HARVARD UNIVERSITY Ernst Mayr Library of the Museum of Comparative Zoology DEC 4 2006 "^fes 'S/ry Y JUL 2 9 TAXONOMY OP Y THE MARSH WREN m SOUTHERN CALIFORNIA by Philip Unitt, Karen Messer, and Marc Thery No. 31 31 May 1996 Proceedings of the San Diego Society of Natural History r ISSN 1059-8707 PROCEEDINGS :^%^ of the ^^( 2 9 San Diego Society of Natural History 1996 Founded 1874 ''^^1^° Number 31 31 May 1996 Taxonomy of the Marsh Wren in Southern California Philip Unitt San Diego Natural Hisloiy Museum. P. O. Box 1390. San Diego, California 92 112 Karen Messer Departmenl of Mathematics. California State University, Fullerton, California 92634 Marc Thery Centre Nationale de la Recherche Scientifique, URA 1 1 S3. Museum National d'Histoire Naturelle, Lahoratoire d'Ecologie Generale, 4 Avenue du Petit Chateau. 91800 Brunoy. France ABSTRACT. — At least three subspecies of the Marsh Wren (Cistnlluinis /xj/i/.t/n.t) occur in southern California C /) plesimlpuhenus is a widespread migrant and winter visitor and breeds in the Owens Valley south to Owens Lake It is charactenzed by large size, pale tawny rump and scapulars, a nearly white breast, mostly brown crown, and bold white back streaks It amves m its winter range mainly in mid to late September and departs mainly in April. A distinction between plesius and puherius seems likely but was not examined in this study. C. p aestiuinmis is resident from the Colorado River and Imperial Valley northwest locally through the Mojave Desert, and along the coast from Ventura County north through northern California to southwestern Oregon. It is charactenzed by its medium size, moderately dark rufous rump and scapulars, a crown partly black and partly brown, and narrow to moderate white back streaks. Its breast ranges from moderately buff (usually in the Colorado Desert; frequently in the Mojave Desert and along the coast) to deep buff or brown (usually in the Sacramento Delta; frequently in the Mojave Desert and along the coast) A distinction between aesluannm and deserticola could be maintained if specimens from the type localities only (Sacramento Delta and Imperial Valley, respectively) were considered, but the area of intergradation is so large (Mojave Desert, coastal central and northern California) that the distinction does not seem broadly useful The resident population of coastal southern California (Los Angeles to San Diego counties) is differentiated by its small size, largely black crown, and deep rufous rump and scapulars At least 94% of specimens of this population can be distinguished from all other Marsh Wren subspecies, so we propose it be known by the new name Cistolhoni.s paluslris ilarkae. No California Marsh Wren population agrees with that of southwestern Washington, with its largely dark brown crown, broad brown nape collar, and consistently narrow whitish streaks on a reduced brownish black back patch That subspecies, paludicola. extends south only to northwestern Oregon INTRODUCTION ranges in v/idth from practically zero to about 9 mm. The patch on the back ranges from small and brownish black with obscure The subspecies of the Marsh Wren (Cistothorus palustns) in whitish streaks to large and completely black with white streaks California were reviewed most recently by Rea (1986). The fonnat as broad as the black between them. The scapulars, rump, under of that review (Phillips 1986), however, did not allow a species with tail coverts, and background color of the tail range from me- variation as complex as the Marsh Wren's to be analyzed as thor- dium buffy brown to deep rufous. The underparts range from oughly as appropriate. Also, specimens collected over the last ten practically white to mostly buff or brown with only small pale years now allow that variation to be described more accurately than patches on the throat and in the center of the belly. The variation was previously possible. Here we attempt to describe the Marsh in size is most easily expressed by wing length. Wren's vanation as it relates to southern California and to apply the information to understanding of the species' distribution and migra- HISTORICAL BACKGROUND tion. Our central goal is to place the resident Marsh Wrens of coastal southern California within the framework of the species' variation. Ridgway (1904) listed two subspecies of the Marsh Wren for We also attempt to describe the Marsh Wren's southern California California, the smaller, darker paludicola Baird, 1858, from the distribution in detail, as this is still poorly understood [e.g., the map "Pacific coast district." west of the Cascade Range and Sierra Ne- in Zeiner et al. (1990) has many inaccuracies]. vada (type locality Shoalwaler [now Willapa] Bay, Pacific County, In western North America, various populations of the Marsh Washington), and the larger, ^dXtx plesius (Oberholser, 1897) from Wren differ in color, pattern, and size. The crown ranges from the "Rocky Mountain plateau district," west to northeastern Califor- entirely brown to black with only a small brown patch in the nia (type locality Fort Wingate, New Mexico, in the winter range), center of the forehead. The dark crown patch is separated from Gnnnell ( 1903) first recognized />/f.v;n.v as a winter visitor to coastal the black and white patch on the back by a brown collar, which southern California. Philip Unitl, Karen Messer, and Marc Thery Swarth (1917) described aesluannus, type locality Grizzly Is- land, Solano County, California, as darker still than paludicola but as large as plesius. He ascribed to aestuarinus a range extending Crom Suisun Bay through the San Joaquin delta and valley, with scattered specimens of migrants trom more coastal localities. Cistothorus p. paliuiicola he believed resident along California's entire coastline, extending inland in the lowlands of southern California, though he cited a few inland specimens from northern California as well. Grinnell and Miller (1944) followed Swarth (1917) closely, desig- nating coastal specimens of aestuarinus and inland specimens of paludicola as migrants. They extended the breeding range of aestuarinus to the Imperial and Colorado River valleys. Aldrich (1946) subdivided plesius, describing a duller, less rufescent subspecies pulverius (type locality Sprague, Lincoln County. Washington). He ascribed lo pulverius a breeding range from eastern Washington south to northeastern California and northwest- em Nevada. The fifth edition of the American Ornithologists' Union (1957) checklist followed Grinnell and Miller (1944) for the dark lowland forms but did not distinguish pulverius. Phillips et al. ( 1 964) took an even more conservative approach, reverting to Ridgway (1904) in recognizing only one dark lowland subspec\ef,, paludicola, and one pale plateau subspecies, /)/<\v/h.?. Mon.son and Phillips ( 198 1 ), however, accepted pulverius and equivocated on aestuarinus. Not until Phillips (19X6) was the species as a whole revised again. In this work, Rea contnbuted the characterizations of the dark lowland subspecies of western Marsh Wrens, while Phillips himself covered the remainder of the species. Phillips recognized both pulverius and plesuts, while Rea recognized aestuarinus, segregated the desert population from aestuarinus as deserlicola, and suspected the coastal California population to represent an undescribed subspe- cies, "brighter and richer" than paludicola. aestuarinus, or deserticola. Figure 1 shows the general breeding distribution of the Marsh Wren in western North America, the type localities of the named subspecies addressed m this study, and the sites in coastal southern California where Marsh Wrens were collected for this study. METHODS Specimen Resources The taxonomic identity of the Marsh Wrens of coa.stal California cannot have been adequately assessed previously because of a dearth of specimens known to represent the local breeding populations. Migrants of pulverius and plesius, especially the latter, are common in southern California in winter, and their arrival and departure dates have not been known precisely. The ideal specimens on which to base a taxonomic study of the Marsh Wren should be collected as soon as possible after a complete molt and therefore with the least worn plumage with its maximum genetically determined information content. The birds' constant contact with rough, often damp vegetation wears and stains the plumage more than in many species. Though in the eastern United States the Marsh Wren has a complete prealtemate (prenuptial) molt in spring (Kale 1966). such a molt is not common in the West. Of 1 1 February-May specimens from southern California examined dur- ing this study, all are too worn to have undergone an extensive recent molt, and of the three whose molt status was described on their label, all specified "no molt." Likewise two March specimens from south- western British Columbia [San Diego Natural History Museum (SDNHM)] are rather worn and not molting. Of eight May-June specimens from southeastern Oregon, at least four are too worn to have molted recently. But three February-March specimens from Solano County, California (Carnegie Museum of Natural History, Pittsburgh) are molting their throats, and one of these was replacing its tail (K. C. Parkes pers. comm.). Since a prealtemate molt in western Marsh Wrens is rare or at least inconsistent, we