Avian Conservation and Management

INTRODUCTION

Cassin’s Sparrow’s environmental preferences

Cassin’s Sparrow (Peucaea cassinii, Woodhouse 1852) is a small, gray, ground-dwelling endemic of the arid grasslands of the southwestern United States (U.S.) and northern Mexico (Woodhouse 1852, Williams and LeSassier 1968, Dunning et al. 2020). It is migratory, constructs nests near the ground, and is notable for its secretive nature and highly developed flight song. Studies have shown that Cassin’s Sparrow’s behavioral ecology is closely linked to wide-spread, seasonal weather patterns as well as highly localized, ground-level climatic conditions (Williams and LeSassier 1968, Schnase et al. 1991, Ruth 2000, Dunning et al. 2020). Regional precipitation appears to be of particular importance at the macroclimatic scale, presumably due to its influence on vegetation availability and insect abundance. In fact, field studies over the past century suggest that Cassin’s Sparrow is an itinerant breeder, so responsive to regional rainfall patterns that birds make seasonal, inter-clutch moves within their range to find optimal conditions for breeding (Williams and LeSassier 1968, Ohmart 1969, Dunning et al. 2020). At the microclimatic scale, wind speed and ambient temperature are important influences. In a behavior that is uncommon among North American birds, male Cassin’s Sparrows defend territories and secure mates using aerial flight song displays. Elevated winds and temperatures reduce the birds’ capacity to perform this energetically demanding skylarking behavior, which is an essential element of the species’ breeding biology (Schnase and Maxwell 1989, Schnase et al. 1991).

Although much of what is reported about the natural history of Cassin’s Sparrow is poorly established, Cassin’s Sparrow’s apparent responsiveness to environmental conditions raises a concern that the species could be among the many North American grassland birds that are vulnerable to climate change (Schnase et al. 1991, Ruth 2000, Wilsey et al. 2019, Dunning et al. 2020). Accurate range-wide assessments, however, are often difficult with these species (Hubbard 1974, Huntley et al. 2010, Reside et al. 2010, Heenan and Seymour 2012, Sohl 2014, Lipschutz 2016, Norman and Christidis 2016, Iknayan and Beissinger 2018, Rosenberg et al. 2019). With Cassin’s Sparrow, for example, numerous studies over the past half-century have painted an inconsistent picture. Some studies find evidence for a retraction of viable habitat and declining populations across North America (Lynn 2006; North American Breeding Bird Survey, https://www.pwrc.usgs.gov/bbs/), whereas others find mixed results and too little data to establish with confidence an overall status (Ruth 2000). Many sources identify the species as stable and of low conservation concern (Rosenberg et al. 2016, Dunning et al. 2020; National Audubon Society, https://www.audubon.org; NatureServe, https://explorer.natureserve.org). State-level studies are often no more definitive. The State Wildlife Action Plan for New Mexico, for example, lists Cassin’s Sparrow as a declining species, susceptible to shifting environmental conditions that could lead to rapid population changes (New Mexico Department of Game and Fish 2016). At the same time, of nine grassland birds in New Mexico, recent work has shown Cassin’s Sparrow to be the only species for which gains in suitable habitat are projected over the next 50 years (Salas et al. 2017).

These differing views of Cassin’s Sparrow’s status suggest that a fundamental aspect of the species’ relationship to the environment is not well understood. From first principles, we know that the multidimensional niche space of a species may shift, through adaptation or acclimation, as may its demography in response to climate changes (Maguire 1973). We also know that climate change can drive shifts in the spatial distribution of many bird species as they track suitable conditions (Pecl et al. 2017). In addition, recent work underscores the importance of non-constant processes, such as long-term trends, seasonal or temporal cycles, or random variation, which can cause climatic drivers to vary in complex ways over time and space, a phenomenon referred to as non-stationarity (Rollinson et al. 2021, Ward et al. 2022). It is generally agreed that the most important contemporary driver of non-stationary environmental trends is global climate change (Ward et al. 2022).

