Scroll Top

Historical changes in bird communities at Quitobaquito Springs, Arizona and impacts of climate, spring flow, and shift from Indigenous to federal-agency management

By: Aaron D. Flesch, Gary Paul Nabhan, and Peter Holm

For nearly a century, bird communities at the oasis of Quitobaquito Springs in the Sonoran Desert have attracted desert ecologists. This spring-fed oasis is among the most biologically and culturally significant sites in arid southwestern North America, but has experienced marked changes in management and is threatened by climate change, border, and other development. We assembled data on birds at Quitobaquito across 83 years (1939–2022) and evaluated the patterns and potential drivers of changes in communities. We found evidence of marked community shifts that included apparent loss and replacement of species dependent on riparian and wetland vegetation, mud flats, agricultural fields, and human settlements, to species that use deeper more open-water environments, generalists, and desertscrub. Strong associations between community composition and shift from Indigenous to federal-agency management, temperature, and spring flow suggest important drivers of change, but associations with precipitation and border development were weaker. Such patterns match those for bird communities at broader scales in nearby and other aridlands and indicate impacts of increasing aridification. Our work represents the first analysis of changes in faunal assemblages at Quitobaquito that quantitatively assesses potential drivers of change, and provides insights for management and conservation.


Desert oases and other similar mesic riparian environments are major contributors to biodiversity in aridlands, and have attracted wildlife and people for millennia (Naiman et al., 1993; McNeely, 2003). These and other freshwater aquatic ecosystems, however, face multiple and often compounding threats linked to increasing anthropogenic pressure and climate change worldwide, especially in aridlands (Shepard, 1993; Naiman and Turner, 2000; Jackson et al., 2001). The Quitobaquito Springs habitat complex is considered an outstanding example of a spring-fed desert oasis and among the most biologically and culturally significant sites of comparable size in arid southwestern North America (Cohn, 2001; Nabhan et al., 2023). Located along the U.S.-Mexico border in the central Sonoran Desert, Quitobaquito Springs is mostly within Organ Pipe Cactus National Monument in Arizona, but is closely connected to the adjacent Rio Sonoyta riparian corridor and Pinacate Biosphere Reserve in neighboring Sonora, Mexico. Quitobaquito Springs is also significant culturally and spiritually as part of the ancestral homeland of the Hia c-ed O’odham and other affiliated tribes. Despite being protected as part of a U.S. national monument and registered UNESCO Biosphere Reserve, Quitobaquito has undergone profound changes in management and is now bisected by the U.S.-Mexico border wall. Like many globally-important sites in arid western North America, Quitobaquito is highly threatened by climate change and regional groundwater extraction, and also faces additional threats linked to border-related activities.

Quitobaquito Springs includes a diverse mosaic of environments that include open water, wetlands, hydroriparian, and mesoriparian areas that interact within adjacent xeroriparian and Sonoran Desert upland environments (Stevens et al., 2021). Ecological interactivity at Quitobaquito is amplified by the rarity of other sources of perennial surface water in the region and hyper-aridity of the surrounding landscape that increases exchange of subsidies between wetland and surrounding upland environments (Stevens, 2020). As a result, Quitobaquito Springs provides important stopover and breeding habitats for migratory and resident birds and many other species of wildlife (Nabhan, 2003). In addition to birds, the site supports some endemic aquatic species including endangered Quitobaquito desert pupfish (Cyprinodontidae: Cyprinodon eremus), the rare Sonotya mud turtle (Kinosternon sonoriense longifemorale), endemic Father Kino’s tryonia springsnail (Cochliopidae: Tryonia quitobaquitae), and a diverse array of dragonflies and other invertebrates (Kingsley 1998). Overall, Quitobaquito Springs embodies the definition offered by Cartwright et al. (2020:245): “Natural springs in water-limited landscapes are biodiversity hotspots and keystone ecosystems that have a disproportionate influence on surrounding landscapes despite their usually small size.”

Here, we evaluate the patterns and potential environmental drivers of faunal change at Quitobaquito Springs between the late 1930s and 2022 by using bird communities as a focal group. Despite being just small facets of overall biodiversity, birds are useful surrogates for assessing environmental change. This is because most bird species are conspicuous and easy to survey, linked to specific sets of resources and conditions required for nesting, foraging, and other needs, and given relatively high availabilities of historical data (Temple and Wiens, 1989; Suarez and Tsutsui, 2004). We amalgamated a fairly large set of historical and recent observations of birds from multiple sources and evaluated patterns of community change across time. To guide evaluations, we focused on species’ traits linked to habitat use and critical limiting resources because they provide insight into the patterns and potential environmental drivers of change. We assessed associations between changes in bird communities and variation in a broad suite of environmental factors by using historical and recent data on spring flow, precipitation, temperature, and differences in management across time. Among the many historic changes in management, we gave special attention to shifts from Indigenous to federal-agency management that occurred in the late 1950s and 60s, and factors linked to recent construction of the U.S.-Mexico border wall, which now bisects the site. Finally, we evaluated the implications of observed patterns for future management and conservation efforts of the Quitobaquito Springs habitat complex.

