Minutes of Fourth Meeting of Science Advisory Panel

Minutes

Courtney opened the meeting and emphasized how much has been achieved. The panel have been major contributors, and have ensured the scientific standards of the process. The credibility of the HCP has depended on outside review and the direction of the panel. New analyses have been carried out at the panel's suggestion, and these have been helpful.

The original intent of the scientific process was to respond to the agencies' needs for additional information. The science was crafted to answer these particular information needs. Since the last panel meeting, the agencies collectively decided on an appropriate conservation strategy (setting out areas to be conserved) which was ultimately accepted by Pacific Lumber Company (PalCo). The company has now prepared a Habitat Conservation Plan (HCP) under this strategy. There are now some issues on which both the agencies and PalCo need additional guidance in the final phases of HCP preparation.

The HCP has been based on the best available science. The agencies' decision, and PalCo's plan have both depended heavily on the available data and analyses. Where analyses were unavailable or incomplete, or where there was substantive scientific disagreement, a conservative or 'worst-case' scenario was always adopted. Of course, there have been continuing analyses, and our understanding of Marbled Murrelets continues to improve. Two recent analyses have been carried out (on offshore population trends, and inland survey detectability) and are presented as part of the HCP. However these two analyses have not resulted in any change to the HCP, because of the plan's conservative strategy or 'precautionary principle'.

Offshore analyses
Linear regression analysis on offshore data to 1997 are now available. The Redwood Sciences Lab group of CJ Ralph have prepared a document, which will be incorporated into the HCP. The main results of this analysis are that in the northernmost sections of California, the Marbled Murrelet population appears stable or increasing; off southern Humboldt County, the population appears to be decreasing at 4% annually. However all populations seem to show an increase in 1997.

Despite these analyses, the HCP has been drafted under the assumptions that the predictions of Beisinger (1995) are correct, i.e. that the population is declining range-wide at 4 to 6 % annually. Given uncertainties over the analyses carried out to date on the offshore data, the HCP was drafted under the 'worst case assumption' that declines were ongoing. This approach was also consistent with the opinion of the Science Advisory Panel at the Arcata (2nd) meeting.

Kareiva reported on the results of analyses carried out by a graduate student at the University of Washington, on a partial data set (as reported by RSL on their web-site). These data, which refer only to the years up to 1995, were subjected to Maximum Likelihood Estimation techniques, which are probably the most appropriate technique. The partial data set suggest that the rate of decline may be larger than 4% for the period to 1995, and that there is ' no chance of no decline'. However the estimate of rate of decline is quite sensitive to the way that data are blocked, and to the assumptions of the analyses. The panel discussed these analyses at length, and suggested that they should be repeated with more recent data as these become available. This will be particularly important to monitoring programs. However the HCP itself was crafted with the assumption that there is an ongoing real decline, and hence should be robust against changes in our understanding of the rate of decline.

Reid suggested that the group should attempt to distinguish among three sorts of ongoing information:
1. any work that is essential to the completion of the HCP (such as agreement on take estimates)
2. work that is useful in evaluating the HCP (e.g. in the EIS process).
3. work useful in supporting the approach taken in the HCP (e.g. the strongly conservative assumption of ongoing decline can be supported by more complete analysis of the offshore data).

Inland survey data: analysis of occupancy
At the San Francisco (3rd) panel meeting, it was decided to engage a statistical consultant, G.White (Colorado State University) to determine whether the assumptions of RSL's analysis of occupancy were reasonable. Specifically, the group was interested in determining whether the assumptions of the 'Relative Bird Value' (RBV) approach, that the probability of detection of occupancy was constant across habitat types and time, were reasonable. If the assumptions were invalid, it might still be possible to develop a correction factor to weight different habitat types.

White's report on these analyses indicates that there is indeed significant variability in the probability of detecting occupied behavior. This probability varies with year, season, and to a lesser degree with habitat. The significant between-year variation is probably due to an artifact: 1995 data are heavily skewed by surveys carried out in that year in Humboldt Redwoods State Park (prime habitat). Within year variation however is large, and is consistent with our knowledge of nesting chronology of the species.