Past response to a changing climate can offer insights

Cassin’s Sparrow’s past response to a changing climate could help explain the variability we see in the species’ status assessments as well as the underlying biological mechanisms for this response. This is especially true if macro- and microclimatic patterns vary across Cassin’s Sparrow’s range and the variability among these assessments arises from the species’ response to non-constant trends in key environmental drivers. Ecological niche modeling (ENM) is a set of techniques and tools that is often used in this type of research. ENM uses species occurrence records and environmental data to estimate the probabilities of suitable habitats across a study area (Peterson et al. 2011). Although ENM is commonly used to project future responses of a species to climatic change, less consideration has been given to historical responses to a changing climate that could shed light on core biological questions or the present and future status of a species (Johnston et al. 2020). In this study, we combine data from NASA’s Modern-Era Reanalysis for Research and Applications, Version 2 (MERRA-2; M2) with field observations spanning the past 40 years to perform a retrospective ENM analysis of Cassin’s Sparrow’s evolving climatic niche to help answer three questions that bear on this issue:

1. How has Cassin’s Sparrow responded to changing climatic conditions in the past?
2. What are the specific attributes of climatic change that the species is responding to?
3. Could Cassin’s Sparrow’s response to changing climatic conditions account for the variability seen in assessments of the species’ conservation status?

To address the first question, we use changing historical patterns of climatic suitability as an indirect indicator of the species’ response to changing conditions. For the second question, we use changes in the types of predictors of climatic suitability over time to identify the key attributes driving Cassin’s Sparrow’s biological response. To gain a perspective on the third question, we look at changing patterns in the values of key predictors over time to gauge their potential as a source of variability.

Retrospective ENM’s three-element approach

To address these questions, we have taken an approach that embodies a novel integration of three elements. First, we base our analysis on environmental variables obtained solely from the M2 reanalysis. Climate reanalyses combine past observations with numerical models to generate a consistent time series of hundreds of fundamental, physical drivers of the Earth system. They offer a comprehensive description of Earth’s observed climate as it has evolved over the past half century at a fine temporal scale (Edwards 2010). The current study spans the 40-year period from 1980 to 2019 and is based on two climatic variable time series. In one time series, we use 30 M2 variables selected to reflect key attributes of the microclimate that are known to relate directly or indirectly to the biological or ecological functioning of a species. This collection of predictors is intended to help us understand the influence of climate-related conditions in localized areas near the Earth’s surface and includes variables, such as temperature near and on the ground, humidity, wind speed, cloud cover, soil moisture, evapotranspiration, vegetation coverage, and the incoming and reflected solar radiation that provides energy to drive ecosystem functioning at the micro scale (Schnase et al. 1991, Cabello et al. 2012, Bosilovich et al. 2016, Gelaro et al. 2017, Arenas-Castro et al. 2018, Pettorelli et al. 2018, Regos et al. 2022a, 2022b). In the second time series, we use 19 M2-derived bioclimatic variables modeled after the classic bioclim predictors commonly used in ENM (O’Donnell and Ignizio 2012). This collection of predictors is intended to help us understand the influence of wide-spread, yearly, seasonal, and monthly temperature and precipitation patterns. Together, these two sets of predictors provide a more detailed view of macro- and microclimatic factors influencing environmental suitability for Cassin’s Sparrow than could be realized by using either set of predictors alone.

Second, we employ time-specific ENM in our analysis: our dependent and independent variables are temporally aligned across the time span of the study. Detailed information about the geographic distribution and changing dynamics of climatic suitability for a species is critical to conservation planning (Guisan et al. 2013, Porzig et al. 2014, Gonçalves et al. 2016, Rollinson et al. 2021). ENM is often applied within a time-averaged framework in which the values of environmental variables are averaged over time spans that are not in temporal registration with the occurrence records upon which models are calibrated or tested (Bede‐Fazekas and Somodi 2020, Ingenloff and Peterson 2021). Although useful for exploring species distributions at a broad level, modeling within a time-averaged framework can elide complex effects of the environment on an organism, especially highly mobile or behaviorally complex species (Roubicek et al. 2010, Ingenloff 2020). Of particular concern to conservation work, studies have shown that temporal mismatches in the time period spanned by occurrence data and the climate baseline can decrease the utility and accuracy of ENM products (Roubicek et al. 2010, Goberville et al. 2015, Guida et al. 2019, Pérez‐Navarro et al. 2021). In this work, we use five-year averaged values for our environmental variables across a sequence of eight time intervals spanning the 40-year period of the study. We then use time-specific species observations corresponding to these five-year intervals as dependent variables in the models that form the basis of our analyses, thereby enabling a temporally explicit view into the major, range-wide, historical patterns of changing climatic suitability for Cassin’s Sparrow.