2. Materials and methods

2.1. Study area

The Quitobaquito Springs habitat complex is located in both Pima County, Arizona and the adjacent Municipio de Plutarco Elias Calles in Sonora, Mexico (31.9445089°N, −113.0176651°W), and has a base elevation of 351 m. This area has supported various environments and vegetation communities over time. They include: 1) open-water environments with aquatic plants and a small pond, 2) small but variable areas of barren mudflats on the water’s edge, 3) seeps and wet reservoir edges covered by emergent, semi-aquatic hydroriparian vegetation, 4) a small patch of tall mesoriparian gallery forest dominated by broadleaf tree species, 5) dense xeroriparian woodlands and scrub dominated by microphyllous trees and shrubs, 6) salt flats dominated by halophytes, and 7) larger surrounding upland areas of desertscrub typical of the Arizona Uplands and Lower Colorado River Valley subdivisions of the Sonoran Desert (Nabhan et al., 1982; Felger et al., 1992). There are also remnants of a historic orchard of pomegranates, figs, and other fruit trees that was tended for decades by Hia c-ed O’odham and other inhabitants, until the U.S. National Park Service (NPS) assumed control of the site in 1957 (Bennett and Kunzmann, 1989; Nabhan, 2003, Fig. 1). In 1961, NPS removed the historic buildings and further impounded the pond to increase water levels but maintained the orchard until the early 2000’s when pond outflow declined and orchard trees died. The 1851 border delineation follows the Gadsden Purchase that essentially bisected the ancient Hia c-ed Tohono O’odham settlement. This area drew upon at least seven artesian springs associated with a fault line that runs along the base of the Quitobaquito Hills, which is a series of narrow mostly granitic ridges that rise 25 m above the valley floor. This historic cultural landscape has several designations and levels of formal protection but experienced a range of changes in management. Here, we focus on bird communities in the ∼5 ha site immediately north of the U.S.-Mexico border that is centered on the pond and where most survey effort for birds was focused over time.

2.2. Bird surveys and data

To assess patterns of temporal change in bird communities, we synthesized historical and recent data on the occurrence and abundance of birds at Quitobaquito, and information on seasonality and habitat use. We used data from the literature, unpublished data from colleagues, NPS, and our own studies, and supplemented these sources with lists from eBird (Fink et al., 2020; eBird, 2022). We only considered observations from the protracted breeding season of most species that use the site (Feb.–July), but included all breeding, possibly breeding, and migratory species detected during this period in analyses. We focused on this broad breeding season because it was when most past survey effort was focused and because occupancy during this period is often prolonged and strongly tied to specific resources and conditions, which is useful for understanding environmental change. Moreover, we only used data sets from formal surveys rather than incidental observations to ensure efforts considered entire bird communities rather than just species of interest or special status.

Multiple field methods and survey strategies have been used to sample bird communities at Quitobaquito since 1939. The first professional survey of birds was completed by Lawrence Huey of the San Diego Museum of Natural History in 1939, just three years after Organ Pipe Cactus National Monument (OPCNM) was designated by President Franklin D. Roosevelt. Huey surveyed birds over 10 days in late winter (4–9 March) and spring (28 April – 1 May), and five days in fall (26–30 Nov; Huey, 1942) by recording species and approximate numbers he observed or collected in areas around the springs. Max Hensley (1954) used more standardized methods during 13, 1-hr surveys during which he walked the circumference of the pond and recorded all species and individuals he detected between 28 Feb. and 16 June 1949, and on 24 Aug 1948. These historical efforts provide an important baseline because they were completed when Hia c-ed O’odham residents still lived in and managed Quitobaquito and were before NPS had much management influence (Nabhan et al., 1982; 2003).

