, is the intrinsic potential useful in highly modified areas, or areas that have undergone climate shifts). What about those cases where the future potential is vastly different than past (historical) situations – e.g., climate change. The concept of IP highlights historic potential, but in some cases the environment has changed so substantially (temperature, flow regimes,etc., S.CA), that the concept of identifying what could habitat could support fish from a historic perspective may not be pertinent. With changing climate, even if mean annual flow has changed over time, drainage area probably has not (ie, this might be a better intrinsic variable than mean annual flow, which is more sensitive to human interactions OR changes in climate). Precipitation and mean annual flow may be shifting with climate changes – mainly in the seasonal timing. Seasonal changes in timing of precipitation and peak flows may be changing, which impacts IP for fish What about extreme events? Consider scale and accuracy of data layers in context of results and inference IP models and data, regions 1. What projects/examples are missing from this table (Table 1), Appendix A or B? Send in comments (make edits, add content, references, etc.) Suggestions for other fields to include: life stage, contact info 2. What should be done (if anything) to keep track of projects or share results? List of IP projects with contact info for researchers Links to IP results and summaries o Ex: work in CA NOAA SWFSC (Brian Spence) Links to data o Metadata o Full dataset o Use existing web-based data repositories KRIS coho StreamNet 4 IP APPLICATIONS SESSION Notes NED Portal Gather info from participants at workshop first o Use existing communication/collaboration resources o Consider data repository later Can all these IP models go into NetMap? NEXT STEP IDEAS Need for coordination exists o How to handle use of different rule sets? Long-tern goal of standardizing DEMs? o Still in experimental phase to determine best approach Updates in NetTrace o How are these documented o What are concerns o How are barriers documented in NetTrace Need to consider LiDAR o Relevance of IP when LiDAR comes online? o Puget Sound LiDAR consortium o New OR Consortium (DOGAMI leads): o Links? Areas to watch for IP work o Puget Sound o Interior Columbia o ODF – Wilson & Siuslaw Rivers o Alberta (NetMap) 3. Is there overlap in regional IP projects – should we do something about this (Appendix A and B)? Group did not get to this question 4. Which of these projects have reports, metadata, or papers (even drafts) that document or present the efforts or results of the project(s) (see Table 1)? Group did not get to this question IP function – Geomean, variable weighting, binning IP results 1. IP approach gives all variables equal weight. What is the consensus on lack of weighting factors in the IP Geomean score? Are there ways of weighting certain variables, or is this necessary? FISH BARRIERS/GRADIENT Since gradient is a primary driving/limiting factor in IP analyses, is it possible to come up with a probability-driven cutoff to counteract erroneous gradient estimates? Perhaps gradient (?) cutoffs or other curves (?) could be derived using a regression (estimates incorporating geology) Perhap use ‘stoplight’ method to define barrier threshold o Red – fish cannot get through barriers o Green – fish can get through barriers o Yellow – undetermined 5 IP APPLICATIONS SESSION Notes How fish respond to gradient is dependent on reach width and flow in the reach. The gradient curve (for fish) should reflect this dependency. Key thing is interaction of gradient with magnitude of high flows Note that the distribution of fish varies by mean flow and the flow in the 7-10 days before spawning. Both the IP curves and how gradient is seen will vary by this (temporal variation). Gradient cutoffs (at 20% gradient, as indicated by WDFW suitability cutoffs) L. Columbia tribs are not so useful for steelhead Noted that small waterfalls are not picked up in the DEM. Point data (cascade and waterfall locations) may not be picked up by the DEM, but luckily the states of point data on many of the waterfalls and substantial cascades. DR is working on a project using regression to look at transition zones for steelhead, chum, and chinook(?), assessing attributes for 4-100m reaches above and below the stopping point for the species. CA used a separate barrier layer, as did many others. States have a natural barriers layer that most use USFS (CLAMs?) uses ODFW fish distribution maps combined with fish bearing streams at 20% gradient to run IP model, report results up to 10% gradient o Need note in document about