based our study on specimens collected shortly after the prebasic molt, in late summer or fall. For the resident Marsh Wrens of southem California, therefore, the ideal specimens are those collected after the completion of molt but before the amval of migrants. For such a sample, in 1994 Unitt mist-netted Marsh Wrens at three sites along the coast of San Diego County where the species was known to breed. Birds still in juvenal plumage or heavy molt were released, while those whose molt was nearly or quite completed were collected and prepared as study skins. The field work began on 23 August and ended on 7 October, when pale migrants had cleariy arrived and were outnumbering the dark local birds. In all, 24 specimens from coastal San Diego County were collected for this study. Since 1984. for basic data on land birds' distribution and migra- tion. Roger Higson and Unitt have been collecting in the Imperial Valley, southeastern California, whose avifauna has not been studied in detail. Over 12 years, we have accumulated 33 specimens of the Marsh Wren, on dates from 4 August to 12 February. This sample includes both representatives of the local population and migrants from farther north. In October 1984 and September 1986, Amadeo Rea, Kem Hainebach, and Unitt visited Grizzly Island and adjacent Joice Island in the delta of the Sacramento River, collecting 30 Marsh Wrens around the type locality of aestuarinus ( 12 in October 1984. 18 in September 1986). On 12 September 1986. we also took two at Gray Lodge State Wildlife Area. Butte County. For a sample of nominate paludicola, we borrowed from the Burke Museum, University of Washington (UW), 20 specimens collected in Pacific and Grays Harbor counties, Washington, on 20 September 1985, 18 November 1985, and I and 2 October 1986. To gain a broader sample from southern California, we borrowed 13 specimens from the San Bemardino County Museum (SBCM), 41 from the Los Angeles County Museum of Natural History (LACM). 4 from the Santa Barbara Museum of Natural History (SBMNH), 2 from the Museum of Systematics and Ecology, University of Cali- fornia, Santa Barbara (UCSB), and 1 from the Museum of Vertebrate Zoology, University of California, Berkeley (MVZ). The sample of specimens on which we base our characterizations of Marsh Wren subspecies were all collected within the past 15 years. We excluded older specimens because comparison of new and old specimens revealed obvious foxing of specimens collected in the first half of the century: among the darker individuals, the rump and scapulars of all the old specimens were more mfous than the recently collected ones. Because mmp and scapular color appeared to be an important variable, and we know of no means for correcting for this shift, widespread among birds, we used only the recent specimens. Once the framework was established with the recent specimens, we used a few older specimens to augment our geographical and histori- cal perspective. Character Assessment The features in which the Marsh Wren varies geographically differ qualitatively, so they had to be assessed in vanous ways. AH measurements and visual assessments were made by Unitt. Wing chord was measured to the nearest 0.1 millimeter Body size, as reflected in wing length, is the only feature in which the sexes differ. The mean wing chord of males, in the entire sample of specimens, was 1 .06 that of females. Therefore, so that the sexes could be combined in the statistical analysis, we multiplied the measurements of the females by 1 .06 to yield an adjusted result, which we used in the remainder of the analysis. In support of this adjustment, Messer ran a two-way analysis of variance on our "enlarged" sample (see below under Categorization of specimens for analysis) with wing length as the response and population of origin and sex as the explanatory variables. Residual plots showed the data to be approximately normally distributed; the ratio of the smallest Taxonomy of the Marsh Wren in Southern Cahfornia Figure 1 . Approximate breeding distnbution ofthe Marsh Wren in western North Ainenca (shading). In many areas within this range, the species occurs in only scattered localized colonies. •, type localities of subspecies addressed in this study; ▲. sites in coastal southern California where Marsh Wrens were collected for this study. (0.72) to the largest (2.12) standard deviation was less than .^, with most sample sizes between 10 and 15. The two smallest groups had standard deviations near I. There was no significant interaction between the sex effect and the population-origin effect (/> > 0.05), justifying the use of a single adjustment for sex across all groups. Had we used the estimated main effect for sex from this analysis in our adjustment, the factor would have been 1 .077. a change of 1 .6%. We did not feel this would affect our results materially and retained the original adjustment for simplicity. A multiplicative rather than additive factor was used to bnng the females' standard deviation (2.01 in the enlarged sample) clo.ser to the males' standard deviation (2.30 in the enlarged sample). The width of the brown nape collar (between the black of the crown and the black of the back) was measured to the nearest Philip Unitt, Karen Messer. and Marc Thery millimeter. In specimens in which the rear of the crown is brown, the crown is a darker shade than the nape collar, and in these the measurement was made from this fairly abrupt transition. The width of the collar is affected to some extent by the amount to which the neck IS elongated or compressed when the specimen is prepared. A few specimens were not scored for this variable because they were poorly made or had lost neck feathers. The great majority of the California specimens were prepared by Unitt; preparation by a single technician enhances uniformity. Nape-collar width is a significant variable only in comparisons of the broad-collared puhidicola from coastal Washington with other populations. The sample of paliulicola consisted of specimens with naturally proportioned necks all beauti- fully made, largely by C. S. Wood. The vanation in crown pattern was assessed by ranking each specimen on a scale from 1 to 6, in comparison to six specimens serving as standards. In category 1 (standard SDNHM 43970) the crown is essentially entirely brown, with only a few black feathers at the sides. In category 2 (standard SDNHM 44592) the crown is mostly brown, with some black along the sides. In category 3 (standard SDNHM 44532) the black extends around the rear of the crown as well as along the sides. In category 4 (standard SDNHM 48937) the crown is about half brown and half black. In category 5 (standard SDNHM 48932) the crown is mostly black with some brown extending from the forehead into the center In category 6 (standard SDNHM 48982) the crown is black with only a .small brown patch on the forehead. The variation in back pattern we assessed by ranking each speci- men on a scale from 1 to 4. In category 1 (standard UW 40570) the back is brownish black with very narrow dull whitish streaks. In category 2 (standard SDNHM 43379) the back is deeper black with whiter but still narrow streaks. In category 3 (standard SDNHM 48992) the back is deep black with white streaks broader than in category 2 but still narrower than the intervening black streaks. In category 4 (standard SDNHM 42843) the pure white and pure black streaks are of about equal width. The variation in the color of the scapulars and rump and of the underparts we assessed by two methods. First, Unitt ranked the color of the scapulars and rump on a .scale from 1 to 10. Category 1 (standard SDNHM 43469) corresponds to a medium tawny brown, close to color 26, Clav Color, of Smithe ( 1975). Categories 2 (stan- dard SDNHM 43972) and 3 (standard SDNHM 44592) are .some- what darker, the latter close to Smithe's color 121C, Mikado Brown. Category 4 (standard SDNHM 48954) is close to color 223, Verona Brown. Categories 5 (standard SDNHM 47685) and 6 (standard SDNHM 43456) are darker yet, the latter close to color 121B, Brussels Brown. Categones 7 (standard SDNHM 48937) and 8 (standard SDNHM 48938) are a deeper cinnamon-rufous, category 8 being close to Smithe's color 23, Raw Umber. Finally categories 9 (standard SDNHM 4898 1 ) and 1 0 (standard SDNHM 489 1 2) are the darkest rufous, the latter close to color 1 2 1 A, Front's Brown. Later, Thery and Unitt measured the retlectance spectrum of the rump of each specimen in percentage of a Spectralon (Ancal, Inc.) white standard, using an Ocean Optics, Inc., PSIOOO diode- array portable spectroradiometer upgraded for near-ultraviolet light (range 300-800 nm), a bifurcated fiber-optic retlectance probe, and an Ocean Optics LS-1 tungsten-halogen lamp. To avoid specular reflectance, measurements with the reflectance probe were done at an angle of 45" against the feather surface, measuring an oval spot 3 mm wide. One measurement was made at the center of the rump for each specimen, as long as the reflectance curve averaging 5 scans was stable. Reflectance spectra were recorded between 350 and 700 nm with a resolution of 1 nm. The measure- ment range includes some near-ultraviolet light that is not per- ceived by humans but is by many birds. From retlectance curves, Thery then assigned each recorded spectrum a score