Wintering Artemisiospiza sparrows: patterns of segregation between Sagebrush Sparrow ( A. nevadensis ) and Mojave Bell's Sparrow ( A. belli canescens ) across Lower Colorado Desert vegetation assemblages

. Bell’s Sparrows ( Artemisiospiza belli ) have only recently been recognized as distinct from Sagebrush Sparrows ( A. nevadensis ), and the “Mojave” subspecies ( A. b. canescens ) shares an overlapping wintering distribution with Sagebrush Sparrow in the Mojave and Sonoran deserts of southeastern California and western Arizona. We lack understanding of the two species’ respective wintering habitat preferences and the degree to which they interact or segregate on their wintering grounds due to the difficulty in separating them in the field and to their previous classification as one species. We captured and sampled 74 Artemisiospiza sparrows from 5 sites across western Arizona, supporting field identifications with genetic analyses of mitochondrial DNA to confirm species and molecular sexing of sampled individuals. Bell’s Sparrows and Sagebrush Sparrows segregated into different habitat types across our study area, with only one species detected at four of five study sites. Bell’s Sparrows comprised 82% (n = 33) of Artemisiospiza sparrows captured at the 5th site at Robbins Butte. Broadly, Sagebrush Sparrows were found in more upland, well-drained locations that were less vegetated with xerophytic scrub. Bell’s Sparrows were found in more vegetated locations with halophytic Mojave seablite ( Suaeda nigra ) and saltbush ( Atriplex ) adjacent to mesquite and tamarisk woodlands. Bell’s Sparrow sex ratios were significantly female-biased (binomial test: n = 56, observed k = 48 females, expected k = 28 females for assumed p = 0.5, Pr [ k < = 8 or k = > 48] < 0.0001; 95% CI = 0.369 – 0.631 for assumed p = 0.5) at Fort Mohave and Robbins Butte, the 2 sites where Bell’s Sparrows were found. Our observed sex-ratios and well-documented year-round presence of Bell’s Sparrows on and near the breeding grounds suggest that Bell’s Sparrow males and females employ different migration strategies, a phenomenon not previously documented for this taxon.


INTRODUCTION
The Artemisiospiza "Sage Sparrows" are demonstrably difficult to separate in the field (Pyle 2013), and their taxonomic status has been debated for over 120 years (Cicero 2010). Bell's Sparrow (Artemisiospiza belli) and Sagebrush Sparrow (A. nevadensis) were once grouped together as Sage Sparrow in the genus Amphispiza, along with Black-throated Sparrow (A. bilineata). Klicka and Spellman (2007) found that the genus Amphispiza was not monophyletic, Sage Sparrows were not as closely related to Blackthroated Sparrows as believed, and Sage Sparrows were more closely related to the clade of "grassland" sparrows (genera: Pooecetes, Oriturus, Passerculus, Ammodramus [in part, and now labeled Ammospiza], Melospiza, and Xenospiza; Klicka and Banks 2011). Klicka and Spellman (2007) also suggested a new Both Sagebrush and Bell's sparrows winter in southeastern California and western Arizona (Phillips et al. 1964, Martin andCarlson 2020a, b). Species-specific wintering habitat preferences and the degree to which the two species segregate or interact on their wintering grounds nonetheless remains largely undescribed (Rosenberg et al. 1991). This knowledge gap is due to the species' apparent niche overlap during the nonbreeding season, the difficulty of separating the two species in the field, and to the very recent taxonomic separation of A. b. canescens and A. nevadensis into separate species.
Following previously published methods for morphometric and genetic analyses on the Artemisiospiza sparrows (Johnson and Marten 1992, Cicero and Johnson 2006, Cicero and Koo 2012 and a photographic guide to plumage differences (Pyle 2013), we sought to identify locations across western Arizona where the two Artemisiospiza sparrow species were found during the winter, to estimate their relative proportions at these sites and to describe the vegetation communities present at these locations.

