[Federal Register: March 21, 2002 (Volume 67, Number 55)]
[Rules and Regulations]               
[Page 13095-13098]
From the Federal Register Online via GPO Access [wais.access.gpo.gov]



Fish and Wildlife Service

50 CFR Part 17

RIN 1018-AH73

Endangered and Threatened Wildlife and Plants; Re-opening of 
Comment Period on the Sacramento Splittail Final Rule

AGENCY: Fish and Wildlife Service, Interior.

ACTION: Final rule; re-opening of comment period.


SUMMARY: We, the U.S. Fish and Wildlife Service (Service), announce the 
re-opening of the comment period for the final rule on the Sacramento 
splittail (Pogonichthys macrolepidotus). Comments previously submitted 
need not be resubmitted as they will be incorporated into the public 
record as part of this re-opened comment period, and will be fully 
considered in the final rule. We are re-opening the comment period to 
invite comments and to obtain peer review on the statistical analysis 
completed by us to re-analyze the available splittail abundance data. 
We are also inviting additional comments on the status of and factors 
affecting the species, as first solicited in the January 12, 2001 (66 
FR 2828), comment period and re-solicited in the May 8, 2001 (66 FR 
23181), and August 17, 2001 (66 FR 43145), re-openings of same.

DATES: We will accept public comments until October 15, 2002.

ADDRESSES: Comment Submission: If you wish to comment, you may submit 
your comments and materials concerning this proposal by any one of 
several methods:
    1. You may submit written comments and information by mail to the 
Field Supervisor, Sacramento Fish and Wildlife Office, U.S. Fish and 
Wildlife Service, 2800 Cottage Way, Suite W-2605, Sacramento, 
California 95825.
    2. You may hand-deliver comments to our Sacramento Fish and 
Wildlife Office, during normal business hours, at the address given 
    Comments and materials received will be available for inspection, 
by appointment, during normal business hours at the address under (1) 

FOR FURTHER INFORMATION CONTACT: For general information, Susan Moore, 
at the above address (telephone 916/414-6600; facsimile 916/414-6713).



    The Sacramento splittail (hereafter splittail) represents the only 
extant species in its genus in North America. For a detailed 
description of the species, see the Recovery Plan for the Sacramento/
San Joaquin Delta Native Fishes (Service 1996), references within that 
plan, and Moyle et al. (2001 in prep.).
    Splittail are endemic to certain waterways in California's Central 
Valley, where they were once widely distributed (Moyle 1976). Splittail 
presently occur in Suisun Bay, Suisun Marsh, the San Francisco Bay-
Sacramento-San Joaquin River Estuary (Estuary), the Estuary's 
tributaries (primarily the Sacramento and San Joaquin rivers), the 
Cosumnes River, the Napa River and Marsh, and the Petaluma River and 
Marsh. The splittail no longer occurs throughout a significant portion 
of its former range.
    Pursuant to the Endangered Species Act of 1973, as amended (Act), 
the splittail was listed as a threatened species on February 8, 1999 
(64 FR 5963). In this previous listing determination, we found that 
changes in water flows and water quality resulting from export of water 
from the Sacramento and San Joaquin rivers, periodic prolonged drought, 
loss of shallow water habitat, and the effects of agricultural and 
industrial pollutants were significant factors in the splittail's 
    Subsequent to the publication of the final rule, plaintiffs in the 
cases San Luis & Delta-Mendota Water Authority v. Anne Badgley, et al. 
and State Water Contractors, et al. v. Michael Spear, et al. commenced 
action in Federal Eastern District Court of California, challenging the 
listing of the splittail as threatened, alleging various violations of 
the Act and of the Administrative Procedure Act (5 U.S.C. 551 et seq.). 
We, as directed by the court, and pursuant to the Act, provided notice 
of the opening of a comment period regarding the threatened status for 
the splittail, from January 12, 2001, to February 12, 2001 (66 FR 
2828). In addition, we re-opened the comment period on two additional 
occasions; from May 8, 2001, to June 7, 2001 (66 FR 23181), and from 
August 17, 2001, to October 1, 2001 (66 FR 43145). We are now re-
opening the comment period for a fourth time to obtain peer-review and 
public comment on the statistical analysis used to analyze the abundant 
data available for splittail, and to seek public comment on the status 
of the species (as first solicited in 66 FR 2828). Upon the close of 
this comment period, we will make our determination whether the 
splittail warrants the continued protection of the Act.