Ornithologist Stephen M. Russell visited Quitobaquito on 15–16 April 1965 and recorded all birds he detected (Russell, pers. comm.). In the 1980s, ornithologist Amadeo Rea and colleagues surveyed Quitobaquito over three days in mid-March and three days in mid-May 1982 recording the maximum number of individuals they detected of each species (Nabhan et al., 1982; Reichhardt et al., 1986). R. Roy Johnson et al. (1983) used a spot-mapping method to survey birds in two plots around the pond 14 times between 28 Mar. and 26 May 1983. They reported the number of breeding pairs of each landbird species within plots but did not record waterbirds or estimate abundances of species visiting plots that were breeding in uplands outside plots.

More recently, in 1995 and 1996, M. Halterman and S. Laymon (2000) used the point-count method to survey birds at Quitobaquito. They completed 10-min. counts at four stations spaced 200 m apart during 6 visits per year between early Mar. and late May, and recorded all individuals they detected. Point counts involve recording birds seen or heard at fixed points at which observers are stationary, and then moving quickly to subsequent points. A long gap in survey effort persisted into the 2000s as OPCNM staff focused bird monitoring at other sites from 1997 to 2004 and then suspended monitoring after 2004. In 2021, bird monitoring at Quitobaquito and elsewhere in OPCNM was reestablished by P. Holm and A. D. Flesch at 25 sites including Quitobaquito. In 2021 and 2022, eight point-count stations were surveyed three times per year between mid-Mar. and late May (Flesch 2022a). To augment recent observations, we also amalgamated complete checklists from eBird for years 2016, 2020, and 2022 during which 10–12 visits were completed per year between Mar. and July by various observers (Fink et al., 2020; eBird, 2022; PH, unpubl. data). Data from eBird were gathered by professional biologists and skilled amateurs, included the number of observations of each species and often effort in time or distance covered, and both landbird and aquatic species. Data from eBird were quality controlled by assessing questionable species identifications (e.g., outside typical distributional limits) and removing them before analyses. Cumulatively, surveys we considered were from 11 different years between 1939 and 2022 and based on 90 daily visits during the breeding season.

2.3. Environmental measurements

To assess associations between variation in bird communities and environmental factors, we considered data on spring flow, precipitation, temperature, changes in management, and temporal factors. For management, we created an indicator variable (e.g., 0/1) to note presence and management by Indigenous residents that had small houses, agricultural fields, and a small orchard that were tended until NPS acquired the site in 1957 (e.g., 1939-49 vs. other years). To assess potential impacts of border wall construction along the international boundary, we created an indicator variable for presence of the new expanded wall (e.g., 2022-21 vs. other years). The expanded wall included extensive road widening, construction of a concrete footer and associated groundwater pumping, and more than doubled the height of the existing border fence; in OPCNM expansion began in 2019 but was not completed at Quitobaquito until after the 2020 breeding season. To quantify time, we computed the number of years between the initial baseline year of 1939 (e.g., year 0) and 2022 the last year of observation (year 83). We also classified the study period into three time periods: a historical period from 1939 to 1949, an intermediate period from the mid-1960s to mid-1900s (1965–1996), and a contemporary period from 2016 to 2022.

To describe climatic variation, we considered a small set of weather factors that we expected could influence bird communities through either direct (e.g., heat stress) or indirect mechanisms (e.g., impacts to food or vegetation). For precipitation (P), we considered total annual P from March of the prior year through February of the current year (e.g., 12 months before survey period), prior warm-season P from April through September one year prior, recent cool-season P from prior October through March of the current year, and running averages of annual P over the prior 3 and 5 years. For temperature (T), we computed mean annual T (MAT) and mean maximum T (Tmax) and running averages of MAT and Tmax in the same periods as for P. To describe historical and recent weather, we used locally downscaled estimates of P, MAT, and Tmax interpolated from weather stations (Wang et al., 2016) for each year from Climate NA version 7.4 software.

To quantify spring discharge (liters/min.) across time, we used three data sources from Quitobaquito. During early years before NPS and U.S. Geological Survey (USGS) monitoring, we obtained an estimate from 1938 gathered by Gould (1938; and reported by Cole and Whiteside, 1965) that we used to describe historical discharge in 1939 and 1949. For the period between Jan. 1974 and May 2021, we used NPS data (provided PH) to estimate spring discharge that covered most other calendar years. Data from Cole and Whiteside (1965) for 1964 were not used to quantify spring discharge during Russell’s bird surveys because they only measured pond outflow that was biased low. Instead, we used mean annual discharge across the three earliest years (1974–1976) of NPS monitoring to estimate approximate flow in 1965. Because discharge often varied monthly (e.g., was higher during cooler winter months) and there were some missing values, we first computed monthly means in each year and used these values to compute annual means. Data from 1995 were missing and so we used mean discharge from 1991, 1992, and 1994 to estimate approximate flows in 1995. Data on spring flow in late 2021 and 2022 were unavailable from NPS, and so we used data from USGS that are based on data loggers placed at the same location. Long-term data on vegetation cover and extent of surface water were not available from early years and hence not used.