Finally, we employ variable-specific ENM in our analysis: a tailored set of independent variables is used for each of the five-year intervals in the time series. ENM generally uses a fixed set of variables in a given study. These variables are typically selected through a manual process based on an ecological understanding of the species being studied or one of a variety of statistical approaches (Petitpierre et al. 2017). However, retrospective ENM using reanalysis data presents an opportunity to examine changes in the drivers of climatic suitability themselves over time. In this study, we use NASA’s MERRA/Max system to perform automatic variable screening within each of the five-year time intervals of the analysis. MERRA/Max provides a scalable feature selection approach that enables direct use of global climate model (GCM) outputs in ENM (Schnase and Carroll 2022). The system accomplishes this selection through a Monte Carlo optimization that screens a collection of variables for potential predictors of suitable conditions (Schnase et al. 2021). With MERRA/Max, variable selection is guided by the indirect biological influences injected into the algorithm’s selection process by the species occurrence files, identifying biologically and ecologically plausible predictors in large, multidimensional data sets where selection through ecological reasoning or other means is not feasible (Searcy and Shaffer 2016, Smith and Santos 2020). Variable selection for each five-year time interval, as applied here, enables a view into the changing patterns of environmental determinants of climatic suitability that would otherwise be difficult, if not impossible, to obtain. That view, in turn, provides a way of examining the possible underlying biological basis for Cassin’s Sparrow’s response to a changing climate.

Collectively, these three aspects of the study offer a more detailed look at Cassin’s Sparrow’s relationship to the environment than has been previously reported. In the sections that follow, we describe our method and results, discuss what we see as the important takeaway lessons from the study, and conclude with recommendations for next steps.

METHODS

Modeling environment and study approach

We framed the study as an analysis of broad-scale, multi-decadal changes in the climatic suitability for Cassin’s Sparrow. To that end, we did not focus on changes in the species’ range, population size, or abundance, and, in the current work, we do not look at higher temporal resolution or inter-annual trends in climatic suitability, which would have added significantly to the computational burden of the analysis. We used MaxEnt for our modeling environment (Phillips et al. 2006, Elith et al. 2011, Phillips et al. 2017). Based on a machine learning approach to maximum entropy modeling, MaxEnt calculates the distribution of suitable habitats using species occurrence data and a set of environmental variables (Phillips et al. 2006, 2017, Elith et al. 2011, Merow et al. 2013, Kalinski 2019). The system uses presence data only, comparing the locations of where a species has been found to environmental data from across the study region. MaxEnt defines the overall environment by sampling a large number of background points throughout the region, including locations where the species is known to occur as well as locations where the species’ presence or absence is unknown. Background points are intended to be so numerous that they capture the full environmental space of the study area and, importantly, so outnumber the occurrence locations that favorable conditions for a species can be mathematical distinguished from the overall environmental context. MaxEnt produces accurate predictions with even small sample sizes and often produces better results when the number of presence locations is reduced by statistical thinning or filtered in an ecologically meaningful way (Pearson et al. 2007, Boria et al. 2014, Aiello-Lammens et al. 2015, Phillips et al. 2017).

Preparing species occurrence data

We used annual, time-stamped, point locations for Cassin’s Sparrow occurrences across the species’ full North American range, making no distinction between breeding and non-breeding season observations. We obtained occurrence data from the Global Biodiversity Information Facility (GBIF, https://www.gbif.org), an international information infrastructure funded by the world’s governments that provides open access to data about all types of life on Earth. Occurrence datasets make up the core of what is published by GBIF. GBIF’s standardized formatting, documentation, and quality requirements are enforced through close collaboration with data providers. We downloaded a total of 32,518 records for the years 1980 to 2019 (GBIF occurrence download, https://doi.org/10.15468/dl.x33grq [15 January 2022]). More than 95% of the records were originally sourced from the eBird citizen-scientist observational dataset (https://ebird.org/species/casspa), but the download also included research-grade records from iNaturalist (https://www.inaturalist.org), the National Ecological Observational Network’s breeding land bird point count collection (https://www.neonscience.org), and museum specimen records from the National Museum of Natural History (https://naturalhistory.si.edu), American Museum of Natural History (https://www.amnh.org), Harvard University’s Museum of Comparative Zoology (https://mcz.harvard.edu), Kansas University Biodiversity Institute and Natural History Museum (https://biodiversity.ku.edu), and the University of Arizona Museum of Natural History (https://www.arizonamuseumofnaturalhistory.org).