[[Page 13096]]

    The approach currently used by us to analyze the best 
scientifically and commercially available splittail abundance data 
differs from methods employed previously. In the February 8, 1999, 
final rule and the January 12, 2001, and May 8, 2001, re-openings of 
the comment periods, we relied primarily on the unstratified Mann-
Whitney U-test approach utilized by Meng and Moyle (1995), first 
published in the Transactions of the American Fisheries Society. See 66 
FR 2828 for a complete description of the Meng and Moyle (1995) method. 
In the August 17, 2001, re-opening of the comment period, we employed 
permutation-based exact calculations of p-values for stratified Mann-
Whitney U-tests to analyze data derived from the Meng and Moyle (1995), 
Sommer et al. (1997), and California Department of Fish and Game (CDFG) 
methodologies. We also employed a polynomial regression model and a 
crude exponential decay analysis in the August 17, 2001, comment 
period. See 66 FR 43145 for a complete description of the revised 

Statistical Analysis of Multiple Linear Regression Model

    We have carefully considered all comments and responses. In regard 
to the analysis of splittail population trends, we now employ a 
statistical analysis of an abundance index and Multiple Linear 
Regression (MLR) model jointly developed and submitted by the CDFG 
(Rempel 2001) and the United States Bureau of Reclamation (USBR) 
(Michny 2001). The model is hereafter referred to as the CDFG/USBR MLR 
model and provides the most sound basis, to date, for statistically 
evaluating temporal trends of splittail abundance data.
    The CDFG/USBR MLR model includes HYDROLOGY and TIME (year) as 
independent variables and ABUNDANCE INDICES as the dependent variable. 
It also incorporates corrected splittail abundance data (Rempel 2001). 
We consider this statistical approach superior to the previous practice 
of using unstratified Mann-Whitney U-tests (Meng and Moyle 1995; Sommer 
et al. 1997) because it does not require arbitrarily dividing an 
inherently continuous data set into ``before'' and ``after'' categories 
(see previous discussion of this issue in 66 FR 43145). We also 
consider the CDFG/USBR MLR model superior to the permutation-based, 
exact calculations of p-values for stratified Mann-Whitney U-tests 
discussed in 66 FR 43145 because of substantive scientific issues 
raised by Rempel (2001), Michny (2001) and others, specifically, that 
such an analysis inappropriately combines results from differing survey 
methods (i.e. midwater trawl, otter trawl, beach seine, salvage) and 
considers primarily adult age class splittail. We further consider the 
CDFG/USBR MLR model superior to the polynomial regression model 
presented in 66 FR 43145 because existing abundance index monitoring 
programs have not been conducted for a sufficient duration to provide 
for reasonably conclusive application of the polynomial model (as 
concluded in 66 FR 43145). We also support use of the CDFG/USBR MLR 
model because of the facility with which it can be applied to all sets 
of splittail age class data from all seven abundance monitoring data 
sets (a total of 20 discrete sets of age-specific abundance monitoring 
data). Lastly, we have omitted the exponential decay model found in 66 
FR 43145 because: (1) It was found by respondents to be insufficient to 
describe interactions in a complex aquatic ecosystem; and (2) the CDFG 
Mann-Whitney U-test results upon which the exponential decay 
calculation was based have since been superceded by the CDFG/BOR MLR 
    The CDFG/USBR MLR model explicitly controls for potential 
confounding effects of hydrological year type, the factor that is 
nearly unanimously viewed as the single strongest predictor of 
splittail year class strengths (e.g., Moyle et al. 2001 in prep.), by 
utilizing the number of days total delta inflow (DAYFLOW, California 
Department of Water Resources) exceeds 1,558 cubic meters per second 
(cms) (55,000 cubic feet per second (cfs)) during the February through 
May spawning/rearing period as a predictor (independent variable). This 
is conceptually comparable, yet superior, to the stratified Mann-
Whitney U-tests presented in 66 FR 43145, which also controlled for 
hydrological year type. There is, however, one potentially important 
assumption associated with the CDFG/USBR MLR model that remains 
untested: The assumption that there is a lack of interaction between 
the HYDROLOGY and TIME variables. The CDFG/USBR MLR model assumes that 
the long term probabilities of high and low Delta inflow years are not 
systematically changing over time. If in fact those probabilities are 
systematically changing over time (due to either changing climate or 
changing water management policy), the coefficients for the TIME 
variable would be incapable of detecting the influence of the 
potentially changing HYDROLOGY component of splittail abundance trends. 
We believe this assumption can and should be tested against existing 
longitudinal hydrological data bases, with future changes to be 
determined once: (1) Sufficient splittail abundance data exist to 
ensure conclusive application of the polynomial model (i.e. multiple 
peaks and troughs); and (2) the cumulative expected hydrologic effects 
of potential large-scale water resource projects (i.e. potential 
projects such as the Folsom Dam reoperation and height increase, Shasta 
Dam height increase, Sites Reservoir, Colusa Basin off stream storage, 
increased pumping at export facilities, etc.) are more clearly 
understood. These potential future actions, and possibly long term 
climate changes, may appreciably change the timing, duration, magnitude 
and/or frequency of floodplain inundation within the splittail's range, 
thus influencing future population trends.