2.4. Data analyses

We used ordination and other multivariate techniques to assess temporal variation in bird communities and associations between communities and environmental and temporal variation. We considered only diurnal bird species because few observers surveyed at night and eliminated species that breed late in the survey period to reduce bias (e.g., Blue Grosbeak; see Appendix 1 for scientific names) as surveys in some years were completed before they typically arrive. To reduce noise in the final community dataset, we censored records of 33 rare species detected only once, typically with just 1–2 individuals (Table S1). All quantitative analyses reported here focus on 111 bird species and 2 species groups (Solitary-type Vireo, Western-type Flycatcher) that are summarized in Table S2. To generate an index of relative abundance that accounted for differences in survey methods and effort, we computed the mean number of individuals of each species detected per daily visit, or point for point-count surveys, during each study year. Effort in survey time or distance covered was not always known, so we relativized data by standardizing (z-scoring) estimates and rescaling them such that the lowest estimate for each study year equaled one. Hence, estimates we used quantified relative abundance in each study year and represented the number of standard deviations from mean overall abundance across time of each species. To improve linearity and decrease emphasis on highly abundant species, we then square-root transformed these estimates of relative abundance.

To help understand how environmental variation contributed to changes in bird communities, we evaluated patterns in the habitat use and affinities of species contributing to observed variation in communities. These trait-based groups were derived from information on habitat use and limiting resources from the literature and our knowledge of bird-habitat relationships. Groups we considered included species associated with desertscrub (e.g., Black-tailed Gnatcatcher, Cactus Wren), exotic species or those linked to humans (e.g., Eurasian Collared-Dove, Barn Swallow), agriculture or open country (e.g., Killdeer, Lark Sparrow), wetlands birds that use semi-aquatic and emergent vegetation (e.g., Sora, Red-winged Blackbird), shallow water or mud flats at the reservoir edge (e.g., most shorebirds and waders), open water (e.g., waterfowl, American Coot, Wilson’s Phalarope), water edge (e.g., Belted Kingfisher, Song Sparrow), cottonwood trees and other mesoriparian elements (e.g., Hooded Oriole), dense xeroriparian shrub-obligates (e.g., Bell’s Vireo, Yellow-breasted Chat), aerial foragers (e.g., most swallows), generalists (e.g., Common Raven), and a broad other category of species that use a wide range of environments many of whom are migrants.

To describe spatiotemporal variation in bird communities, we performed nonmetric multidimensional scaling (NMDS) on a Sørensen distance matrix in up to six dimensions and used stress values to select an optimal solution. NMDS is an iterative ordination technique that generates a monotonic or “low stress” configuration based on ranked distances among sampling units (Kruskal, 1964). Unlike principal component analysis, NMDS does not assume linear relationships among variables, which in our case were species (McCune et al., 2002). Sørensen distances provide a biologically intuitive fixed maximum distance for sampling units with no species in common. The community matrix included estimates of relative abundance for all 113 species or species groups (columns) in each year (rows). We used additional approaches to describe patterns of temporal variation in bird communities and associations between communities and environmental and temporal factors. First, to test for overall differences among time periods, management, and before and after border wall expansion, we used multi-response permutation procedures (MRPP). Second, we used indicator species analysis (ISA) to identify species and species traits indicative of each time period, Indigenous vs. NPS management, and before and after border wall expansion. This approach utilizes proportional species abundances and frequencies to compare observations among periods and computes indicator values between 0 and 1 to quantify associations between species and groups, with P-values based on random permutations. Indicator values reflect the strength of association between species and each group based on the concentration of abundance and faithfulness within groups and can suggest deterministic changes in communities. Finally, to assess how variation in management, time, weather, and spring flow was associated with spatiotemporal variation in bird communities, we fit vectors onto ordination plots and computed Pearson correlation coefficients between vectors and ordination results. These vectors point in directions factors vary most in ordination space and have lengths scaled to the magnitude of correlation. To describe trends in environmental factors we used simple linear regression with year as a fixed explanatory factor.