NetTrace method of max_grad_downstream estimate,which is helpful in allowing this kind of reporting and summarizing-results efficiency. o Compare LIDAR derived stream with 10M DEM-derived LIDAR will get around a lot of issues, but will still need to decide appropriate reach length scale Will solve problems of what we can measure, but will not resolve issue of what we should measure 2. What are effective ways of binning and normalizing scores? What is the consensus on IPKM? Depends on question – looking between watersheds versus within a watershed Is a .1 IP score difference meaningful? No. Too many variable factors. It is a geometric mean on correlated variables. In CA, they used a binary system (envelope) – similar to geometric mean (CA) – fit step functions based on occurrences of species. Used the envelope versus the curve function. Weighting factor to use? What to do when you end up with a lot of 0s? Use Core tool in NetMap Delineated reaches at 1000m reaches you would get less 0s within a contiguous block – 2-teired process, pick gradient to reflect barriers, generalize habitat or results – could do barrier or distribution analysis outside IP, use IP scores within truncated distribution Participants that knew about IPKM thought it was a usefull way to display, use the IP data. DISPLAYING DATA For displaying and presenting data to managers. Need to answer – “What is the management question you are trying to answer with IP score?”, and “Who are you presenting information to?” Management questions being asked now are different than the MQs asked for historic IP scores Is this for Restoration or protection? 6 IP APPLICATIONS SESSION Notes Display by 6th field HUC For display or further analysis, use the Clustering Tool in NetMap – Core Area Tool – allows you to set clusters (cluster tolerance factor) It would be useful to display estimated IP versus actual spawning ground aggregations (potential versus actual) It would be useful to come up with a probability index for a certain percent (??) of an index (suite of probabilities that give insight into particular parameters) Summary scale 1. What are the most appropriate scale for summarizing the IP scores? How to determine this? SCALE of ANALYSIS IP scores based on reach length used NetMap Geomorphic reach length – is limited by resolution of data used to calculate reach gradients; use variable reach lengths – set reach breaks at geomorphically relevant reach gradients o Another approach is to make all reach lengths the same – consensus was that this makes analysis easier. Reaches can be grouped later. How long of a reach do you need to get an accurate gradient? Rely on contours; Calculate where the contour crosses the stream. The gradient is then assigned to the local reach, even if the reach length does not match the length between contour crossings. Lengths over which gradients are calculated change based on slope (in NetTrace) Has anyone looked at IP scores at 100m, 200m, 400m – to look for opportunities at smaller scale o Smaller the spatial extent, the higher resolution data you need o Watershed analysis in Wilson Basin (ODFW) – prioritize riparian management, debris flow inputs from upslope wood to fish habitat - looked at spot in channel and delineated upslope; resolution of 10m DEM was insufficient to do this – now trying to use LIDAR 2. What are some ways of comparing IP scoring in overlapping or neighboring regions to other studies? Group did not address. New variables 1. What are other important physical thresholds that could be (and sometimes are) included? What are other useful intrinsic variables? Need to be remotely sensed (ie, need to be available at a certain scale at a regional map extent) Use to determine spatial extent (?? ) Physical variable chosed depends on what you’re trying to model and the scale of your model – is it coast-wide, watershed, ESU? Certain physical / ecoregional features may change fish response to the particular physical variable added. Temperature, Air temperature, elevation, vegetation Is it possible to model likelihood of gravel dist and pair this with flow as a useful intrinsic variables? Yes. FISH CONSIDERATIONS A systematic, biology-centric approach to coming up with IP 7 IP APPLICATIONS SESSION Notes Cleary identify what species and life stage you’re calculating an IP score for (ex. Coho overwintering) Identify the limiting habitat metrics Determine GIS surrogates for those metrics It would be nice to standard definition for barriers – to have consistency across datasets For species (steelhead and coho) people tend to pick GIS attributes or factors that are related to juvenile abundance. Different for Chinook and Chum 8