Discussion of CDFG/USBR MLR Model Results

    The TIME variable captures temporal trends in the population index 
data. Its regression coefficient will be negative if splittail 
abundance is trending downward over time and positive if splittail 
abundance is trending upward over time. The probabilities of any given 
coefficient reflecting a true nonzero time trend are 1-p, where p is 
the standard statistical probability for the null hypothesis (of a zero 
trend). Thus, a ``p-value'' of 0.05 would be the same as a 95 percent 
probability that the corresponding TIME coefficient reflects a true 
nonzero downward or upward trend in splittail abundance. Results of the 
CDFG/USBR MLR model as presented by Rempel (2001: Table 3) for CDFG and 
Michny (2001: Table 1) for USBR reveal that 14 of 20 abundance 
monitoring data sets for splittail show downward trends (i.e., have 
negative coefficients for the TIME variable). In addition to a high 
frequency of negative coefficients that would be highly unlikely by 
chance alone (exact one tailed p = 0.0577; binomial test, 
H0=0.50, N1=6, N2=14)(StatXact 4: 
CYTEL Software Corp. 2000), the median (middle value) probability of 
nonzero negative trends (0.81 or 81 percent) is also clearly greater 
than the median probability of nonzero positive trends (0.59 or 59 
percent) (Figure 1 below), to an extent that would be highly unlikely 
by chance alone (exact one tailed p= 0.0303, Wilcoxon-Mann-Whitney 

[[Page 13097]]

Figure 1.