There were relatively few occurrence records in the early years of the study’s time span. We found that binning records into five-year intervals gave us enough observations to produce useful models across the long time period of the study. We therefore merged the GBIF observations into a time-series comprising eight, five-year aggregated collections: 1980–1984, 1985–1989, 1990–1994, 1995–1999, 2000–2004, 2005–2009, 2010–2014, and 2015–2019 (Fig. 1A). These five-year collections initially ranged in size from 263 records in the 1980 group to over 14,000 records in the 2015 collection. After removing replicates, we thinned the records to non-overlapping observations within a 16 km (~10 mile) buffer around each point to avoid double counting the same individuals. For count uniformity across the series and to reduce record-densities that can diminish model performance, we extracted random 250-record samples for each five-year span in the time series (Boria et al. 2014). In trials using various sample sizes, we found that 250 records per five-year interval yielded the most consistently accurate models across the 40-year time series.

Preparing environmental variables

We created two sets of environmental variables for the analysis. We began by defining a study area that encompasses Cassin’s Sparrow’s summer and winter ranges across the Continental U.S. using information from the U.S. Geological Survey (USGS) National Gap Analysis Program (GAP) (https://doi.org/10.5066/F7Q81B3R). We selected a study area that extended from latitude 24.8° N to 44.0° N and longitude 93.5° W to 115.6° W. We then obtained a base collection of maximum, minimum, and mean values of 30 gridded M2 variables across our spatial and temporal domain, natively available from NASA’s Goddard Earth Sciences Data and Information Services Center (https://disc.gsfc.nasa.gov; Table 1, Fig. 1A). This collection contained the precursor temperature and precipitation variables from which the 19 classic, bioclim predictors used in ENM are derived (O’Donnell and Ignizio 2012). In addition, it included environmental attributes of more direct biological significance that are not explicitly represented in the 19 traditional bioclimatic variables, such as soil moisture and evaporation from land, wind direction and speed, and various solar radiation fluxes. We aggregated these data to our five-year time intervals to create working set collections of the five-year averaged mean values for each variable, which we then formatted for use by MaxEnt. To smooth the representation of local environmental conditions, we resampled the M2 layers from their native spatial resolution of 1/2° latitude × 5/8° longitude to 5.0 arc-min (1/12°) resolution (~7.6 km at latitude 35.0° N) using bilinear interpolation. We also built working set collections of M2-derived bioclimatic variables, one set per five-year interval, which we refer to as the MERRAclim-2 (MC) working set collection (Table 2). These were modeled after Worldclim’s 19 bioclim variables; however, because the classic Worldclim collection is based on 30-year averaged values, we created an M2-based version to accommodate the time-specificity requirement of the study. For this, we used the R dismo library following the method of Vega et al. (Vega et al. 2017, Hijmans et al. 2023). We employed M2’s monthly maximum and minimum temperature values (T2M) and the averaged values for monthly precipitation (PRECTOTCORR) to create the MC working set collections, which conveyed the added consistency of MC’s 19 variables being derived from the same data sources as M2.

Building the time series

We built two MaxEnt climatic suitability time series using the aggregated GBIF occurrence collections. In one, we used the M2 working sets as independent variables; in the other, we used the MC working sets as independent variables. A two-step processing workflow was applied to each five-year interval of each time series (Fig. 1B).

Selecting and refining time-specific variables

First, we used MERRA/Max to automatically select top contributing M2 and MC variables in each of the eight, five-year spans of the 40-year series. We performed three screening runs, each returning a set of top ten selected variables using MERRA/Max’s standard screening configuration and a per-variable sampling rate of 100 (Schnase and Carroll 2022). We averaged the results of the three runs to create a selection set of the top ten predictors for each five-year interval. We then used variance inflation factor (VIF) analysis to reduce collinearities in the selected predictors (Pradhan 2016). This resulted in a final, model set of top predictors for each five-year interval in the two time series that had few if any collinearity issues.