for hue (domi- nant wavelength), chroma (purity or saturation of the color), and total brightness, computed following Endler (1990). Our procedure with underpart color was parallel to that for rump and scapular color. The specimens were ranked among six catego- ries. In category I (standard SDNHM 43972) the breast is practically white with a light buff tinge only along the sides. In category 2 (standard SDNHM 44592) a faint buff wash extends across the breast. Specimens in category 3 (standard SDNHM 47685) have a distinct buff breast band. In category 4 (standard SDNHM 48912) the breast is darker brownish buff and the throat is tinged brown. In category 5 (standard SDNHM 48938) the breast is still darker, medium brownish. Category 6 (standard SDNHM 43386) represents the Marsh Wrens with the darkest breasts; the entire underparts are brown with only a triangular patch in the center of the belly being whitish. In the darker-breasted specimens collected very shortly after molt, in September, the underparts are more rufous, whereas in those from the same locality collected just one month later, in October, the color has dulled to a drab medium brown. Because this change appears to result from some adventitious process, the specimens were ranked for underpart color on the basis of paleness or darkness alone, not hue. Again, we evaluated the breast color of each specimen with a spectroradiometer, placing the sensor over the darkest point along the midline of the breast. The results were converted into values for hue, brightness, and chroma in the same way as those for the rump. Data Analysis Speciromdiomeiry. A graphical assessment of the spectrora- diomelnc results for breast and rump color suggested that brightness was the variable with the greatest (and probably only) systematic variation. Scatterplots of spectroradiometric assessment versus vi- sual ranking for rumpAscapular bnghtness and for breast brightness (Figures 2 and 3) reveal in both cases a positive but only moderate correlation (Pearson correlations of 0.51 and 0.58, respectively). Some difference might be expected because the spectroradiometer measured a range of wavelengths broader than that to which the eye IS sensitive, but there may be other confounding factors as well. Because the visual assessments consistently gave better results, we did not use the speclroradiometnc results further in the analysis. The reason(s) why the spectroradiometer proved less satisfactory than the eye in these comparisons are unclear. Endler (1990) dis- cussed several reasons why the human eye's and brain's perception of colors is not proportional to the electromagnetic characteristics of the light reflected from an object and reaching the eye. He recom- mended the use of spectroradiometers to circumvent this problem. He did not address, however, other problems that may affect the applicability of spectroradiometers in taxonomic studies of birds, where often subtle rather than gross contrasts require quantification and testing. We suspect that such problems more than the nonlinear response of the human eye and brain accounted for the mediocre agreement between the spectroradiometric and visual assessments in our study. Possibly variation arising from the irregulariy multilay- ered structure of plumage overwhelmed the rather subtle vanation in color we were trying to record. Though we tried to ensure that the sensor recorded only the pigmented tips of the feathers, possibly it was influenced irregularly by some of the dark gray bases of the feathers, confusing the results. The sensor read a much smaller area of plumage than that embraced by the visual assessments; possibly because of differences of scale the two are not always comparable. The low retlectance of the rump and the dullness of both the rump and breast colors may exceed the equipment's sensitivity. Though spectroradiometry of plumage has been used occasionally in taxo- nomic studies of birds (e.g., John.son 1980, Atwood 1988), to our knowledge, the reliability and sensitivity of vanous spectroradio- meters and various techniques for using them have not been com- pared and tested. Since more precise and replicable quantifications Taxonomy of the Marsh Wren in Southern CaHfornia e o n. 95 — - -■^ 85 — • • J • 75 — • • 1 1 1 ""-^ • --. .•,^ 65 — • • 1 1 • II * • • • • t 55 — X t • ^^ • t ■•^ ---, t 1 45 — • • -* ^' : t t 35 — t ---- ^_ — ■ - 2.5 — "■ 1 1 1 1 1 1 1 1 1 1 OJ) 30 c -o a Urn o o < 0.005, but because of the stepwise procedure used in vanable selection, the nonnomial nature of several of the variables, and the nonrandom nature of the samples, it is questionable how much inference may be drawn from this fact. For similar reasons, we decline to present confidence regions or prediction regions based on an assumption of normality. For ease of interpretability in the pairwise compansons, we present graphs intermsof the pair of vanables that we felt were most useful in making a visual assessment of the group separation. TTiis is usually the pair of variables with the strongest univariate group mean separations (as measured by the F-to-enter test statistics at step 0 in the stepwise vanable-selection procedure). Occasionally, this pair did not produce the most obvious .separation, and in that case the pair of vanables that, considered together, was most informative was used (as measured by the f-to-entertest statistic at step 1 in the vanable-selection procedure). In the figure legends, these are described as "the pair of variables in which (the groups] differ mo,st." While these plots are more easily interpretable than the corresponding canonical-vanable plots, they do not separate the groups as well (compare Figures 4 and 1 1 ). In assessing which characters were most informative in the pairwise compansons, we again chose to present them in order of univariate mean group separations, as we felt these would be the easiest to use visually. Note that this means the first two variables listed may not be the/jairof most informative vanables but rather the two variables that when considered in isolation best differentiate the groups. When we say a pair of populations "differs significantly" in a variable, we mean the group means differ significantly at /? < 0.05. The variables actually used in the discriminant function are indicated by italics. These are the variables that, when the others are already in use, conlnbute additional information. Figure 5 Histogram of the canonical variable resulting from a stepwise discnminant analysis of the core sample of 10 specimens of Cislothorus patustris plesnislpuheniis and all Marsh Wren specimens from the San Joaquin delta. The discnminant function was based on the core specimens, and was used to classify the remaining specimens. Underpart color, rump/ scapular color, wing length, back pattern, and crown pattern, in decreasing order of importance, are the informative variables. C, core specimens of plesiuslpuherius: D. core specimens oi aesmarimis. A, inferred specimens of plesnislpulvenus. B, inferred specimens of aesiuanints. RESULTS Cluster Analysis Cluster analysis of the initial core sample of 65 specimens, covering all five key populations, identified three well-defined groups (Figure 6). The sample from coastal San Diego County was the most distinct; the 10 specimens of plesiuslpulvenus formed a second branch. Among the remainder of the specimens, the coastal Washing- Taxonomy of the Marsh Wren in Southern Cahfornia 27.11 plesius/pulverius paludicola aestuarinus Observations s. Calif. Figure 6. Cluster analysis of the core sample of 65 specimens of the Marsh Wren, covenngall five key populations on which further analyses were based Letters designate exceptional specimens not clustering with the rest of their population as labeled A, aestuarinus: C. coastal southern California; D. deserticolu: W, paludicola. The height of the horizontal bar separating each cluster is proportional to the distance between clusters, as measured by Ward's algonthni The first dichotomy separates all but one of the coastal southern California sample from the remaining specimens; the second dichotomy separates the Great Basin population [plesius/pulverius) from the remaining specimens. The population of southwestern Washington {paludicola) segregates only partially from the remaining specimens, from the Colorado Desert (deserlicola) and San Joaquin delta (aestuarinus). ton (paludicola) and Sacramento Delta (aestiianiuis) samples clus- tered together but did not segregate clearly from each other, some paludicola forming a subcluster, others falling with aeslitannus. Of the three Colorado Desert (deserlicola) two fell among the cluster of aestuarinus. one with the sample from San Diego. The sharp distinction of the San Diego sample from the remain- ing specimens suggested that it be evaluated as a separate group in the discnminant analyses; we then used the results of these analyses, both of the entire sample and of comparisons to each other popula- tion individually, to assess the level of and basis for this distinction. Discriminant Analysis: All Five Populations Simultaneously Core sample, deserlicola excluded. The core sample consisted of specimens that we are certain, on the combined basis of their locations and dates, to represent one of our five key or "parent" populations, corresponding to