Because variation in methods and effort across time required some assumptions linked to relative abundance calculations, we assessed how ordination results based on an alternate dataset varied to confirm our approach. We used Mantel’s asymptotic approximation method (Mantel, 1967) to compare ordinations based on the dataset of standardized relative abundance estimates, with that based on occurrence (e.g., 0/1) data. The standardized Mantel statistic (r) ranges from 0 to 1, with 1 indicating perfect correspondence between two ordinations and has a null hypothesis of no relationship between matrices. We implemented NMDS, MRBP, ISA, and Mantel tests with PC-ORD V.6.0 (McCune and Mefford, 2011).

3. Results

3.1. Effort

The final bird community dataset included records from 12 annual survey efforts over 11 different years across an 83-year time period (1939–2022). A total of 149 bird species (or species groups) and 10,159 individuals were detected during 90 days of survey effort during the breeding season; ∼60% of species were migrants and 40% were breeders or likely breeders (Appendix 1). Effort was lowest during the historical period (1939–1949) with 23 days of effort over two years, and similar during the intermediate and contemporary periods with 4–5 survey years and 28–39 survey days. After eliminating the rarest species (Table S1), we considered data on 113 species (or species groups) of which 57 were breeders or likely breeders and 56 were migrants (Table S2).

3.2. Variation in communities

NMDS ordination of bird species’ relative abundances summarized relationships among species composition, environmental variation and time (Fig. 2; Table S3). NMDS projected the distance matrix into a two-dimensional solution with low stress (0.111, final instability = 0.0000) and produced a cumulative R2 of 0.83. Axis 1 explained 46.6% of variation in the bird community matrix, and axis 2 explained another 36.4% of variation. Species composition varied among the three time periods (A = 0.16, P ≤ 0.001; MRPP) and communities in each period occupied fairly discrete areas of community space (Fig. 2). Average distance among communities was highest (0.40) in the intermediate time period (mid-1960s to mid-1900s), which included more years (31; 1965–1996), somewhat lower in the historical period (0.38), and lowest in the contemporary period (0.30) when communities where tightly clustered in ordination space indicating similar species composition. Pairwise comparison from MRPP indicated significant differences in communities between all three combinations of time periods (A ≥ 0.10, P ≤ 0.01) suggesting marked temporal shifts in communities.

Temporal variation in bird communities was characterized by increasing values along both ordination axes (Fig. 2 top and bottom), and time measured in years was fairly strongly positively correlated with both axes (r = 0.61–0.65: Table 1). Historical communities had low (negative) values along axis 1 and moderate values along axis 2, whereas contemporary communities had moderate values along axis 1 and high values along axis 2. Most species associated with the historical period were migratory shorebirds dependent on shallow water or mud flats (e.g., Western Sandpiper, Wilson’s Snipe), waterfowl and open-water birds (Green-winged Teal, Wilson’s Phalarope), birds that once bred in agricultural fields (Lark Sparrow; Huey, 1942) and openings (Killdeer; Hensley, 1954), and species that use marshes (Sora) or man-made structures (Barn Swallow). Species associated with the intermediate time period included those that once bred in tall cottonwood trees (Hooded Oriole, Bullock’s Oriole) and breeding (Bell’s Vireo), potentially breeding (Yellow-breasted Chat), and migratory (House Wren, Hermit Thrush, MacGillivray’s Warbler) species tied to dense riparian shrubs or mid-story vegetation, but many of these same species also occurred in the historical period. Species associated with the contemporary period had broader habitat affinities and included the expanding Anna’s Hummingbird, the generalist Common Raven, exotic Eurasian Collared-Dove, and some open-water birds such as Ring-necked Duck and Pied-billed Grebe. Importantly, contemporary communities occupied areas of ordination space that were much closer to that occupied by many broadly distributed species that use desertscrub. Results of indicator species analysis (ISA) among time periods highlighted some of the same species conspicuous in ordination results, with many water and open-country birds indicative of the historical period (Table S4). However, ISA highlighted some additional species of interest including four desertscrub species (Costa’s Hummingbird, Ladder-backed Woodpecker, Gilded Flicker, Cactus Wren) indicative of the contemporary period (P ≤ 0.026).

Temporal variation in bird communities was highly correlated with variation in several environmental factors (Table 1). Most conspicuously, management by Indigenous people was nearly perfectly negatively associated with temporal shifts in bird communities along axis 1 (r = −0.93). Temporal variation in communities was also highly positively correlated with variation in most temperature (T) factors (r ≥ 0.71), especially T over the prior year (r ≥ 0.74) and three years (r ≥ 0.77), and negatively correlated with spring flow (r = −0.63). All vectors representing T pointed in directions of community space occupied by contemporary bird communities, with a vector representing spring flow pointing in the opposite direction toward areas of community space typical of historical and intermediate time periods (Fig. 2, bottom). Vectors representing precipitation (P) pointed in similar directions as T but associations were limited largely to annual P over the prior three years (r = 0.51) and weak for more recent annual and seasonal P. Associations with border wall expansion were weaker (r = 0.42) but also oriented in directions similar to that for T and P.