Study area
We compiled recent sight reports of Artemisiospiza sparrows in late 2013 and early 2014 to help us locate individuals to sample from 5 locations spanning approximately 250 km across western Arizona in Maricopa, Yuma, and Mohave counties (Fig. 1). Sampling sites (Table 1) were selected to represent a variety of plant assemblages that could potentially host wintering Sagebrush and Bell's sparrows. Field sampling was from 6 to 16 February 2014. Figures 2-6 depict the five study sites, with sparrow flushing areas outlined in yellow and five vegetation plots for each site depicted with yellow circles.

Robbins Butte Wildlife Area
The Robbins Butte study site (Maricopa County, AZ; 33. 316˚N, 112.634˚W) is within the southern periphery of the Gila River floodplain in western Maricopa County, west of Arizona Highway 85. Tamarisk (Tamarisk sp.) riparian woodland and small cultivated plots of winter wheat (Triticum aestivum) and alkali sacaton grass (Sporobolus airoides) planted for wildlife forage extend to the north and northwest of the site (Fig. 2). The toe of an alluvial fan that descends from the Buckeye Hills extends uphill to the south, transitioning to mixed Sonoran Desert scrub. A portion of the site appears to have once been irrigated by a small canal that bisects the site, suggesting that the area was previously used for pasture or cultivated crop production several decades ago. The LANDFIRE US 220 Existing Vegetation Type classification system (LANDFIRE 2022) classifies this site as a mix of western warm temperate fallow/idle cropland, North American warm desert ruderal and planted scrub, Sonoran Paloverde-mixed cacti, and Sonora-Mojave creosotebush-white bursage desert scrub.

Tacna
The Tacna study site (Yuma County, AZ; 32.718˚N, 113.924˚W) is on a sandy mesa at the northern edge of the Mohawk Valley, approximately 2 km south of cultivated cropland in the Gila River floodplain (Fig. 3). The study site was once irrigated cropland, perhaps used for citrus production, and is in a ruderal state, not yet returned to creosote bush (Larrea tridentata) and white bursage (Ambrosia dumosa) scrub that persists in untilled land with similar elevation, physiography, and soils south of the study site (December 1985

Data collection
Mist-netting Artemisiospiza sparrows were captured following a flush-netting method described in Gordon (2000). Scouting was required at each site, given that (1) preceding reports of recent Artemisiospiza sightings were vague as to the precise location of potential captures, the reports occasionally dated, and (2) wintering Artemisiospiza are nonterritorial (Martin and Carlson 2020a). Presence of Artemisiospiza sparrows determined where nets were placed; all vegetation types were scouted at sites with multiple vegetation assemblages (Robbins Butte, Fort Mohave, Boundary Cone), and the yellow-outlined capture areas depicted in Figures 2-6 correspond to where Artemisiospiza sparrows were found and rounded up toward net lines.
Upon locating Artemisiospiza sparrows during scouting, we aligned 4 to 6 mist nets (2 x 10 m, 32-and 36-mm mesh) from end to end in a line near sighted birds, with care taken to hide nets in natural gaps in shrub cover when possible. We spaced ourselves and typically 8 to 15 volunteers into a large C-shaped semi-circle then slowly herded sparrows toward the net line. The net line was then dismantled and placed in a new location, contingent on where uncaptured birds moved while herding them toward nets or where new groups of potential captures were found. Occasionally the net line was left in the same location, and we herded birds toward the nets from the opposite direction. Two to approximately 12 net placements were attempted at each site, depending on the success of preceding flushing attempts.

Field measurements
We banded all captured Artemisiospiza with uniquely numbered aluminum USGS bands and recorded a series of morphological measurements on each capture (Table 2). Specific individuals are described in text and in tables and figures with the final four digits of their band's unique number. We were constrained in the number of measurements we could take within a safe handling time

Measurement Description
Wing Chord Unflattened wing length to the nearest millimeter using a wing rule. Measurement taken while the wing was in its natural closed position, from the tip of the carpal joint to the end of the longest primary Flat Wing Chord Wing length as above, but with the wing gently flattened on the wing rule, producing a slightly longer measurement than unflattened wing chord Body Molt Assessment of the amount of molt visible in contour feathers and wing and tail coverts, with 0 representing no body molt, and 3 representing heavy body molt, with greater than 1/2 of body feathers in molt because we also wished to obtain blood samples and photographs from captured birds, and our method of capturing birds required netting several birds simultaneously. The entire suite of morphological measurements was not taken for some individuals (e.g., the bird escaped during processing, or the bird showed signs of distress and was released prior to completion of measurements).