either named subspecies or the resident population of coastal southern California. Because the core sample included only three complete specimens of deserlicola. possibly insufficient to de- fine a discrete group, we ran the analysis both with and without these three specimens. Discnminant analysis of the core sample containing only the remaining four groups yielded a function capable of catego- nzing 58 of the 62 specimens into the four groups defined by their ongins, for an overall correct classification rate of 94%. The San Diego and Great Basin samples did not overlap with any other; the only misclassification was belween paludicola and aestuarinus. When the classification rates were corrected via the jackknife procedure, the results differed only in that one specimen from the San Diego sample and an additional specimen of paludicola were misclassified with aestuarinus (Table 1 ). From 86 to 100% of each group was classified as its origin suggested; for each of the four groups this figure exceeds the 75% traditionally regarded as the threshold for formal taxonomic recognition, suggesting each of the four may constitute a valid subspe- cies. A plot of the two most informative canonical variables associated with the discriminant function (Figure 7) best illustrates the segrega- tion among the populations, reducing to two dimensions most of the information from the six variables with which we describe variation in western Marsh Wrens. Core sample, deserticola included. A repetition of the discrimi- nant analysis including the three August/September specimens from Table 1. Jackknife-corrected discriminant-function classification of the core samples representing Cislothorus pcdustris plesiusi pulverius, C. p. paludicola. C. p. aestuarinus. and the population of coastal southern California. Number of specimens classified as plesius/ pulverius palu- dicola aestu- arinus Coastal S Calif Percent Correct C p. plesius/ pulverius C p paludicola C. p aestuarinus Coastal S Calif 10 0 0 0 0 0 100.0 2 0 900 12 0 8.5 7 1 17 94.4 Philip Unitl, Karen Messer, and Marc Thery 3 - UJ ry > (J •E o c o -2 - o o o o -5 0 1 5 canonical variable Figure 7. Scatterplot of the two most informative canonical variables generated by a stepwise discriminant analysis of the core sample of 62 speci- mens of the Marsh Wren, covenng four key populations (desertkola excluded). O. specimens from the Great Basin or inferred to have onginated there [plesiusl pulvenus). +. specimens from southwestern Washington ljkiludicola).x. speci- mens from the San Joaquin delta Uiesliuiiiinis), D, specimens from coastal southern California. the Colorado Desert did not greatly alter the results among the other four populations, but the attempt to define a group on the basis of these three specimens alone was unsuccessful. With the jackknife procedure, two of the three were classified by the discriminant function among other groups, while the desert "cluster" attracted one or two specimens from each of the other groups except plesiusl pulverius. A plot of the canonical variables implies a tendency of the Colorado Desert population, despite its marginal position geographi- cally, to be intermediate among the other four clusters in plumage. Clearly, a larger sample from the Colorado Desert was necessary for the taxonomic position of that population to be assessed, and this was the goal of the following step. Enlarged sample. We repeated the discriminant analysis, with an attempt to define five populations, on the core sample augmented as described above under Categorization of specimens for analysis (including the 12 October specimens from the San Joaquin delta, all inferred to represent aestiuinnns. the additional 1 3 and 1 6 specimens inferred, respectively, from companson of deserticola and plesiusl pulverius to represent those populations; Figure 4). This analysis of 104 specimens (Figure 8) suggested that all five populations could be considered subspecies. The poorest distinction was between aesluannus and deserlicola, but even after jackknife correction the discriminant function still classified 81% of each of these groups as expected (Table 2). Discriminant Analysis: Pairwise Comparisons of Five Populations Cistolhorus p. plesiuslpulverius vs. paludicola. This pair of popu- lations differs significantly in all vanables, in order of decreasing importance, wing length, rump/scapular color, underpart color, back pattern, nape-collar width, and crown pattern. The discnminant function revealed a wide separation belween paludicola and the core sample of plesiuslpulverius. With the enlarged sample of plesiusl pulverius. the separation was not complete; three specimens fell within the cluster of paludicola. reducing the rate of correct classifi- cation in the total sample to 93% after correction for over- classification via a jackknife procedure. Figure 9, a scatterplot of plesiuslpulverius versus paludicola in wing length versus rump/ scapular color, reveals these three misclassificd specimens. Possibly, some of the misclassificd plesiuslpulverius in the enlarged sample, from the Imperial Valley, were misidentified deserlicola (see above under Categonzation of specimens). But even if not, the two popula- tions are differentiated well enough to be continued to be recognized as subspecies. Cislothoriis p. plesiuslpulverius vs. aestuarinus. This compari- son yielded one of the strongest distinctions; the two groups differ significantly in five of six vanables (all except nape-collar width), and the discriminant analysis, after jackknife correction, classified 100% of the specimens as expected. Rump/scapular and underpart color are the two variables most responsible for this separation (Figure 10). Cistolhorus p. plesiuslpulverius vs. deserticola. These two groups differ significantly in four variables I rump/scapular color, under- part color, wing length, and crown pattern), but in the discriminant analysis the second two did not add any discriminating power be- yond the first two. Therefore, in Figure 1 1 we present a scatterplot of the two groups in just these two variables; this allowed us to include the two specimens in the core sample of deserticola that were molting their primaries. In this plot, the groups appear well sepa- rated, hut, as discussed above under Categorization of specimens and shown in Figure 4, several fall and winter specimens from the Impenal Valley bridge the gap. Study of a larger sample of both subspecies may reveal some overlap not evident with our core samples. Nevertheless, the strong separation obvious in Figure 11, and the 1 00% separation achieved by the discnminant function, even after jackknife correction, implies that any ovedap is not extensive enough to invalidate the distinction between deserticola and plesiusl pulverius. Cistotlwrus p. plesiuslpulverius vs. coastal southern California Table 2. Jackknife-corrected discriminant-function classification of the en- larged samples representing Cistolhorus paluslris plesiuslpulverius, C. p. paludicola, C. p. aestuarinus. C. p. deserlicola. and the population of coastal southern California. Number of spe cimens c assified as Pl esiusi palu- aestii- deser- Coastal Percent pii veriiis dicola anntis licola S. Calif Correct C. p. plesiusl puheriii.s 22 1 0 1 0 91.7 C. p. piiludicola 0 17 2 1 0 85.0 C. p. aestuarinus 0 2 21 3 0 80.8 C. p. deserlicola 0 0 2 1,1 1 81.3 Coastal S. Calif. 0 0 0 1 17 94,4 Taxonomy of Ihc Marsh Wren in Soulhem California 4 3 (N Id 0 o • ^1^ c -1 o c a o -2 -3 ▲ < A < i^ A A A < o • •P • • A \l O o o A A A < < <1 ( o o8 • • • f A 1e ■ «p o o ^^* ^ t ■ ■ ■ ■ -5 0 5 canonical variable 1 Figure 8- Scatterplol of the two inost informative canonical vanables generated by a stepwise discnminant analysis of the enlarged sample of 104 specimens of the Marsh Wren, consisting of the core sample of 65 specimens augmented with 14 inferred specimens of plesius/pulverius. 12 inferred specimens of aestiuirinus. and 13 inferred specimens of deseriicoUi. •, core-sample specimens of plesius/pulverius: O. inferred specimens of plesiusi puhenus: H. specimens from southwestern Washington [puludicola): *, core-sample (September) specimens from the San Joaquin delta (aesluarinus): +. inferred (October) specimens of aesluctnitus from the San Joaquin delta; M, core-sample specimens from the Colorado Desert (desenicohi). 