Results of ISA for management and wall expansion highlighted a range of species indicative of each period (Table S4). Species composition varied significantly between periods of Indigenous vs. NPS management (A = 0.08, P = 0.004; MRPP) and before vs. after border wall expansion (A = 0.05, P = 0.016; MRPP), but less so than among time periods (A = 0.16, P ≤ 0.001). With regard to management, 18 of the 20 species indicative of either Indigenous or NPS management (P ≤ 0.10) matched species indicative of one of the three time periods given the close association between time and management. Exceptions were dense shrub-dependent Crissal Thrasher and marsh-dependent Yellow-headed Blackbird, both of which were indicative of Indigenous management. Fewer species were indicative of periods before and after border wall expansion (n = 10; Table S4) but included Black-throated Sparrow and Common Raven in the period following border wall expansion.

Finally, we assessed whether assumptions linked to relative abundance calculations influenced results by comparing an alternative ordination based on presence/absence data. Each ordination was highly redundant, with a standardized Mantel statistic of 0.87 (P < 0.001). Although the configuration derived from presence/absence data had lower final stress (0.063) and higher R2 (0.91), it converged on a three-dimensional solution compared to that for relative abundance, which described similar variation (0.83) in a more efficient two dimensions.

3.3. Environmental change

Temperature (T) increased markedly across the 83 year time period we considered. Mean annual T (from March of the prior year through February of the current year) increased by an average of 0.33 degrees C. (± SE = 0.05; P < 0.001) per decade with similar but less increase in Tmax (0.26 ± 0.08 degrees C./decade; P = 0.009; Fig. 3). In contrast, there was no trend for annual precipitation (P = 0.89). Spring discharge also declined markedly by an average of 0.26 ± 0.02 L/min/decade (P < 0.001) with annual averages that ranged from 0.58 to 0.61 in 2020–2022.

4. Discussion

Analyses of variation in bird communities at Quitobaquito Springs between 1939 and 2022 indicate marked faunal changes that were associated with a complex set of environmental drivers. These drivers include variation in climate, water availability, vegetation structure, and management practices that can all influence habitats and species distributions (Lindenmayer and Fischer, 2013; Flesch, 2019). Few of these changes were likely driven by a single trigger or cause. Strong associations between bird community composition and Indigenous vs. federal-agency management, temperature, and spring flow suggest marked impacts of these factors, but less effect of precipitation and border wall expansion. Although variation in survey coverage, timing, and methods likely account for some differences, patterns we observed generally match those for bird communities at larger scales across Organ Pipe Cactus National Monument (OPCNM) and in adjacent aridlands, and are consistent with the impacts of increasing aridification (Iknayan and Beissinger, 2018; Riddell et al., 2021; Roberts et al., 2021; Flesch, 2022a). Although moisture availability at Quitobaquito is higher than in the surrounding matrix, this island-like, desert oasis was not immune to the impacts of climate change and anthropogenic disturbance. To our knowledge, this study represents the first analysis of changes in faunal assemblages at Quitobaquito that attempts to quantitatively assess key factors that drive changes in communities.

In the case of Quitobaquito, apparent changes in bird communities include loss and replacement of species that depend on various elements of riparian and wetland vegetation. Specifically, this included species dependent on dense low-to mid-story vegetation and cottonwood trees typical of mesic riparian areas, shifts from species that depend on mud flats and dense emergent vegetation to those that use deeper more open-water environments, and increases in generalists and species that use upland desertscrub, which are more common in the surrounding landscape. These changes have occurred during a period of widespread drought during the 21st Century that has affected Sonoran Desert vegetation (Munson et al., 2012, 2016). Moreover, they have also occurred during periods of increasingly intensive groundwater use in the adjacent Sonoyta Valley in Mexico over the latter half of the 20th Century and associated decline of local aquifers, which has impacted water flows and riparian vegetation along the Río Sonoyta (Zamora et al., 2020). These and other factors have led to apparent local extirpation of at least twenty species of vascular plants that require soil moisture associated with springs, seeps, and canal spillover at Quitobaquito (Felger et al., 1992; Nabhan, 2003).