The most important part of the analysis concerns differences among habitats. White shows that there is statistically significant variation in detection probabilities, with detection of occupancy (when birds are present) being highest in Old-Growth Redwood, intermediate in Residual stands, and lowest in mixed Old-Growth with Douglas Fir. However these differences, although statistically significant, do not translate into large errors in weighting. Because several visits to a site (10 under the current protocol) are necessary to determine occupancy, the actual error introduced by differences in detectability is very small (on the order of 1%). Hence the observed between-habitat variability, while real, has little effect on the analyses carried out by RSL and others.

Despite the positive outcome of these analyses, which suggest that the RBV approach has some merit, the HCP did not make use of this method. Instead a more conservative method of estimating Take was adopted, which assumes that there are no differences in density of Murrelets in different stands. This approach will be highly conservative in that the major reserve areas (e.g. Headwaters, MMCAs) are accorded the same weighting as less suitable areas. The HCP therefore adopts the 'worst-case' precautionary approach.

The group discussed White's analysis and felt that it was valuable, although the strategy of the HCP renders moot the results of the analysis. The group noted that White's finding, that detectability varied with habitat type, itself suggested differences in density of Murrelets in the three habitat types considered. The group also felt that the analysis could and should be taken further, particularly during the monitoring phase of the HCP.

The HCP

There was a general discussion of the HCP, the strategies used in developing the plan, and the placement of MMCAs. The need for buffers was discussed, as was the role of the HCP in the recovery process (the HCP is not required to contribute to recovery of the species, but cannot jeopardize it). Generally, the only potential or actual habitat areas that will be harvested, with the exception of one old-growth stand, are small, unbuffered, and furthest from the ocean of the stands considered.

Mitigation under the HCP was discussed. It was agreed that mitigation was being achieved through habitat set asides. Forest succession is unlikely to result in new nesting habitat in the short-term, but may contribute to habitat quality by buffering, and provision of shelter to marginal nest-sites.

Reid presented a recent analysis of 'Take', or the expected effects of the proposed HCP on Murrelet habitat (see Appendix). The metric uses the most recent data on timber types, and on survey data, to determine the amount of habitat that will be protected under the HCP (most unentered old-growth; some residuals). Reid's goal was to develop a 'worst-case' estimate of Take. He made explicit statements of where his assumptions were conservative, and that a 'best-case' estimate would indicate loss of habitat of far fewer Murrelets than envisaged in the HCP.

The group generally supported the approach, but identified three sources of error in the method used:
1. The analysis attempts to identify sites that have been 'adequately' or 'inadequately' sampled, and to make correction factors for 'inadequately' sampled stands. However many areas were mislabeled as inadequate, and were in fact surveyed to PSG protocol. This systematic error will lead to some areas that are known not to be habitat being assessed as potential habitat.
2. The method also uses GIS data to assign occupied stations to areas within 1/2 mile, without reference to the PSG protocol standard of assigning detections, or to alternative metrics (e.g. that used in the RBV method of RSL). Again, this will lead to a significant error in that occupied behaviors will be assigned to all stands in an area, hence maximizing the areas assessed as potential habitat.
3. The areas that have been surveyed are not randomly selected. Notably, low-quality residual areas are not surveyed (as directed by CDFG staff). High-quality areas are surveyed; estimates of occupancy of residual areas are based only on these surveys in high-quality habitat. Again this will lead to an over-estimate of the amount of habitat on PalCo lands.

In the discussion, the group noted these sources of error, and that, in every case, the error would lead to an over-estimate of Take. The group therefore felt comfortable in assuming that Reid's analysis was successful in setting an upper bound for the Take estimate. The likely amount of actual Take can not be determined at this point, but is less than that assumed in the HCP.

There was extensive discussion of the use of different habitats by Murrelets. In particular, several of those present thought that the best predictor of habitat quality was the number of potential nesting platforms in the stand. Chinnici noted that past timber harvest practices, when first entering a stand, were to remove trees with platforms, and to leave as residuals the straight trees with few side-limbs. Thus past harvest practices may have eliminated most potential nesting trees in residual stands. Courtney reported opinions of others that the numbers of platforms in residual stands were approximately 1% of those in old-growth stands.