    All four coefficients for the TIME variable that exceed a 95 
percent probability (classic 0.05 alpha level statistical significance 
criterion) for a true nonzero trend are negative. A fifth negative TIME 
variable coefficient is nearly statistically significant (p=0.057). Due 
to the very limited statistical power associated with the abundance 
monitoring data sets for splittail (see discussion of this topic in 66 
FR 43145) there is substantive bias in favor of type II statistical 
error, i.e., failing to correctly reject the null hypothesis (of no 
time trends). Low statistical power is not unique to the splittail data 
sets. Due to the inherent high variability in fisheries and wildlife 
abundance data, for applied purposes (such as detecting oil spill 
impacts on marine bird populations) it has become ``customary'' to use 
an alpha level of 0.20 (i.e., an 80 percent probability of true nonzero 
trends) for statistical tests of population trends over time (Day et 
al. 1997; Murphy et al. 1997; Irons et al. 2000; Wiens et al. 2001). 
Wiens et al. (2001:890) further state that even using an alpha level of 
0.20, ``* * * there remains the question of how blindly one should 
follow the results of statistical (significance) testing.'' 
Furthermore, Taylor and Gerrodette (1993) persuasively argue that 
because of the low statistical power that is so often characteristic of 
abundance monitoring data sets for rare species, ``* * * detection of a 
[statistically significant] decline should not be a necessary criterion 
for enacting conservation measures * * *.'' Referring to the management 
of a rare species of porpoise, the vaquita, Taylor and Gerrodette 
(1993) caution that due to the low statistical power of abundance 
monitoring data, ``* * * if we were to wait for a statistically 
significant decline before instituting stronger protective measures, 
the vaquita would probably go extinct first.'' Although splittail are 
not as rare as vaquita, the ``boom-or-bust'' reproductive biology of 
splittail results in such high-variance abundance monitoring data that 
the limitations on statistical power are as severe as Taylor and 
Gerrodette (1993) encountered with the vaquita. We must therefore take 
into consideration the issue of statistical power when interpreting the 
splittail abundance data. We accomplish this by evaluating all trends, 
not just the trends that meet traditional (p=0.05) criteria for 
statistical significance. Those traditional criteria assume a much 
higher standard of statistical power than the splittail data are able 
to meet. The inherent difficulty in effectively surveying splittail is 
likely to result in considerable scientific uncertainty and low 
statistical power to detect actions' effects. Recent studies indicate 
that these constituents, and the uncertainty and risk associated with 
them, favor a precautionary approach (Thompson et al. 2000, Slooten et 
al. 2000). Under such circumstances, and given the intrinsically 
precautionary nature of section 4 of the Act, we must consider the 
preponderance of the data, including both statistically significant and 
insignificant trends. Of 14 negative coefficients, 7 have a probability 
of 80 percent or greater (p=0.20) to reflect true nonzero downward 
trends in splittail abundance. Of 6 positive coefficients, 0 (none) 
have a probability of 80 percent or greater to reflect true nonzero 
upward trends in splittail abundance. This asymmetry in the results is 
highly significant (exact one tailed p=0.022, Fisher's exact test, 
``mid p'' corrected) (StatXact 4: CYTEL Software Corp. 2000) and 
clearly indicates a preponderance of data consistent with an 
``apparent'' declining trend in splittail abundance.
    The four highest, statistically significant (at traditional levels) 
probabilities of a nonzero downward splittail population trend are 
exhibited by the Suisun Marsh survey (Age-0 and adult) and in the data 
collected via fish salvage operations at the State Water Project (SWP) 
Skinner Delta Fish Protective Facility (Age-1, and Age-2 and greater). 
The decline evident in the Chipps Island Trawl (Age-2 and greater) is 
nearly statistically significant at traditional levels (94.3 percent 
probability). Two additional probabilities of a nonzero downward 
splittail population trend are evident at the 80 percent probability 
level; Chipps Island Trawl (Age-1) and SWP (Age-0).
    We fully concur with the statements of various respondents that 
abundance monitoring data for splittail have

[[Page 13098]]

methodological weaknesses of one sort or another; none of the surveys 
were designed specifically to rigorously estimate splittail population 
numbers (see Moyle et al. 2001 in prep.; Meng and Moyle 1995; and 
Sommer et al. 1997 for descriptions of surveys). However, existing data 
sets do constitute best available scientific information for the 

Public Comments Solicited

    We will accept written comments during this re-opened comment 
period, and comments should be submitted to the Sacramento Fish and 
Wildlife Office as found in the ADDRESSES section.


    The primary authors of this notice are Jason Douglas and Joseph 
Skorupa (see ADDRESSES section).

    Authority: The authority for this action is the Endangered 
Species Act of 1973, as amended (16 U.S.C. 1531 et seq.).

    Dated: March 14, 2002.
Steve Williams,
Director, Fish and Wildlife Service.
[FR Doc. 02-6803 Filed 3-20-02; 8:45 am]