Historical shifts in management from Indigenous peoples to the U.S. National Park Service (NPS) had the greatest observed associations with temporal variation in bird communities. Such results are not surprising given these management changes drove marked environmental restructuring that modified habitats at Quitobaquito. Historically, under management by Indigenous peoples, Quitobaquito supported a shallow and broad expanse of open water surrounded by vegetation openings, orchards, agricultural fields, living fencerows, and shaded home sites on both sides of the international border (Bennett and Kunzmann, 1989). At that time, muddy seeps and gravelly areas of open shoreline were present and used by a greater diversity of shorebirds, wading birds, and waterfowl than in more recent times (Nabhan et al., 1982; Johnson et al., 1983). Loss of these areas and the bird species associated with them, together with losses of species that once used agricultural fields, orchards, and human settlements, which were also removed or modified by NPS in the late 1950s and early 60s, explains many observed shifts in bird communities. At this time, NPS drained a large open water area and surrounding areas of seepage and created a more defined and deeper pond that in more recent times is used by more open-water bird species that prefer deeper water. In the past, water also moved from the springs to the pond via an open ditch that supported wetland vegetation. In 1974, this ditch was replaced by a plastic pipe, but there was enough leakage at the upper end that some wetland vegetation persisted. In 1989, a new channel was dug from the springs to the pond and lined with concrete. This further reduced seepage and vegetation dominated by semi-aquatic and riparian understory plants, but the channel was still used by both aquatic and terrestrial birds. Whereas some species of wading and shorebirds are sometimes still observed at Quitobaquito, they are often limited to periods when pond levels subside and mud flats increase in area. Nonetheless, there were also lagged effects of water and vegetation management on resources given some cultivated plants like pomegranate trees persisted for three decades after changes in management.

In contrast to management changes in the more distant past, impacts of more recent management linked to expansion of the U.S.-Mexico border wall were weaker and as of yet, inconclusive. This is despite marked observed and expected impacts of these structures and associated border-related activities on wildlife and vegetation in the region (Flesch et al., 2010; McCallum et al., 2014, Peters et al., 2018), which involve extensive vegetation clearing and groundwater pumping for wall construction (Nabhan et al., 2023). Limited time since completion of these structures and complex connections and lag times between groundwater extraction and spring water discharge (Zamora et al., 2020) complicated our efforts. Hence, more study and time are needed to better understand these processes and estimate the impacts of border development on wildlife and habitats at Quitobaquito.

In arid southwestern North America, extreme weather linked to climate change is having pervasive impacts on wildlife communities (Lovich et al., 2014; Cruz-McDonnell and Wolf, 2016; Flesch 2022b) and climatic shifts are expected to intensify in future decades (Cook et al., 2015; Williams et al., 2020). In the central Sonoran Desert at Quitobaquito, we found that increasing temperatures were strongly associated with shifts in bird communities suggesting marked impacts of climate warming and associated aridification. Such changes were also closely aligned with declines in spring flows, drying of nearby springs (e.g., Aguajita, Burro, Williams) in the 1980s, and depletion of groundwater at larger scales (Zamora et al., 2020), and may be linked, especially across the nearly century-long period we considered. In contrast, precipitation had weaker associations with shifts in bird communities and unlike temperature did not seem to be declining systemically across time. Although such results suggest a smaller role of precipitation in driving faunal change, relationships may have been dampened by high levels of precipitation during recent summer monsoons. Regardless, whereas mesic environments such as Quitobaquito may buffer the impacts of precipitation on faunal communities, our results contrast with stronger observed impacts of low precipitation on birds elsewhere in the Sonoran (Flesch, 2022a, 2022b) and neighboring Mohave (Iknayan and Beissinger 2018) deserts, and suggest more time and data are needed to better understand the role of precipitation.

Mesoriparian forests and woodlands are among the most threatened and important environments for wildlife in arid southwestern North America (Knopf et al., 1988; Naiman et al., 1993; Ohmart, 1994). Although the area of riparian-influenced vegetation at Quitobaquito was always small, evidence of declines in these environments around the pond are suggested by fewer numbers of riparian shrub- and cottonwood-dependent bird species found historically, and recent shifts of bird communities in directions more typical of desertscrub species. Although some riparian and mesic-shrub dependent bird species persisted into the mid-1990s, and some may still be present and breed occasionally today (e.g., Bell’s Vireo), increases in temperature and declines in spring flows may to have eliminated many of the more riparian-dependent species. Recent occupancy by some riparian species such as Crissal Thrasher, however, suggests some auspicious patterns that additional data and analyses can elucidate in the future. Persistence of these and other species may be linked to overall increases in woody vegetation cover, even despite losses of more mesoriparian elements, which is suggested by historical photos and long-term observations. Moreover, declines in mesic riparian conditions, and presumably bird species associated with these environments, in the surrounding landscape along the adjacent Río Sonoyta has been drastic since the 1980s (PH, pers. observation), and have also impacted communities at Quitobaquito.