Generally, there was agreement that Reid's analysis was the best 'worst-case' argument (i.e. the approach was a well articulated and supported estimate of the most pessimistic case), and that this conservative approach was the most appropriate way to proceed with the HCP.

Monitoring

Under the HCP, there will be both compliance and effectiveness monitoring. The HCP sets out a program of several techniques.

The group discussed the goals of monitoring. It was felt that it was not possible to develop a program of monitoring with specific feedback loops to management for two reasons. Firstly, mitigation in this HCP is all 'up-front' by setting aside habitat. No additional mitigation will be feasible. Secondly the long lifespan of the bird suggests that there will be long lags in response to any change in conditions. Hence it was thought that 'adaptive management' was not feasible in this HCP. However monitoring should be able to provide information to the agencies so that they can determine whether the HCP is in fact leading to a greater than expected effect (to the level of causing jeopardy).

The group did feel that a program of monitoring was essential, and should include:

1. ongoing oversight by the Science Advisory Panel
2. marine surveys
3. a program of inland surveys/studies
4. statistical/ data management support

There was extensive discussion of statistical power. The group felt that it was inappropriate to maintain a '5%' confidence level when considering a population in decline. The standard of statistical confidence should be relaxed (to avoid a Type II error). The exact level to be used should be determined from analyses of the offshore data.

Productivity data (the numbers of young produced) were also discussed. While these have value, and cost little to obtain (since marine surveys are being carried out for other reasons), their interpretation is fraught with difficulty at present.

The group also agreed that the areas north of the bioregion might serve as an appropriate control area. Since there is essentially no harvest going on in these areas, we may be able to compare areas, to determine the effect of harvest levels. This will give more power to analyses (comparison of control with 'experimental' areas).

The group also supported the continued study of RBV and similar metrics that will improve estimates of Take. It is essential to understand the levels of Take in order to make prediction for the monitoring program. Detrich also pointed out that improved metrics may allow further consideration of mitigation options.

The panel generally supported the following:
Compliance monitoring of harvest schedules and effects
Offshore monitoring of populations
Effectiveness monitoring of occupancy in MMCAs and Headwaters
Exploratory work to improve understanding of breeding habitat and its use (e.g. platform density, relationship of habitat characters to occupancy, radar, telemetry, predation risks).

Kareiva, who has reviewed many HCPs recently, noted that an essential component of a good HCP was that monitoring data should be freely available to everyone who is interested in them. This ensures credibility, and that the analyses are well carried out. Data accessibility has been a problem in the HCP. However it should be noted that raw data are essentially useless - very often they must be pre-digested through partial analysis, before another party could use them.

Murdoch stressed the need for independent scientists to carry out monitoring. The regulated party (PalCo) should not be carrying out the monitoring. Bacik stressed the company's willingness to work with outside scientists, as has been the case with the offshore surveys carried out by RSL.

The temporal scale of monitoring was discussed. Murdoch suggested that Ackakaya's models might have value in determining the expected period over which an effect on population trend would become visible.

Panel recommendations

The panel discussed among themselves the HCP and the science used in its development. They were asked to provide guidance on the design of monitoring, the appropriateness of take estimates, and invited to give their opinions on the HCP itself. Their opinions were recorded and presented by Cody (see appended summary).

Important points from their discussions include:

* The HCP is a compromise and should be presented as such.
* Mitigation that can be done, has been done
* The HCP itself is well presented and well organized, although some background information can be supplemented
* The science was used well, but gave ambiguous results; it is appropriate that the HCP at this point takes a conservative and cautious approach
* The PVA was a 'good go', but should not be (as it is not) a major guiding component, because of the ambiguities it reveals.
* The RBV approach could be refined, and would provide a batter and less conservative estimate of take than Reid's analysis,
* Science was essentially used to craft the 'worst case scenario', and to manage to that worst case. This was a sensible thing to do.

The panel generally thought that the HCP had done a good job of identifying areas for conservation value, and that the MMCAs were well placed.