Calibrating and building time-specific models

Next, we built calibrated models for each set of occurrences and predictors in each of the five-year intervals in the M2 and MC time series. We used the R ENMeval package to identify optimal MaxEnt parameter settings for each interval (Muscarella et al. 2014, Kass et al. 2023). ENM automatically creates a collection of models across a range of settings and uses Akaike’s Information Criterion corrected for small sample size (AICc; Akaike 1974) to measure information loss among the models. The combination of MaxEnt parameter settings that results in the lowest AICc value is taken to be an optimal tuning configuration for a given set of inputs. We used these optimal settings to construct a final model for each five-year interval in the two time series. For each of the ENMeval calibration runs, and in the final MaxEnt model run, we used 10,000 background locations randomly selected from across the study area and performed a 10-fold cross validation in which 70% of the occurrences were selected for training and 30% for testing in each repetition (Phillips et al. 2017). Cloglog format was used throughout. This model calibration and construction process was performed in triplicate for each five-year interval to produce the final models that we used in the subsequent time series analysis.

Analyzing time series results

We analyzed our time series results by creating trend maps showing regions of positive and negative change in climatic suitability per five-year interval across the study area over the 40-year span of the study (Fig. 1C). We used the Theil-Sen median slope estimate to build these trend maps. Theil-Sen provides a non-parametric means of robustly fitting a line to a set of points by finding the median slope of all the lines through all pairs of points in the set. In our case, the estimate was applied for each pixel across the eight raster layers comprising the final models of climatic suitability for each 40-year time series. There were, thus, eight points, one per five-year interval, in the slope calculation for each pixel in the resulting trend maps. We used Mann-Kendall Z values to determine the statistical significance of the resulting trends (Theil 1950, Sen 1968).

We then quantified the overall direction and speed of shifts in climatic suitability observed in the trend maps. We used the movement of weighted centroids to characterize these shifts. Weighted centroids represent the geometric center of all pixels in the studied area, weighted by their suitability index (VanDerWal et al. 2013). We chose weighted centroids for this purpose because of their ability to account for information derived from across a species’ entire modeled range. We quantified movement by measuring length and direction of the vector between weighted centroids of the 1980 and 2015 probability maps.

Finally, we examined trends in the values of the most important variables across the time series. Top contributing variables were those that appeared in half or more of the 24 final models in the triplicated time series; these variables were ranked by their average permutation importance across the three runs (Searcy and Shaffer 2016). For the top, study-wide contributing variables, we computed Theil-Sen slopes to visualize patterns of change in the values of the M2 and MC variables across the study area. We then examined trends in the relative contributions (i.e., permutation importance) of each of these top variables per five-year interval across the 40-year span of the study.

The final models in both time series were qualitatively judged for reasonableness on the basis of first-hand knowledge of the species and its environmental preferences and what can be inferred about climatic contexts from published estimates of Cassin’s Sparrow’s North American range. We used the Area Under the Receiver Operating Characteristic (ROC) Curve (AUC; Fielding and Bell 1997), the True Skill Statistic (TSS; Allouche et al. 2006), and Percent Correctly Classified (PCC; Warren and Seifert 2011) as measures of model accuracy.

RESULTS

Evaluation of model performance

Our models exhibited moderate performance across the 24 final models in each triplicated time series according to their mean AUC, PCC, and TSS values (Table 3). In evaluating model performance, we regarded AUC measures < 0.7 as low, 0.7–0.9 as moderate, and > 0.9 as high; similarly, we regarded TSS and PCC measures < 0.5 as poor, 0.5–0.8 as useful, and > 0.8 as good (Arenas-Castro et al. 2018). The AICc estimate of information quality in the final models within and across the two time series was highly consistent. Against these metrics, we judged the observed level of performance to be a sufficient basis for examining the broad trend patterns displayed by the time series models, which was the primary focus of the study. The observed spatial distribution of suitable conditions corresponded well with what is known about the natural history of the species and published range maps.