0. inferred specimens of deserucola from the Colorado Desert; A. core-sample specimens from coastal southern California. population. Rump/scapular color, wing length, underparl color, and crown pattern all contributed significantly toward the strong separa- tion of these two groups. Figure 1 2 is a scatterplot based on the two strongest vaiiables, rump/scapular color and wing length. Only two specimens were misclassified by the discriminant function. Both are inferred specimens oi plesius/pulverius from the Imperial Valley and not used in the generation of the discnminant function. Cistolhorus p. paludicola vs. aestuannus. Five variables contrib- uted significantly toward the differentiation of the Washington and San Joaquin Delta samples; in order of decreasing importance, underparl color, nape-collar width, crown pattern, rump/scapular color, and back pattern. Figure 13, a scatterplot depicting the first two of these variables, reveals some overlap. Even with all variables combined in a discnminant function, however, a 100% separation could not be achieved. Marginal overlap left three specimens of paludicola and one of aestuarinus on either side of the line of best separation decided by the discnminant function. Jackknife correc- tion of the function suggested that 88% of the combined samples could be assigned correctly. All of the inferred (October) specimens of aestuarinus were correctly classified. Though the separation is not total, it appears adequate to support continued recognition of aestuarinus as distinct from paludicola. Cistothorus p. paludicola vs. deserticola. Rump/scapular color. nape-collar width, and back pattern contributed toward a good separation between this pair of samples (Figure 14). Only one speci- men was misclassified by the discriminant function, an inferred specimen of deserticola that by virtue of its rather narrow white back streaks (rated 2) and rather pale rump (rated 5), and despite its narrow nape collar (3 mm, outside the range of paludicola), fell in the middle of the cluster of paludicola as defined by the canonical vanable generated by the discriminant function. Following the jack- knife correction, the discriminant function predicted a 97% separa- tion of these two populations. Cistolhorus p. paludicola vs. coastal southern California popula- tion. This pair of samples segregated 1 00%, differing significantly in all variables, in order of decreasing difference, rump/scapular color, crown pattern, wing length, nape-collar wuith, hack pattern, and underparl color. The plot of rump/scapular color versus back pattern (Figure 15) shows the clearest separation in two variables. The canonical variable generated by the discriminant function suggested 10 Philip Unitt, Karen Messer, and Marc Thery 2 3 4 5 rump / scapular color Figure 9. Scatterplot for Cistolhonis pahislns plesiuslpulvenus and C p. pahulicokt of the two vanahles in which they differ most, vving length and rump/scapular color So that the sexes could be considered together in one statistical process, the wing chords of females were multiplied by 1.06, the factor by which the average male Marsh Wren exceeds the average female •, core-sample specimens of plesius/putverius; O, inferred specimens of plesitis/pulverius.M. specimens from southwestern Washington (piitudicdki). In this and subsequent figures in which one or both of the axes represents a variable ranked in discrete categones, some "jitter" has been added to avoid overstnkes and thereby show all points plotted. not only complete separation but a gap between the two populations (Figure 16). It is evident that the name pahuiicola does not apply to the resident Marsh Wrens of coastal southern California. Cistothorus p. aesluarimis vs. deserticola. Among the 10 pairs of "parent" populations, this comparison yielded the weakest separa- tion. The stepwise variable-selection procedure used only a single variable, underpart color, in the discriminant function. Back pattern 2 3 underparts Figure 1 1 . Scatterplot for Cisunlumis paluslris plesiuslpulvenus and C p. deseriicoki of the two vanables in which they differ most, nimp/scapular color and underpart color 9, core-sample specimens of plesiuslpulvenus, O, in- ferred specimens oi plesiuslpulvenus; ■ , core (August/September) specimens of the breeding population of the Colorado Desert (.desenicola): D, additional October-February specimens from the Colorado Desert inferred as desenicola (see Figure 4). is the next most informative variable; its distribution among so few categories reduces Us contribution to the generation of the discrimi- nant function. When the function was modified to include back pattern, however, it placed 3 of the 30 specimens of aestuannus with desenicola, 1 (not from the core sample of 5 specimens) of the 18 specimens of desenicola with aestiiarinus (Figure 17). Thus a dis- tinction can be drawn so that only 4 of 48 specimens in the samples from at or near the type localities overiap, suggesting a valid distinc- tion between these two subspecies. Application of the discriminant functions to specimens from elsewhere in California, however, mud- o ■o 5 •S 4 - n. y 3 "d. E 2 0 12 3 4 5 underparts Figure 10. Scatterplot for Cislolhorus palustris plesiuslpulvenus and C p. aestuarinus of the two vanables in which they differ most, underpart color and rump/scapular color 9. core-sample specimens of plesiuslpulvenus; O. inferred specimens of plesiuslpulvenus; H , core-sample (September) speci- mens from the San Joaquin Delta (aesiuarinus); D, additional (October) specimens from the San Joaquin Delta (inferred aesiuarinus) 10 9 8 7 6 5 4 3 2 H 1 0 CO O ^ (P *) o© • o • • • 45 50 55 adjusted wing chord Figure 12. Scatterplot for Cislolhorus paluslris plesiuslpulverius and coastal southern California Marsh Wrens of the two vanables in which they differ most, rump/scapular color and wing length. So that the sexes could be considered together in one statistical process, the wing chords of females were multiplied by 1 ()6, the factor by which the average male Marsh Wren exceeds the average female •, core-sample specimens of plesiuslpulvenus; O. inferred specimens of plesiuslpulverius; ■ , specimens of the breeding population of coastal southern California Taxonomy of ihe Marsh Wren m Southern Cahfornia 11 5 - c« 4 — a, (U T3 3 - nape-collar width Figure 13. Scatterplot for Cisiorhonis paluslns pahidicoUi and C. p. aesluarinus of the two vanables in which they differ most, underpart color and nape-collar width. A. specimens from southwestern Washington (paludicola), •, core-sample (September) speciinens from the San Joaqum Delta (aestuaimus). O. additional (October) specimens from the San Joaquin Delta (inferred aesuuinnus). died this distinction substantially, so it does not seem useful on a broader scale (see below). Cislollumis p. aestitarimis vs. coastal southern California popula- tion. The discriminant function analysis, with jackknife correction, separated these two groups completely, with a substantial gap between them. Rump/scapular color, wing length, undetpart color, crown pal- tern, and back pattern, in decreasing order, all differed significantly. Figure 18, a plot of rump/scapular color versus underpart color, shows this separation in just two vanables. Evidently, the name aestitannus does not apply to the Marsh Wrens of coastal southern California. Cistothorus p. deserticola vs. coastal southern California popu- lation. The discriminant analysis, from the direct result and after 2 3 back pattern Figure \^- Scatterplot for Cislalhorus pulustns paludicola and coastal southern California Marsh Wrens of the pair of vanables in which they differ most, rump/scapular color and back pattern A, specimens fi-om southwest- em Washington (paludicola): •, specimens of the breeding population of coastal southern California. cross-validation via jackknifing. classified 38 and 37, respectively. of 39 specimens as expected on the basis of their origins. The discriminant-function analysis identified four variables as infor- mative in making this distinction, in order of decreasing impor- tance, wing length, rump/scapular color, crown pattern, and un- derpart color. Between this pair of populations, back pattern and nape-collar width do not differ significantly. Wing length alone separates the samples totally (with adjustment for sex, the coastal sample measures 44.0-50.4 mm, the desert sample 50.7-55.9 mm), though the approach is so close some overlap should be expected in large samples. Though the two groups differ substantially in rump/ scapular color (only 2 of 1 8 desert specimens rating darker than 7; only 2 of 1 8 coastal specimens rating paler than 8, and one of these, collected in February, was likely faded in comparison to the rest of the sample, collected in August and September), the computer- o o 3 o C/1 11 -f 10 9 8 7 6 5 4 H 3 2 back pattern Figure 14. Scatterplot for Cistothorus palusiris paludicola and C p. deserticola of Ihe pair of variables in which they differ most, nimp/scapular color and back pattern A. specimens from southwestern Washington [paludicola). •. core-sample (August/September) specimens of the breeding population of the Colorado Desert (deserticola), O. additional Oclober-Febm- ary specimens from the Colorado Desert inferred as deserticola (see Figure 4) 20 10 r3 0 - -13 -12 -11 -10 -9 -8 -7 -6 canonical variable Figure 16 Histogram of Ihe canonical vanable resulting from a stepwise discnminant analysis of Cistothorus paluslns paludicola and coastal south- ern California Marsh Wrens Rump/scapular color, nape-collar width, and back pattern are the vanables contnbuting to the separation. Shaded bars, C p. paludicola; white bars, coastal southern California Marsh Wrens. 