Kareiva stated that, having reviewed approximately 50 HCPs recently, the PalCo HCP was 'one of the better ones' in its use of science.

4TH MEETING OF SCIENCE ADVISORY PANEL, SANTA BARBARA CA
MAY 26-27 1998

"MARBLED MURRELET HABITAT CONSERVATION PLAN"

NOTES DATED: May 29, 1998

COMPILED BY: M.L. Cody

FOR: SEI/Steven Courtney

RESPONSE OF PANEL TO MURRELET HCP AND GROUP QUESTIONS

A. General

We were asked for a general opinion of the HCP: was it a "good" plan, perhaps the "best" plan, and had the scientific data been utilized to best advantage? To the former, we could say only that, as it was a done deal, a compromise in which we were not party to the negotiations nor to what the options were, what was on the table, what was flexible and what was not, we cannot comment with any rational analysis. It seems worth saying somewhere that the HCP is a compromise, between on the one hand retaining as much Old-growth redwood as possible for conservation of Marbled Murrelet breeding habitat and on the other PalCo's continuing viability as a lumber concern. To the extent that the agreement is already forged, any mitigation that may take place is already described in the document. Nevertheless, it seems somewhat disingenuous to use phrases to describe the plan such as (p.1, HCP) "provide for improved conditions",... "survival of the species will be assured", ....and "minimize the effects on Marbled Murrelets." It is easy to envisage alternatives which would do these better, but perhaps those alternatives are not available due to constraints we do not know about.

With regard to the use of existing data and projections from it, we noted that the HCP relies to no great extent on existing data, rather adopts a conservative approach in the light of the wide range of uncertainties inherent in the murrelet data at all levels. We concurred that, given the uncertainty in e.g. PVA predictions, it is wisest not to employ these predictions specifically. The HCP is crafted in an appropriately conservative fashion, where worst case scenarios are assumed. We were asked whether any potential risk factors were omitted from the HCP, and note that the PVA mentions a variety of risk factors from the more to less deterministic, from incremental to catastrophic; we agree that it is difficult to be more specific about risk given the nature of the PVA predictions. Nevertheless, there are places in the HCP where more data might be cited to better effect. We noted for example that a) no relative worth is assigned to different habitat patches, and only area is used as a conservation criterion, although there are data that assign relative value to patches, via data on occupancy behavior, within proposed MMCA's and outside (i.e. Relative Bird Value of Ralph et al); b) the trend data are used (see below), but likely not with the most appropriate of analyses. In both cases, the HCP elects to use a 'worst-case' scenario.

Notably, the proposed MMCA's do appear to be those areas that should be highest priority for Marbled Murrelet breeding habitat conservation. In general, science was used appropriately in the HCP.

B. Impact

No mention is made of impact (as opposed to "Take"), perhaps justifiably so, because the PVA is equivocal on this matter. While the PVA suggests that no serious impact is likely, it nevertheless does describe a reduced probability of persistence with decreased carrying capacity. If carrying capacity is set by area/quality of breeding habitat, some impact is likely as a result of 25% or whatever reduction in breeding habitat.

C. Take

Estimations of Take are fraught with unknowns and complexity, given which we thought that the figures presented by T.Reid were reasonably derived and reasonably conservative. They provide an 'upper limit' or 'worst-case' estimate of take. We noted, however, that Marbled Murrelet populations are down perhaps 85-95% from pre-logging levels, and assuming linearity now between "take" of breeding habitat and "take" of Marbled Murrelet's is an act of faith. All sorts of non-linearities and second-order effects might be operating and thresholds approached: habitat "stepping stones", edge and fragmentation effects, critical levels re: facilitation, predation etc. These unknowns should be mentioned, and if a high-end figure of +/- 23% is to be cited it should be well justified.