Evaluation of suitability trends

Overall trends

The study area encompassed approximately 3.91 x 10⁶ km². Broad areas of climatic suitability change across the study area were observed in both time series over the past 40 years. The most favorable climatic conditions for Cassin’s Sparrow have generally concentrated in the southeastern regions of the study area according to both the M2 and MC time series; however, over time, a northwesterly shift in areas of high suitability is apparent in both series (Figs. 2A, 3A). The Theil-Sen trend maps also show a northwesterly movement in increasing climatic suitability over the past 40 years, although the patterns of change are more broadly diffuse in the M2 series, as described in greater detail below (Figs. 2B, 3B). An overall increase in climatic suitability was identified over approximately 54.9% of the study area (~21.49 x 10⁵ km²) in the M2 time series and 69.5% of the study area (~27.21 x 10⁵ km²) in the MC series (Table 4). Areas showing an overall decrease in climatic suitability ranged in size from approximately 40.1% (~15.72 x 10⁵ km²) in the M2 time series to 24.4% of the study area (~9.54 x 10⁵ km²) in the MC series.

Statistically significant trends

Areas of statistically significant change at the 95% confidence level are relatively small, and the rates of positive and negative change in those regions are low (Table 4). Theil-Sen results in the M2 series indicate that the estimated probabilities of climatic suitability have been increasing by an average of 0.03 every five years for the past 40 years across 6.2% of the study area, with a corresponding average decrease in estimated probabilities of 0.05 every five years over 0.5% of the region. Statistically significant positive trends were concentrated in the central, northwestern, and far southwestern regions of the study area, whereas statistically significant negative trends concentrated in the southeast and central southwest, imparting a west to northwesterly axis to these positive shifts that is consistent with the weighted centroid analysis, as described below.

Theil-Sen results for the MC time series paint a similar picture. Probability estimates of climatic suitability, within areas of statistically significant change, have been increasing by an average of 0.03 every five years for the past 40 years across 5.5% of the study area, with an accompanying average decrease in estimated probabilities of 0.06 every five years over 0.9% of the region (Table 4). Statistically significant positive trends concentrated in three clusters, one each in the northwest and northeast, the third in a central southwest region, which, by contrast, showed a negative trend in the M2 analysis. However, there was a statistically significant negative trend in the southeast region for the MC series that coincides with the pattern observed for the M2 series.

Patterns of climatic suitability shift

Overall shifts in the climatic suitability for Cassin’s Sparrow were observed in both time series. The 40-year displacement of the weighted centroid for suitability in the M2 time series was approximately 104 km and had a westerly trajectory (282°). This pattern contrasted with that seen in the MC time series, which showed a 40-year displacement of 75 km along a northwesterly trajectory (309°; Fig. 4). Centroid movement in the M2 time series shows a higher change velocity than that seen in the MC time series (2.60 km/yr vs. 1.85 km/yr, respectively).

Evaluation of variable trends

Trends in the MERRA-2 (M2) predictors

Seven M2 variables appeared in more than half of the models in the M2 time series (Table 5). Each variable demonstrated complex patterns of non-constant spatial and temporal trends across the study area (Fig. 5), making it difficult to generalize about increasing or decreasing conditions. For example, Theil-Sen analysis showed generally positive trends across the study area for SPEED, QV2M, PREVTOT, and SWGNT, with positive change centrally located in the study area for SWGNT, concentrated in northern, northeastern, and southeastern regions for QV2M, and broadly scattered in SPEED and PREVTOT. Predominantly negative trends were seen in EVPSOIL and EVPINTR, with EVPSOIL’s negative changes concentrated in the southwest and EVPINTR’s broadly distributed across the entire study area. Nearly equivalent areas of overall positive and negative change were observed for TAUTOT (Table 6), with positive change concentrated in the north and southeast. The M2 trends generally lacked statistical significance.

Trends in the MERRAclim-2 (MC) predictors

Seven MC variables also appeared in more than half of the models in the MC time series (Table 5). These variables demonstrated complex patterns of changing spatial and temporal trends across the study area as well (Fig. 5). Theil-Sen analysis showed predominantly positive trends across the study area for MC_Bio08, MC_Bio15, and MC_Bio05, with large expanses of positive change concentrated centrally and in the northwest region of the study area for MC_Bio05, concentrated in the southwest for MC_Bio08, the west and southwest for MC_Bio15. Negative trends dominated in MC_Bio14 and MC_Bio18, especially in the western half of the study area. Nearly equivalent areas of positive and negative change were seen in MC_Bio03 and MC_Bio13 (Table 6), with broadly scattered patches of change apparent for each variable throughout the study area. Most of the trends observed in the MC variables also lacked statistical significance.