Photographs
We photographed captured Artemisiospiza sparrows with a Nikon digital single lens reflex camera and a 300 mm lens. We photographed each individual's back, tail, nape, crown, profile, front, and opened wing, and all images are currently available at https://doi.org/10.5061/dryad.gmsbcc2tp. Additionally, notes were taken on quality and extent of dorsal streaking in the back, on malar streak color and extent, outer rectrix patterning, and coloring in the supercilium.

Blood sampling
Captured birds were gently placed in small cloth bags with drawstrings and brought to a banding station ca. 100 m from the net line. A blood sample was drawn from the brachial vein with a 27-gauge ½-inch needle following Owen (2011). One to two drops of blood were absorbed onto both sides of an Advantec Nobuto Blood Filter Strip. Samples were allowed to air dry and then placed into paper coin envelopes. Sample envelopes were stored outside of direct sunlight at ambient temperature and labeled with the bird's unique band number, date collected, sampling location, and suspected species. Remaining blood is archived with the Kovach Laboratory at the University of New Hampshire.
Genetic analysis DNA was extracted from blood samples using a Qiagen DNeasy blood and tissue kit (Qiagen, Valencia, California, USA), following manufacturer's protocol. The sex of the individuals was determined using a molecular sexing protocol that leverages a length polymorphism in the CHD 1 gene of Z and W chromosomes. The CHD 1 gene was amplified using primers P2 and P8, following methods of Griffiths et al. (1996) and Fridolfsson and Ellegren (1999). Amplified products were resolved in a 2% agarose electrophoretic gel. This assay produces two fragments of different size for females and a single fragment for males.
Species identification was performed using the mitochondrial DNA restriction fragment length polymorphism assay described in Cicero (2004). A 900-bp fragment of the cytochrome B gene was amplified with primers L14987 and H15916, as in Cicero and Johnson (2001). Restriction digests were performed in 10 μl volumes, with 1 μl of the AluI enzyme, 1 ml of enzyme-specific primer (New England Biolabs, Beverly, Massachusetts), 4 μl of DNA, and 4 μl of water, overnight at 37°F. Restriction fragment banding patterns were visualized after electrophoresis in a 2.5% agarose gel. Interpretations followed the haplotype patterns identified by Cicero and Koo (2012).

Vegetation sampling
We characterized vegetation at each study site in December 2014, following a modified protocol described in Fleishman et al. 2014 (1) 2 vegetation plots were set using the net placements used on the 2 days of sampling at Fort Mohave, and the 3 other vegetation plots were set 250 m apart heading eastward from the 2 net placements (Fig. 5); and (2) the first net placement location for each of the 5 days spent at Robbins Butte was used for that site's 5 vegetation plots (Fig. 2).
Within each plot we walked 3 30-m transects radiating from the plot's center, with the first transect direction set randomly and the following 2 transects established at 120° from the previous transect. Assessments were located at 8 m, 16 m, and 24 m along each transect to determine species presence (or in the event of no plant species present, bare ground) in each cardinal direction and along the transect line, for a total of 5 assessments per stop, 15 assessments per transect, 45 assessments per plot, and 225 assessments per site. At each assessment location, we used an ocular tube approximately 11.5 cm long and 2.5 cm in diameter to take measurements at a 45° angle downward from the line of sight. A "hit" was tallied at each assessment when a perennial plant species was observed within the tube while looking downward (e.g., at assessment point 1, at 8 m on the transect tape, looking northward through the tube, only creosote bush was observed; looking eastward, only bare ground was observed through the tube; looking southward, creosote bush and boxthorn (Lycium spp.) were observed through the tube; looking westward, only creosote bush was observed through the tube; and looking forward on the transect tape, only white bursage was observed through the tube, totaling three creosote bush hits, one boxthorn hit, one white bursage hit, and one bare ground-only hit. All hits were aggregated by plot and then by site. Low and high heights (cm) were also recorded for each species recorded as hits on each transect.
At the end of each transect, an 11.5 m-radius circle was set and tree canopy radius (cm), tree height (cm), and trunk diameter (cm) were recorded for each arborescent species. Diameter was typically measured at ground level or 10 cm above ground level.
We used aerial imagery provided by ESRI (2022)