D. Status

"Status" pertains to the present status of the Marbled Murrelet population and the extent and direction of its population trend, in s. Humboldt Co., in Zone 4 and at large. The best data available are the at-sea census results dating from about 1987. A figure of around -4 to 6% annual decline is cited in the HCP several times, together with a -30% cumulative decline in an appendix. The panel strongly supported better analysis of the at-sea data; P Kareiva believes there is justification to support a figure of -10% with narrow confidence limits, for a partial data set (to 1`995). There may be some evidence for non-linear trends (e.g. upturn following 1995). Log-likelihood estimation of trend should be conducted on the full data set (up to 1997) and separately for different sections of Zone 4, in particular for s. Humboldt Co. where some evidence argues for a steeper decline relative to other sections of Zone 4. This makes sense in terms of continued logging activity over the last decade in s. Humboldt Co. and little if any such activity further n. in Zone 4. If a trend of -10% or steeper is in fact detectable for s. Humboldt Co., it would behoove the HCP to employ such a figure, as trends in the immediate future will likely be more steeply negative than they are now, and their comparison with -4% would be foolhardy if poorly justified.

For a broader evaluation: a) There might be a possibility of relating the trends determined by at-sea censuses with logging and habitat removal ashore, within Zone 4. b) The at-sea data have never been standardized by indices of detectability, although the data for such corrections are apparently in hand; this might enhance the worth of the trend data and confidence with which they might be invested. c) No use is made of the ratios of juveniles to adults (J/A) at sea. These ratios might still be very useful indices of Marbled Murrelet productivity, especially if the denominator is spring adult numbers. More should be made of these data; however extensive discussion at the meeting suggested these data were of uncertain value d) We have witnessed several strong El Niño events in the last decade, and ENSO years are reported to be poor for Marbled Murrelet breeding success. No analysis of ENSO on either J/A or trend data has been forthcoming, and this might resolve some of the variance in the at-sea numbers.

All data should be used; it is essential that these be publicly available (e.g. on a website). This will build public confidence, and scientific credibility, and assure that the analysis is completed.

D. Monitoring

In summary, the panel recommends that
1. An independent Monitoring Data Manager/Overseer/ Analyst (MDM/A) be employed, in charge of the design, execution and analysis of the monitoring program;
2. Some form of "compliance" monitoring we assume will be a routine part of the program: the removal of potential or actual breeding Marbled Murrelet habitat over space and time, detailed by stand size, location, age, composition etc. will be required;
3. Marbled Murrelet population should continue to be monitored at-sea, possibly with some refinement but in a fashion such that data comparable to those already in hand are collected;
4. There should be formalized monitoring of Marbled Murrelet activity at actual or potential inland breeding sites.

In more detail,
1. One of the obvious deficiencies of monitoring efforts to date is a wide gap between actual data collection and the completed analysis and summary of such data in the hands of the agencies that need them (and in part paid for them). We have never, as Murdoch emphasized, been treated to a comprehensively managed data set for any aspect of the program. There have been delays in producing the data, i.e. having them available in lucid form for discussion and deployment, and also in their full and statistically defensible analysis. A competent and vigorous MDM/A would go a long way toward solving this chronic problem; it might also obviate the necessity of using outside statistical consultants who are unacquainted with the data collection and Marbled Murrelet biology, though such consultants might still be approached when needed. Given such a person, his/her talents might be employed immediately, as data in hand have been under-analyzed, and the design of monitoring protocols for the next 5-10 years need to be worked out in detail over the next 6 months. A Ph.D. Marbled Murrelet biologist with broad statistical expertise would be best for the position. Data standards and documentation needs should be identified immediately. Timeliness of data transfer should be built into any contracts. The independence of the MDM/A is important. PalCo should ensure capable, independent scientific analysis.

2. Needs no further amplification.

3. Comments on the at-sea data set have been made above (see Trends). As far as we know, there are no better ways to monitor Marbled Murrelet numbers and trends that those which have been employed for many years by RSL; indeed, we think it likely that the data are best collected by RSL in the same way as before. We note, however, that it is essential the MDM/A has immediate access to the full data set and is overseer to its organization and analysis. A framework not only for the data collection protocols (where, when, how often, how made etc.) but for its transit to the MDM/A, assembly, display, analysis, results, conclusions, predictions etc. are required. Continuation of the at-sea counts for at least another decade will be necessary, and for agency purposes perhaps for much longer. Distance data should be used, and density estimated using variable-width methods. Corrections for detectability should be incorporated, as in the inland data. A population trend model (based on harvest schedules and Ackakayya's models) could be used in predicting trends.