Variable contribution trends across the time series

The relative importance of the top contributing variables also varied across the 40-year span of the study (Fig. 6). In the M2 time series, we observed a generally increasing trend in the contributions of EVPINTR to each five-year interval’s final model and a generally decreasing trend in the contributions of QV2M and SWGNT. SPEED and EVPSOIL were consistently high contributors; TAUTOT and PREVTOT were consistently moderate contributors. In the MC series, MC_Bio14 and MC_Bio03 show generally increasing trends; MC_Bio08 showed a sharply decreasing trend. MC_Bio18 was a consistently high contributor, with MC_Bio15, MC_Bio05, and MC_Bio13 making consistent contributions at moderate to low levels.

DISCUSSION

Complex 40-year patterns of changing climatic suitability

Velocity of change consistent with other findings

The shifts in favorable conditions we observed across both time series are consistent with what has been reported for Cassin’s Sparrow and many other North American grassland bird species (Bateman et al. 2016, Huang et al. 2023). Bateman et al. (2016), for example, found an average bioclimatic velocity of 1.27 km/yr to the west, northwest, and north over the past 60 years in the potential breeding distributions of 285 species of land birds across the Continental U.S., with some potential breeding populations shifting at rates of up to 5.51 km/yr. Likewise, a recent study by Huang et al. (2023) reported a mean bioclimatic velocity of 2.25 km/yr in 29 species of grassland birds along predominantly east-west axes of increasing environmental suitability, accompanied by an estimated, mean abundance-based velocity of 5.02 km/yr along more northerly inclined axes of increasing abundance.

Trends and patterns of change vary across time series

Our two time series reveal contrasting histories of climatic suitability change over the past 40 years. The MC time series paints a more favorable picture of changing conditions than what is seen in the M2 time series, with areas of overall improving conditions exceeding areas of declining conditions by 45.1% in the former and only 14.8% in the latter (Table 4). This contrast in overall trend is particularly apparent across the western regions of the study area and within the USGS GAP boundaries of Cassin’s Sparrow’s northernmost breeding range (Figs. 2B, 3B). In the west, M2-driven model results portray a northerly shift in improving conditions across the full extent of the study area. With the MC-driven models, changes are not nearly as distinct nor are they as extensive. The M2-driven model results show sharply improving climatic suitability in the northeastern extent of the breeding range and a fall-off along the western boundaries of Cassin’s Sparrow’s breeding and non-breeding ranges, whereas the MC-driven model results show improving climatic suitability in the northeastern region and a fall-off along the western boundary of only Cassin’s Sparrow’s non-breeding range.

State-level scale highlights differing trends and patterns

The differences between the two time series become more noticeable when results are considered at the state level. Cassin’s Sparrow’s breeding and non-breeding ranges comprise seven states within the Continental U.S.: Arizona (AZ), Colorado (CO), Kansas (KS), Nebraska (NE), New Mexico (NM), Oklahoma (OK), and Texas (TX) (Figs. 2B, 3B). Cassin’s Sparrow is found in four of these states only during the breeding season (i.e., CO, KS, NE, and OK). With KS and OK, we see similar trend patterns across the two time series. On the other hand, for CO, we see a sharp decline in suitability in the M2 time series, especially in the central part of the state, and a general improvement in conditions in the MC series. We see the inverse pattern for NE across the two series, with the decline in the MC series in the north-central part of the state.

Within the U.S., Cassin’s Sparrow has historically been most abundant in AZ, NM, and TX (North American breeding bird survey, https://www.pwrc.usgs.gov/bbs/). These states represent the heart of Cassin’s Sparrow’s U.S. range, and it is here we see the most striking differences in the M2 and MC time series’ model results. In the M2 series, climatic suitability in areas that historically accommodated seasonal range expansion between non-breeding and breeding seasons is shown to have been declining over the past 40 years in AZ and NM. In contrast, climatic suitability in these areas appears to be improving or largely unchanged in the MC series. In TX, M2 model results indicate improving breeding season conditions in northern regions and decreasing suitability along the state’s southern border in breeding and non-breeding areas; similar but less pronounced patterns are observed in the MC time series.