With loss and degradation of mesic riparian forests and woodlands, uplands environments and associated wildlife often expand and homogenize faunal communities. We found that contemporary bird communities at Quitobaquito occupied areas of community space into which many more common desertscrub species are found such as Cactus Wren and Ladder-backed Woodpecker. These results further suggest increasing aridification and associated reductions in riparian vegetation, but may also have been influenced by greater survey coverage of areas further from the pond in recent years. Finally, bird species associated with more recent times also included Anna’s Hummingbird that is expanding its range in the region (Greig et al., 2017), exotic Eurasian Collared-Dove that only recently expanded into the region (Flesch, 2008), and generalist Common Raven that was the only species found to have expanded in the adjacent Mohave Desert in the last century (Iknayan and Beissinger 2018).

Despite challenges in attributing the relative impacts of different stressors and potential drivers of faunal change at Quitobaquito, this small habitat complex is highly significant for regional biodiversity. This complexity is further underscored by shifting management practices from Indigenous O’odham farmers and foragers, to science-based approaches by NPS, complex connections and time lags between current spring flow and local and regional groundwater exploitation, and more recently to construction of security infrastructure that has militarized the border at Quitobaquito. Regardless, our results provide some useful insights that we believe are consistent with the principle elucidated by De Grenade and Stevens (2019): “Desert oases are among the most threatened ecosystems on earth … Given the small land area that these spring-fed ecosystems occupy, their importance as quality habitat, migratory rest-stops, biodiversity, and economic value in arid regions is largely underestimated”.

Several recent modifications and management changes at Quitobaquito occurred shortly after this study, and will continue to alter habitats for wildlife at and around the habitat complex. In spring 2022, two shallow refuge ponds were created on the north side of the main pond to hold pupfish and mud turtles while the main pond was drained, re-contoured, and fitted with a non-permeable liner. The new pond has a gentle slope to provide shallow water environments, which should persist so long as encroachment by rushes and other aquatic vegetation is managed. Dense stands of rushes and woody shrubs around the pond were removed and one of two remaining Fremont cottonwood (Populus fremontii) and one of two remaining Goodding willow (Salix gooddingii) trees died. The new pond liner was intended to keep the pond full with inputs ≥15 L/min, but as of August 2023, this has not happened and either water budget calculations were incorrect or there is a breach in the new liner. Re-vegetation of the perimeter of the pond is on hold while NPS seeks a solution. NPS plans to keep the new refuge ponds as additional shallow-water environments and to trap sediment during floods. NPS has struggled to manage vegetation encroaching into aquatic environments for decades but will continue to do so.

Additional future management focused on 1) planting young cottonwood and willow trees to buffer losses of the few remaining broadleaf trees, 2) protecting trees that reproduce naturally, 3) passive rainwater harvesting structures, and 4) directing seepage or overbank water flows in ways that enhance vegetation cover, can help maintain and bolster riparian conditions for wildlife at Quitobaquito. Future survey data combined with that amalgamated here will continue to elucidate these changes and the implications for management and conservation.


W.K. Kellogg Foundation provided support for GPN and ADF, with additional support for ADF from and the U.S. National Park Service’s Southwest Border Resources Protection Program (Cooperative Agreement No. P20AC00299). PH was supported by NPS and has served as the ecologist for OPCNM since 2003.

CRediT authorship contribution statement

GPN envisaged the study, assisted with initial data assembly and writing, and secured support for ADF. ADF assembled the data with assistance of GPN and PH, undertook the analysis, and wrote portions of the manuscript. PH assisted with data assembly, provided data on spring discharge and birds, and insights. All authors read and contributed to revisions of the manuscript.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

Aaron Flesch reports financial support was provided by National Park Service. Aaron Flesch and Gary Nabhan reports financial support was provided by WK Kellogg Foundation.


We thank R.R. Johnson, L. Stevens, K. Reichhardt, A. Rea, and E. Mellink for help gathering historical data. K. Muddle, S. Rossi, and T. Wilson of NPS provided important support for bird monitoring in Organ Pipe Cactus National Monument. NPS Archivist K. Saba helped us locate historical photos, S. M. Russell provided unpublished bird data, and L. Eiler, Hia c-ed O’odham elder and former Tohono O’odham tribal councilmember provided important support.


Journal of Arid Environments

Related Posts