4. Regarding Marbled Murrelet monitoring at inland breeding sites, again we have no basis on which to second-guess the way in which monitoring has been conducted to date; the "occupancy" criteria seem the best available (and practical) for assessing Marbled Murrelet use of actual or potential breeding habitat. Such monitoring should be continued in State Parks, Headwaters Reserve, and designated MMCA's; we see no utility in monitoring at Old-growth Redwood residuals slated for harvest. Design protocols need to be finalized for the inland monitoring program, and we note that the analysis of data collected to date should have a bearing on these protocols. For example, what is the relation between "occupancy behavior" and stand characteristics such as tree height and density, potential nesting platforms, epiphytes, predator density etc.? Then, in turn, what is the relation between stand characteristics thought conducive to Marbled Murrelet breeding and the actual productivity of the site? This is much more difficult, but seems tractable via 1.monitoring nest-leaving juveniles, 2.location of recently fledged juveniles at sea, 3.use of radar, etc. The data analysis by G. White was useful; its result could be both extended and used. The distinction between detection of occupancy by stand characteristics and differences in occupancy behavior by stand characteristics needs to be sharpened. Then, if detection is very much higher during certain times of the year, i.e. around Julian date 200, perhaps a time when demands of large young for food are maximal, then perhaps monitoring during only those periods is required. The RBV approach of RSL should be further explored, so that relative conservation value of stands can be better evaluated.

There are other aspects of Marbled Murrelet biology that might well be revealed as by-products of the monitoring program, and others that remain as unknown but potentially important; all seem well worth pursuing in order to generate a better picture of Marbled Murrelet biology overall. Some of these data may be collected with little or no additional effort in the Monitoring program. Some of the factors mentioned were:

a) Landscape issues. Originally the PVA was to be integrated, as a second step, with "landscape factors", such that with a better input from the disposition of size, location, stand composition, etc. of breeding habitat the metapopulation results might be sharpened. This has not been done; and it is not clear whether it is worth doing for the sole purposes of the PVA, since metapopulation models assume habitat patches contain all resources required for population growth and have patch size- and location-specific rates of emigration, immigration, extinction and colonization. Clearly patchy breeding habitat for Marbled Murrelet means something quite different, but still a study of these landscape issues may be a productive line of inquiry. Edge and fragmentation effects, the spatial arrangement of patches relative to each other and relative to sea-inland flyways, etc., are landscape factors whose importance or relevance is not yet known.
b) Old-growth Douglas Fir. The estimation of "take" has no component attributable to removal of Old-growth Douglas Fir. This is because no Marbled Murrelet occupancy behavior has been observed in Old-growth Douglas Fir in s. Humboldt Co. We note that elsewhere in CA, as certainly elsewhere in Zone 4 and the 3-state area of concern, Old-growth Douglas Fir is used as breeding habitat, sometimes the preferred habitat. Nearly all of the Old-growth Douglas Fir is outside the MMCA's and slated for harvest. How confident are we that this habitat is not Marbled Murrelet breeding habitat? Is the habitat well-monitored? If it is not breeding habitat, why not? Might it become breeding habitat when Old-growth Redwood declines in area? This seems a worthwhile line of questioning,
c) Human activity. There has been no critical assessment of the effects of human activity on Marbled Murrelet occupancy behavior or breeding success, although it seems that some recognition of these effects is pervasive. Could more be accomplished here?
d) Other factors. From at-sea (more information of El Niño effects, fish densities and locations, foraging areas, better statistical treatment of census data) to in transit (are Marbled Murrelet's flyway or access limited in their approach to or use of breeding habitat?) to inland sites (predators and their regulation, artificial nests or nest platforms? rate at which new breeding habitat anticipated to come on line for Marbled Murrelet) there is a number of factors which seem worthy of further study, some of which will come from monitoring and some not. A good MDM/A appointment would be invaluable here.