Species identification
We captured 86 Artemisiospiza sparrows across the 5 study sites and analyzed blood samples from 74 unique individuals (Table  3).
Genetic results were consistent with 68 of 74 (92%) field identifications. Four of 56 individuals (7%) identified in the field as Bell's Sparrows shared the BC haplotype group consistent with Sagebrush Sparrow (Cicero and Koo 2012). Two of 18 individuals (11%) identified in the field as Sagebrush Sparrows shared the AD haplotype group associated with Bell's Sparrow.
Haplotype designations identified using the Alu RFLP approach do not distinguish the species 100% of the time (Cicero and Koo 2012). Therefore, sample totals (Table 3) are summarized by (1) all 74 individuals as identified by genetic analysis, (2) all 74 individuals as identified in the field by phenotype, and (3) for 68 individuals, excluding the 6 individuals with inconsistent field and genetic assignations. Morphometric measurements for the 6 individuals with inconsistent field and genetic assignations are provided in Table 4 and individual cases are described as follows: . 8022: Male (Robbins Butte), identified in the field as Sagebrush Sparrow and found with the AD haplotype group. Morphometric measurements classify as Sagebrush Sparrow, notably with a wing chord outside of the range of variation for Bell's Sparrow males presented by Cicero and Johnson (2006), moderate but distinct mantle streaking through the back, and pale gray malar streaks consistent with Sagebrush Sparrow.

Segregation across study sites
We found that the two species segregated across the five study sites, with only Sagebrush Sparrows found at Tacna, Kofa Mountains Bajada, and Boundary Cone (Table 3). Only Bell's Sparrows were found at Fort Mohave. Bell's Sparrows comprised approximately 82% of the individuals sampled at Robbins Butte.

Robbins Butte
The Robbins Butte site consisted nearly entirely of allscale (Table  6), with isolated honey mesquites covering 2-5% of the study area, placing it within the Atriplex polycarpa shrubland (allscale scrub) and the Prosopis glandulosa woodland (mesquite bosque, mesquite thickets) alliance (Prosopis glandulosa/Atriplex spp. association) defined by the California Native Plant Society  (Sawyer et al. 2009). Additional perennial species included isolated creosote bush and boxthorn that stood over 2 m in height, as well as a small number of blue palo verdes.

Fort Mohave
The Fort Mohave site consisted primarily of quailbush and Mojave seablite (also referred to as inkbush and bush seepweed) in silty, alkaline soil (Table 6), placing it within the Suaeda moquinii (bush seepweed scrub) shrubland alliance (Sawyer et al. 2009). Large contiguous patches of Mojave seablite and quailbush were found in lower-lying areas with potential to retain moisture after rainfall events, and the site contained patches of desert pavement largely devoid of vegetation left from clearing and grading activities in the mid 2000s. The site held an isolated cluster of 8-10 m-tall Athel tamarisk. The presence of Mojave seablite and tamarisk indicated a high-water table as well as potential for occasional flooding and additional potential hydrologic input from irrigation runoff.

Tacna
The Tacna site was the most sparsely vegetated of the five study sites, consisting primarily of isolated fourwing saltbush clumps, open sandy areas, and annual vegetation (Table 6), placing the site within the Atriplex canescens (fourwing saltbush scrub) shrubland alliance and also described by the fourwing saltbush (Atriplex canescens) shrubland association Hartman 2007, Sawyer et al. 2009).