12 Philip Unitt, Karen Messer, and Marc Thery 5 - <% t 4 — \\ ^ • 3 - ^ °& 2 - 1 (TO 1 1 back pattern Figure 17. Scatterplot for Cixtolhorus paluslris aesluarinics and C. p. deserlicoki of the only two vanables in which they differ significantly, under- part color and back pattern A. core (September) specimens from the San Joaquin Delta (aesluarinus). A . additional (October) specimens from the San Joaquin Delta (inferred aesniannus); 9. core (August/September) specimens of the breeding population of the Colorado Desert (deserticdhi), O, additional October-February specimens from the Colorado Desert inferred as deserticola (see Figure 4) generated discriminant function did not use this variable, the infor- mation it provides being redundant with thai for other variables. The difference in the crown is the coastal sample's averaging more extensively black. A slight difference in the underparts arose from 25% of the desert sainple's being rated 2, paler than in any of the coastal specimens. Figure 19 is a scatterplot of these two groups showing the separation on the basis of wing length and rump/ scapular color Two specimens of these groups were misclassified by the dis- cnminant function. One is SDNHM 42931, the single February (comparatively worn and faded) coastal specimen, the other SDNHM 46003, an Impenal Valley specimen at the dark extreme for the species in rump/scapular color (rated 10), a color matched otherwise 10 o 3 Q. a o 7 - 6 - 5 - underparts Figure 18. Scatterplot for Cistothorus paluslris aestuarimis and coastal southern California Marsh Wrens of the pair of variables in which they differ most, rump/scapular color and underpans. •, core (September) specimens from the San Joaquin Delta Uiesluanmis), O, additional (October) specimens from the San Joaquin Delta (inferred aesluiinmis), ■, specimens of the breeding population of coastal southern California 10 9 - 6 - 5 - 47 49 51 53 adjusted wing chord Figure 19 Scatterplot for Cisunhorus paluslris deserlicoki and coastal southern California Marsh Wrens of the two vanables in which they differ most, wing length and rump/scapular color So that the sexes could be considered together in one statistical process, the wing chords of females were multiplied by 1 06, the factor by which the average male Marsh Wren exceeds the average female. ■. core (August/September) specimens of the breeding population of the Colorado Desert (deseriicola), D, additional October-February specimens from the Colorado Desert inferred as desenicola (see Figure 4); •, specimens of the breeding population of coastal southern California. only along the southern California coast. In wing length and crown pattern, however, this specimen is like deserticola and unlike the coastal sample. The jackknife-corrected discnminant analysis suggested that only 5% of a sample of these two populations should overlap. Therefore, the name desenicola does not apply well to the southern coastal population. ALLOCATION OF SPECIMENS FROM OTHER SITES We applied the discriminant functions generated by both the core and enlarged samples to specimens from elsewhere in the Marsh Wren's breeding range. Coos County, Oregon Of two specimens collected 7. 1 miles north of Coos Bay on 22 October 1982 (SDNHM 42077 and 44078), one was placed by the discriminant functions (from both the core and enlarged samples) with deserticola, the other with aesluarinus. That neither was placed with paludicola suggests that the southern limit of paludicola lies somewhere along the coast of central Oregon and does not reach California. Humboldt County, California A single immature male (SDNHM 3203), collected at Clam Beach on 18 October 1967, is older than the specimens on which the discriminant functions were based but we assessed it anyway, as it was still in heavy molt so presumably represents the local population. With a rump/scapular rating of 7 and a crown rating of 4, it is at the pale extreme for the southern California coastal population, but its short wings (apparently fully grown in at 48.8 mm) are typical of that group, to which the discriminant functions assigned it. With the exception of one specimen from the Imperial Valley, it is the only specimen from outside coastal southern California placed with that group. Taxonomy of the Marsh Wren in Southern Cahfornia 13 Coastal Central California Another somewhat aged specimen (SDNHM 35175), collected 18 September 1958 just northeast of Martinez in Contra Costa County, has paler underparts (rated 3) than the sample from just 10 miles to the northeast across Suisun Bay at Joice and Grizzly islands (all rated 4 or 5). Therefore, it matches deserticola better than aeslitannus, and that is where the discriminant functions placed it. An old, foxed specimen (SDNHM 24642), collected 4 December 1938 at Santa Cruz, though not evaluated by the discrimrnani func- tions, appears closest to deserticola as well. The breast is only lightly tinged buff, while the crown is too brown and the wings loo long (50.7/51.7 mm, female) for the southern coastal population. A specimen from the mouth of Los Osos Creek on Morro Bay, San Luis Obispo County, collected 7 September 1986 (SDNHM 44461), was placed with aestuarinus by the discriminant function based on the core specimens alone, with deserticola by the function based on the enlarged sample. Of two specimens (SBMNH) from Dune Lakes, southwestern San Luis Obispo County (28 September 1962, 12 September 1973). the former was placed by the discnmi- nant functions with deserticola. the latter with aestuarinus. Of two specimens from the Santa Ynez River mouth. Santa Barbara County (UCSB). one (4 January 1992) is clearly a rmgxxAplesiuslpulverius. while the other (10 January 1992) was placed by the discnminanl functions with aestuarinus. Though the base for drawing a conclu- sion is rather meager, evidently the Marsh Wrens of coastal central California bridge the difference between aestuarinus and deserticola. Oases of Mojave Desert We examined specimens from two sites in the Mojave Desert. Harper Dry Lake, San Bernardino County, and Piute Ponds, Los Angeles County. The specimens from the Piute Ponds, within Edwards Air Force Base in the Antelope Valley, are all in the Los Angeles County Museum. Five were collected on 27 October 1989. Of these, two are migrant plesiuslpulverius. while the other three fit with deserticola. Six were collected in spring, from 26 April to 8 May. Three are juveniles; two are badly worn adults. One adult, a male with enlarged testes taken 26 Apnl 1989. is still in a condition good enough to be assessed. It agrees well with deserticola in all variables. From Harper Dry Lake, we assessed 17 specimens, all collected in fall and early winter, the earliest fall specimens being taken on 28 September All were therefore taken at a time when migrants should be expected, and. not surprisingly, of the 17. 6 were identified by the discriminant functions as plesiuslpulverius. Two additional speci- mens (SBCM), identified by the functions as paludicola, were prob- ably migrants from the Great Basin as well. One had a rump/scapular score of 2, typical for plesiuslpulverius but not (or paludicola. With nape-collar widths of 5-6 mm, probably their necks were elongated in preparation in compan.son to the specimens used in the defining samples, as they were made by different preparators. The other 9 Harper Dry Lake specimens are too dark on the rump, scapulars, and underparts for the migratory subspecies, so we infer these represent the resident population. Of the 9, 5 conform with deserticola. while 1 (SBCM 53683), with a nape-collar width of 5 mm and a back score of 2, was placed by the functions with paludicola. Again, this specimen may have had its neck overly stretched; in other variables it agrees with deserticola. Three specimens, by virtue of their darker underparts (especially striking in SDNHM 48952) or narrower back streaks, better fit aestuarinus. Though it is possible that these appar- ent aestuarinus dispersed southeast from the Central Valley, more likely they represent normal variation in the resident population. The dark extreme of underparts crops up in one specimen of deserticola from the Imperial Valley, and the narrower white back streaks (rated 2) in two specimens. As noted above, the distinclion between aestuarinus and deserticola is comparatively weak, and despite the wide separation of the main ranges, the intervening oases may provide an opportunity for gene flow. Even in these nonmigratory subspecies, substantial dispersal ability likely favors survival of birds dependent on widely scattered tiny patches of suitable habitat. Ventura County, California Unfortunately, we located only a single recent specimen from Ventura County, a male taken at the sewage ponds in the Point Mugu military reservation on 13 December 1986 (SBMNH 5090). This specimen falls within the range of the southern coastal population in its fairiy dark rump and scapulars (rated 7) and fairly blackish crown (rated 4) but disagrees in its long wings (53.4). The discriminant functions placed it with deserticola. SUBSPECIES DEFINITIONS These comparisons suggest that plesiuslpulverius, paludicola, aestuarinus, and the population of coastal southern California are all differentiated at a level appropriate for designation as subspecies. Cistothorus p. deserticola presents a more awkward problem. Cistothorus p. plesiuslpulverius Our analyses reaffirm the distinctiveness of the more or less migratory plateau population from the lowland populations nearer the Pacific Coast, a difference universally recognized since 1897. This group is identified by its comparatively long wing (a difference expected between migratory and sedentary populations), entirely brown to moderately black crown (Figure 20), narrow brown nape collar, broadly white-streaked back, comparatively pale tawny rump and scapulars, and pure white or only slightly buff-tinged breast (Figure 