Kofa Mountains Bajada
The Kofa Mountains Bajada site consisted of undisturbed, sparsely vegetated desert pavement populated primarily by creosote bush, with white bursage, brittlebush, big galleta grass, and rhatany associated with small, shallow channels with alluvial soils that cut into the desert pavement (Table 6). Isolated trees (foothills palo verde, ironwood, and catclaw acacia) and saguaro covered less than 1% of the site and were found along larger channels that descended the bajada. The site's vegetation composition is consistent with the Larrea tridentata -Ambrosia dumosa shrubland alliance (creosote bush -white bursage scrub) and the creosote bush -burro bush/big galleta (Larrea tridentata -Ambrosia dumosa -Pleuraphis rigida) shrubland association Hartman 2007, Sawyer et al. 2009). Aspects of the Carnegiea gigantea -Parkinsonia microphylla -Prosopis velutina provisional shrubland alliance (saguaro -foothills palo verdevelvet mesquite desert scrub) were associated with the larger wash channels with emergent arborescent vegetation (California Native Plant Society 2022).

Boundary Cone
The Boundary Cone site consisted of a wide valley of alluvial soil, gravel, and small boulders. Its vegetation composition was similar to the Kofa Mountains Bajada site, with creosote bush, white bursage, brittlebush, and rhatany (Table 6), also placing it within the Larrea tridentata -Ambrosia dumosa shrubland alliance. A large wash that bisected the site also held cheesebush and catclaw acacia. The Boundary Cone site was more typical of Mojave Desert creosote bush -white bursage scrub, lacking arborescent tree species and saguaro found on the Kofa Mountains Bajada that are associated with the Sonoran Desert.

DISCUSSION
We sampled wintering Artemisiospiza sparrows across a variety of common, widespread vegetation community types in western Arizona, obtaining blood samples from 74 individuals to provide genetic context for our field identifications. Our sample set was not an exhaustive representation of all habitats within the overlapping Sagebrush and Bell's Sparrow distributions, but our study provides a first examination of differences in wintering habitat selection between two species that are difficult to separate in the winter without assessments in the hand.
With the aid of previous reports of plumage differences (Pyle 2013) and morphometric differences (Cicero and Johnson 2006), we found that Bell's Sparrows and Sagebrush Sparrows are readily separable in the hand, and 68 of 74 field identifications (92%) were consistent with genetic results. Cicero and Koo (2012) reported a potential for shared haplotype markers between the species, with a small number of A. b. canescens found with the Hap10 BC type haplotype, and one A. nevadensis in their sample set found with the AD haplotype. Four of our six individuals with inconsistent field and genetic assignations were consistent with Bell's Sparrow in size and plumage characteristics but were found to have the Hap10 BC type haplotype. One individual was consistent with Sagebrush Sparrow in size and plumage but had the AD type haplotype, and one individual was consistent with Sagebrush Sparrow in size, intermediate in plumage, and had the AD type haplotype.
We found Bell's Sparrow sex ratios to be significantly female biased at our Arizona study sites. Fort Mohave and, in particular, Robbins Butte are on the southeastern periphery of the Bell's Sparrow wintering distribution. However, the sites are 250 km apart, suggesting a widespread regional pattern of differential migration at the Bell's Sparrow wintering distribution's southeastern periphery.  (Sullivan et al. 2009). We suspect that wintering Bell's Sparrow sex ratios in the western Mojave Desert will be found to be male-biased (Ketterson andNolan 1976, Cristol et al. 1999).
Prior to this study, little information existed on differences in habitat selection by Bell's and Sagebrush sparrows within their overlapping wintering distributions. We found wintering populations of Bell's and Sagebrush sparrows segregated across different vegetation communities within our study area, with mingling between the two species to be the exception and not the norm. Broadly, Sagebrush Sparrows were found at less vegetated, more xerophytic upland scrub than Bell's Sparrows, whereas the latter were found in more heavily vegetated, halophytic allscale and quailbush-seablite scrub with silty soils that were within or adjacent to mesquite or tamarisk woodlands. The two species only co-occurred at Robbins Butte, near the eastern perimeter of the Bell's Sparrow wintering distribution, where the sample was biased toward Bell's Sparrow (26 of 30 individuals with consistent genetic and field identifications). We did not record visual observations of the species interacting or foraging together at Robbins Butte, only that a small number of Sagebrush Sparrows were swept up in our large round ups that flushed birds into net lines. Garrett and Dunn (1981) noted Bell's Sparrow' affinity for Suaeda on the Colorado River, as did Meents et al. (1982), although Meents et al. did not differentiate between Sagebrush and Bell's sparrows. Patten et al. (2003) also reported that Artemisiospiza sparrows could be found in inkbush and saltbush habitats in the Salton Sink, but they did not recognize a separation between Amphispiza belli nevadensis and A. b. canescens, classifying the reports as A. b. nevadensis (see Unitt 2002, Cicero andJohnson 2006). Our findings suggest that the great majority of the Artemisiospiza sparrows previously reported at Suaeda sites in western Arizona and the Salton Sink were Bell's Sparrows, and that most were females.
Georeferenced photographs from locations within the lower Colorado River riparian corridor and the Salton Sink submitted to online citizen science platforms such as eBird and iNaturalist are consistent with this pattern, with identifiable photographed reports from the Salton Sink and the Colorado and Gila river valleys consisting typically of Bell's Sparrows, with uncommon (e.g., at a settling ponds site near Borrego Springs, California, and around Robbins Butte) to only sporadic instances of photographic documentation of Sagebrush Sparrows in expected Bell's Sparrow habitat in southeastern California and western Arizona (Sullivan et al. 2009, GBIF 2023; https://www.inaturalist. org).
Meents et al. also found temporal changes in diet across the wintering season, with Artemisiospiza sparrows consuming more seeds later in the winter as invertebrate food sources presumably became scarcer, resulting in birds clustering at study sites with Mojave seablite later in the season (1982). Our study period occurred during the first half of February, leaving open the possibility that Bell's Sparrows may be found to be more dispersed across the landscape earlier in the wintering season. Declines in available food, such as during prolonged drought, may produce conditions in which wintering Sagebrush Sparrows will be compelled to abandon xerophytic upland scrub with ephemeral food sources and be more likely to be found in Bell's Sparrowassociated wintering habitats, mixing with Bell's Sparrows.
Notably, unlike our other study sites, Robbins Butte was in the midst of moderate to severe drought in early February 2014 (NDMC et al. 2014), perhaps an explanation for the sympatric occurrence of the two species at this site (27:6 Bell's to Sagebrush capture ratio).
We suspect that our observed patterns in wintering Bell's Sparrows' habitat selection may be more specific to sites outside of the Bell's Sparrow breeding distribution, given (1) that a portion of the Bell's Sparrow population remains on the breeding grounds through the winter (Sullivan et al. 2009), and (2) A. b. canescens nest in a variety of upland scrub habitats more similar to the desert scrub habitats in which we encountered Sagebrush Sparrows while sampling in western Arizona (Martin and Carlson 2020a). This appears consistent with photographed reports in eBird and iNaturalist that provide widespread georeferenced photographic documentation of wintering Bell's Sparrows across a variety of habitats within the Bell's Sparrow breeding distribution in the western Mojave Desert (Sullivan et al. 2009, GBIF 2023. Wintering Sagebrush Sparrow reports from within the A. b. canescens breeding distribution in the western Mojave Desert are notably uncommon (Garrett 2020; https://www.youtube.com/ watch?v=6RjC_fpietw), particularly given that a large amount of apparently available Sagebrush Sparrow wintering habitat exists here. This raises the intriguing possibility that Bell's Sparrows may exclude wintering Sagebrush Sparrows from habitats within the Bell's Sparrow breeding distribution or that wintering Sagebrush Sparrows may avoid the Bell's Sparrow breeding distribution in some fashion. We recommend additional winter sampling in the western Mojave Desert to document the two species' presence, habitat use, and sex ratios in this region. Rosenberg et al. (2019) reported large scale declines in North American bird populations, and there is increasing awareness that birds are exposed to different threats across their life cycles (North American Bird Conservation Initiative 2022). Our study has identified different patterns in desert scrub selection by two taxa of Artemisiospiza sparrows that were once considered the same species with the same life cycle. These different vegetation types face different types and degrees of threats from climate change, fire disturbance, and conversion to solar, urban, and agricultural development. Our study also provides evidence that A. b. canescens Bell's Sparrows may employ differential migration, a strategy previously undocumented for the species. If occurring, sex-based differential migration will present additional conservation challenges and opportunities for this taxon.