21). In combination these features sufficed to distinguish 92% of our sample. Assessing the distinction between plesiits and pulverius was not a goal of this study, and such an assessment was not possible from the sample used, which lacked specimens from the breeding range of plesiids. The few specimens from eastern California were among the largest and palest of the sample, however, suggesting the Marsh Wrens breeding along the east side of the Sierra Nevada and Cascade Range (pulverius) represent the extreme development of this group's charac- ters and may be distinguishable from the majority of winter visitors in southern California, in which these characters are on average less extreme and may be migrants from farther east in the Great Basin/ inlermountain region (plesius). The greater abundance of apparent plesiiis in winter in southern California, despite its breeding range's being more remote, may be due to a difference between pulverius and plesius in the winter climate of the breeding range. From the Pacific Ocean to the Rocky Mountains, winter temperatures tend to decrease from west to east, and wintering by Marsh Wrens in the ranges of pulverius and plesius has been reported as irregular and dependent on iheseventy of the winter (e.g.. Root 1988; Gilliganet al. 1994). Cistothorus p. paludicola The sample from southwestern Washington stood apart from the others largely on the bases of its entirely brown to only slightly black- margined crown, broad brown nape collar, and nartowly white streaks on a reduced weakly black-tinged back patch. TTie rump/scapular ranking of pahuhcola overlapped extensively with those of aestuarinus and deserticola on our light-to-dark scale, but the hue of these parts of the p]umage paludicola tends more toward an earth brown, and away from rufous, than in the other lowland subspecies, a subtle variation not captured in our analyses but evident in Figure 20. In wing length and underpart pattern paludicola occupies a position intermediate with and overiapping several other populations. The weakest separation of paludicola was from aestuarinus, but the darker underparts, blacker 14 Philip Unite, Karen Messer. and Marc Thery clarkae pahidicola deserticola plesius aestuarinus pulverius Figure 20 Upperparts of six populations of the Marsh Wren CisUHlwnis paluslns clarkae. based on SDNHM 48915, holotype from Batiquitos Lagoon, Carlsbad, San Diego County, Cahfomia, 23 August 1994 C /> desenicola (best included under aesuiariims). based on SDNHM 44278, a topolype from 2 miles north-northwest of Seeley, Impenal County, California, 29 September 198-'>. C p. aesnianniis. based on SDNHM 445.^2, virtual topotype from Joice Island, Solano County, California, 9 September 1986 C /> paludicola. based on UW 40S65, a topotype from the North River mouth, Willapa Bay, Pacific County, Washington, 18 November 1985 C /> plesius. based on SDNHM 43971, from Picacho Reservoir, Pinal County. Anzona, 24 October 1985. C. p. pulverius. based on SDNHM 4.3469, from Owens Lake, 5 miles northeast of Olancha, Inyo County, California, 20 September 1984. Taxonomy of the Marsh Wren m Soulhcrn Cahlorni; 15 clarkae paliidicola deserticola plesius aestimrmus pulverius Figure 21. Underparts of six populations of the Marsh Wren, based on the same specimens as in Figure 20 16 Philip Unitt, Karen Messer, and Marc Thery crown, narrower nape collar of the latter still served to allow 91% of that pair of populations to be distinguished. Cistothonis p. aestuarimis The Sacramento Delta sample differed from the other groups primarily on the basis of its dark underparts; the specimens ranked darkest (5) on the underparts were from this area almost exclusively. In wing length, rump/scapular color, crown pattern, nape-collar width, and back pattern the delta sample was in an intermediate position, as might be expected from its central position geographi- cally. The extreme developments of some of the peripheral popula- tions, however, plesius/pulveriiis in large size and plumage paleness, paliidicola in its broad nape collar and brown crown, and the popula- tion of coastal southern California in its small size and dark upper- parts, left aesluarinus adequately isolated from these. The distinc- tion from deserlicola. on the basis of darker underparts and narrower white back streaks alone, was the weakest but could be defined so that only 4 of 43 specimens from near the type localities in the combined enlarged samples overiapped with the other. Cislothonis p. dcsi'iticiila Despite its peripheral position geographically, the population of the Colorado Desert is the least distinctive morphologically. In all charac- ters, it lies in a position intermediate among the other subspecies. It differs from plesius/pulveriiis in its tendency toward a darker rump and scapulars, buffier underparts, shorter wings, and blacker crown. It differs from paludicola in its tendency toward darker rump and scapu- lars, a narrower nape collar, ;ind bolder white back streaks. It differs from aestnarmus in its tendency toward paler underparts and narrower white back streaks. Its differs from the population of coastal southern California in its longer wings, tendency toward paler rump and scapu- lars and browner crown, and average slightly paler breast. If only the type localities oi aesluarinus and deserlicola had to be considered, the two subspecies could be recognized fairly easily. But the samples from coastal central California and Harper Dry Lake show the whole range of these phenotypes. A distinction between the two implies a biogeographically oddly shaped zone of intergradation at least as large as the core ranges of the subspecies. Therefore, recognition of deserlicola as distinct from aesluarinus doei not seem practical, a result implied also by the group's failure to segregate in the cluster analysis. The wide range (even to south-coastal Oregon) of Marsh Wrens matching the characters of deserlicola suggests that the deserlicola phenotype could represent the primitive appearance of California's lowland Marsh Wrens and that the dark underparts of aesluarinus are an innovation that arose in or near the San Joaquin Delta and has spread to some degree over most of California but has barely touched the southeastern corner of the state. Conversely, the aesluarinus type may have originally been widespread, and, especially before the damming of the Colorado River and irrigation of the Imperial Valley, deserlicola was confined largely to the Colorado Delta. Changing water-management practices have probably increased the range and population of deserlicola enormously in historic time (Rosenberg et al. 1991), possibly enabling it to invade northwest and mix with aesluarinus. Southern California Coastal Population The sample from San Diego County was well isolated from the others on the basis of having the shortest wing, most extensively black crown, and darkest rump and scapulars. In nape-collar width, back pattern, and underpart pattern it did not differ greatly from several other samples, overlapping in those features substantially. But the sample stands at the extreme for the species in three of the six characters quantified in this study. Cluster and discriminant analyses consistently identified it as an independent group. In all discnminant analyses, both pooled and pairwise, 91% or more of this sample was classified as expected on the basis of origin, while no more than 6% of any other sample was classified with the San Diego group, even when the discriminant functions were cross- validated via jackknif- ing. Because the level of differentiation of this sample well exceeds the 75% threshold, and equals or exceeds that of other subspecies recognized in the Marsh Wren, we propose that it be known as Cistothorus palustris clarkae subsp. nov. Holotype. San Diego Natural History Museum number 48915, collected by Philip Unitt (original number 1357) on 23 August 1994 at the east end of Batiquitos Lagoon, city of Carisbad, San Diego County, Califomia (35" 06' N, 117" 16' W). Adult female (skull completely pneumatized; ovary granular, 4 x 1.5 mm; ova minute). Weight 8.1 grams; slight fat. Length in flesh 123 mm, wingspread 1 56 mm, wing chords 47. 1 mm. Prebasic molt completed except for a few feathers on chin and face. Diagnosis. Differs from other western populations of the Marsh Wren in small size (see Table 3), more extensively black crown (at least 50% black; usually with only a small brown patch in the center of the forehead), and darker rufous scapulars, rump, upper tail coverts, and central rectrices, in fresh plumage all of these close to Raw Umber or Prout's Brown of Smithe (1975). The new subspecies differs from deserlicola, as suggested by jackknife-adjusted discnminant analysis, at a 94% level (at least) by these three characters, plus a slight tendency to a darker breast. A single exceptionally dark specimen of deserlicola and a probably laded February specimen of clarkae were the only two confounding specimens. In the sample examined wing length alone yielded 100% separation. From aesluarinus, clarkae differs at a nearly 100% level by its usually darker rump and scapulars, shorter wing, paler underparts and flanks (breast pale buff), and blacker crown. In combination these characters distinguish 100% of specimens from the type local- ity. From the broader range of aesluarinus, as we redefine it, includ- ing all of coastal California south to Ventura County and the Mojave and Colorado deserts (deserlicola), only 2 of 60 specimens (3%) were placed by the discriminant functions with clarkae. From paludicola, with which it has traditionally been linked, clarkae differs al a 100%> level by its bolder white back streaks on a blacker background, and narrower brown nape collar (2-7 mm in clarkae. >5 in only 2 of 23 specimens; 4—9 m paludicola, <5 in only 3 of 20 specimens) in addition to the three other characters. From the migratory plateau subspecies pulverius and plesius, which invade its range in fall and winter, clarkae differs even more in the three main characters than it does from the other lowland popula- tions, plus it has a more intensively buff breast. From the Marsh Wrens of northeastern North Amenca Uamgi, iliacus, dissaeplus, and nominate paluslris), clarkae differs in its barred upper tail coverts as well as its darker rump and scapulars and buff breast band. From the Marsh Wrens of the coastal southeastern United States clarkae differs as follows; from lliryophilus by its largely blackish crown, from marianae and waynei by its more rufous rump and scapulars and lack of dusky speckling or barring on the flanks or breast, and from griseus by its far more rufous rump and scapulars and more extensive black and white patch on the back. From lolucensis of central Mexico clarkae differs by its less extensive black and white back patch and paler, less rusty underparts. Dislnbulion. Coastal lowland of southern California, from the Tijuana River immediately north of the Mexican border north to Los Angeles. Of 25 old October-February Marsh Wrens from Los Ange- les and Orange counties in the Los Angeles County Museum of Natural History, 22 appear to be migrants of plesius/pulverius, while Taxonomy of the Marsh Wren in Southern California 17 Table 3. Wing chords of various subspecies of the Marsh Wren. Standard n Mean Range Deviation Males C. p. plesiusi puherius II .S4.9 51.0-57,1 1.93 C. p. paludicola 16 51-3 49.5-52.4 0.70 C p. uestuanmis 15 51.4 49.5-53.3 0.97 C. p. deserticola 5 52.4 51.7-53.8 0.81 C. p. clarkae 12 48.2 44.0-50.4 1.84 Females C. p. plesiusi pulverius 13 50.3 48.2-52.2 1.16 C. p. paludicola 4 46.4 45.3-48.2 1.09 C. p. aestiiaritnts 12 48.0 46.5-51.3 1.36 C. p. deserticola II 49.0 47.8-52.7 1.39 C. p. clarkae 12 45.6 43.3^7.1 1.09 3 agree with clarkae in their extensively black crowns (rated 5 or 6), dark (now foxed and unscorabie) rump and scapulars, and short wings (46.2 in the one female; 44.7 and 48.2 in the two males). These are LACM 12289, from the San Gabriel River on 23 December 1895, LACM 2376, from Nigger Slough on 12 February 1918, and LACM 17629, from Play a del Rey on 20 December 1931. These specimens demonstrate that the characters of clarkae are long estab- lished in this population, not an artifact or innovation of the recent sample from San Diego County. The breeding Marsh Wrens of western Riverside County prob- ably belong with clarkae. though this needs confirmation with recent specimens. The only specimen we have seen from this area, SBCM 3912, is a migrant plesius collected 12 March 1967. For further details and historical changes, see below. One specimen implies short-distance dispersal of clarkae across unsuitable habitat. This is LACM 1 9635, a male collected by George Willett on 13 November 1939 on San Clemente Island, where the Marsh Wren is only a rare nonbreeding visitor (Jorgensen and Ferguson 1984). The rump is as dark as in the three old specimens of clarkae from Los Angeles County. The crown seems to have lost a few feathers so is difficult to judge but looks about half black. At only 48.0 mm. the wing measurement is typical of clarkae. Etymology. We name this Marsh Wren in honor of Mary Hollis Clark, in appreciation of her 33 years of support and service, through good times and bad, to the San Diego Natural History Museum and its scientific and educational mission. With the help of Mrs. Clark and her family, many San Diegans have increased their appreciation of the uniqueness of their natural environment. With its narrowly restricted range, Cistolhonis palustns clarkae reveals itself as yet another element of that uniqueness. Discussion. Our results confirm Rea's (1986) conjecture that the coastal southern California Marsh Wrens constitute a subspecies, though not quite on the basis that he suggested. As noted above, the pigment coloring the rump and scapulars of Marsh Wrens is subject to foxing, increasing its redness. Evidently, the greater the concentration of this pigment, the more grossly the specimen foxes. In the pale extreme of the species {pulverius) the difference is slight; in the dark extreme (clarkae) the difference is great, as seen in companson of our 1994 specimens with the three collected in Los Angeles County from 1895 to 1 931. The rump and scapulars of old specimens, in which the natural darkness of clarkae has been foxed into rustiness. look "brighter and richer" (Rea 1986) than in the other subspecies. Swarth (1917) noticed that the Marsh Wrens of coastal southern California were exceptionally small but did not quantify the differ- ence. Neither he nor subsequent revisers applied this variable to the population. The difference in crown pattern has not been reported previously. Revelation of an undescribed subspecies of bird in a region as heavily populated as coastal southern California, which has been studied by thousands of biologists, may seem incongruous. In the ca.se of Cistothorus palustris clarkae, however, several factors contributed toward concealing it for so long. In existing collections, the great majority of specimens of the Marsh Wren from coastal southern California were taken in winter, and consist largely of migrants from the plateau region. Since there is .substiuilial variation among these migrants, probably interpretable as two subspecies, the additional variation arising from the few specimens of clarkae was less notice- able. Some individuals of paludicola and aesluarinus were thought to make long-distance migrations, reaching coastal southern California, for which we have seen no evidence. The few breeding-season speci- mens (e.g., SBCM 36797, Nigger Slough, Los Angeles County, 13 May 19l7)are in poor condition, often with wom crowns, faded rumps and scapulars, and broken primary tips; they do not preserve well the defining features of clarkae evident in fresh plumage. The very urban- ization of the range of clarkae discourages collectors from working in it; collecting birds at most sites in coastal southern Califomia where Marsh Wrens breed now requires special authorizations beyond the standard permit from the Califomia Department of Fish and Game. DISTRIBUTIONS OF THE SUBSPECIES OF THE MARSH WREN IN SOUTHERN CALIFORNIA Cistothorus p. plesiuslpulverius Marsh Wrens from the Great Basin/intermountain plateau region winter throughout southern Califomia, invading many areas where the species does not breed. Though Garrett and Dunn ( 1981 ) reported the Marsh Wren as absent from the "colder northern deserts" in winter, T Heindel (pers. comm.) finds it more common in the Owens Valley m winter than in summer West of the Sierra Nevada, Gnnnell and Miller (1944) reported plesius north only to Tomales Point and Petaluma. Mann and Sonoma counties. Gabrielson and Jewett (1940) did not report it from west of the Cascade Range in Oregon, but in Jewett's collection (SDNHM) are three specimens, from Taft. Netarts. and Portland, that conform with the intenor population well in both color and size. The specimen from Portland (6 October 1928, SDNHM 24646, wings 55.5/56.0 mm) apparently represents the northwestem- most documented point of these migrants' dispersal. Understanding of the schedule of arrival and departure of plesiusi pulverius in southern Califomia needs further refinement. We have not seen any specimen eariier than one taken for this study in the Tijuana River valley, San Diego County, on 23 September 1994 (SDNHM 48983). But the migration unquestionably begins some- what earlier than this; M. Heindel (pers. comm.) finds Marsh Wrens migrating commonly through the desert oases of eastern Kem County by 10-20 September; M. A. Patten (pers. comm.) has earliest dates for migrants in the California deserts of II and 17 September; Lehman (1994) reported migrants as fairly common in Santa Bar- bara County by the beginning of September Marsh Wrens have been reported away from breeding sites as early as 29 August in eastern Kem County (M. Heindel pers. comm), 14 August in San Diego County (Unitt 1984), and 23 July in Santa Barbara County (Lehman 1994). We suspect the few early records represent short-distance dispersal from breeding sites nearby and that migrants from the plateau region do not reach the coast of southern California until around the middle of September. But the exact schedule and possible annual variations of this arrival remain uncertain, requiring testing via further collecting. Spring departure takes place largely in April. We have not seen specimens of plesiuslpulverius from the winter range later than 13 Apnl (1923, salt marsh at Santa Catanna Landing, 29" 30' N, Baja 18 Philip Unilt, Karen Messer. and Marc Thery •O Santa Barbara Ventura O \ Los Angeles San Bernardino