Implications for conservation practice

The variability we see across the two time series and across the species’ geographic range has practical implications for conservation policy development and implementation. Population size, species abundance, and similar attributes are a central component of essentially all state-level conservation status formulation processes (Association of Fish and Wildlife Agencies State Wildlife Actions Plans, https://www.fishwildlife.org). It has been shown that the combination of bioclimatic predictors with other environmental attributes, such as ecosystem functional attributes, edaphic variables, topographic data, vegetation indices, and microclimatic variables, can capture factors affecting abundance rather than just occurrence, thereby yielding suitability model results that are often highly correlated with abundance (Cabello et al. 2012, Weber et al. 2017, Leitão and Santos 2019, Regos et al. 2020, Arenas-Castro and Sillero 2021, Cavalcante et al. 2022). In our results, we see that patterns of historical change in environmental suitability can vary widely depending on the spatial extent considered and on whether time series models are driven by bioclimatic variables alone or by variables more aligned with ecosystem functioning. It is possible that our results are an indicator that demographic and other observational survey results may vary greatly across the species’ range as well, which could lead to differing conclusions regarding the conservation status of Cassin’s Sparrow and possibly account for the inconsistencies we observe in the literature and among states regarding this species’ status.

Complex 40-year patterns of changing environmental drivers

Drought and seasonal precipitation are important macroclimatic drivers

The MC time series provides insights into how climatic changes have influenced seasonal trends that are important to Cassin’s Sparrow (Table 5). The three most contributory variables in the MC time series are mean temperature of the wettest quarter (MC_Bio08), which appears to have been generally increasing across Cassin’s Sparrow’s range over the past 40 years while declining in importance as an environmental driver; total precipitation of the driest month (MC_Bio14), which has been decreasing across Cassin’s Sparrow’s range while increasing in importance as an environmental driver over the same period of time; and precipitation of the warmest quarter (MC_Bio18), which has also been generally decreasing across the species’ range but displays a discontinuous patchwork of positive and negative trends while remaining one of the top environmental drivers over the 40-year span of the study. Collectively, these three variables appear to confirm that the long-running drought conditions across southwestern North America have had a major influence on climatic suitability for Cassin’s Sparrow (Stahle 2020, Williams et al. 2020). That MC_Bio18 is consistently the most important variable across the time series is notable, given that this variable essentially characterizes the seasonal precipitation pattern of the North American Monsoon (Adams and Comrie 1997, Becker 2021). The next two most contributory variables are isothermality (MC_Bio03) and precipitation seasonality (MC_Bio15). The consistent contribution of these two variables to the MC time series suggests that variability in both temperature levels and precipitation amounts has also had a historical influence on climatic suitability for Cassin’s Sparrow.

Wind speed and drying soils are important microclimatic drivers

The M2 time series provides insights into how trends in ground-level factors have influenced environmental suitability for Cassin’s Sparrow (Table 5). The two most important variables in the M2 time series are surface wind speeds (SPEED), which appears to have been generally increasing across the study area over the past 40 years while remaining a consistently top model contributor, and bare soil evaporation energy flux (EVPSOIL), which is generally decreasing across the study area over this period, especially in the southwestern region, while also remaining a top contributor across the time series. Wind has a mixing effect on air near the ground that can increase the evaporation of water from soil and plant surfaces until a point is reached where drying conditions take over (Tran et al. 2016, Lian et al. 2022). Further evidence of this phenomenon is seen in the water loss from plant surfaces, as reflected by interception loss energy flux (EVPINTR), the fourth most contributory variable, which has been decreasing over the past 40 years while its importance as a driver in our models has steadily increased. Total re-evaporation of precipitation (PREVTOT), which can contribute to reduced precipitation efficiency (i.e., the amount of falling precipitation that evaporates or sublimates before reaching the ground), is the sixth most contributory variable. In our results, there has been a generally increasing trend in PREVTOT across much of the study area while the driver itself has remained moderately consistent in importance across the 40-year span of the study. These factors, along with increasing shortwave radiant energy from the sun (SWGNT) and decreasing cloud thickness (TAUTOT), the seventh and fifth most important variables, respectively, appear to be contributing to decreasing water content of the atmosphere (QV2M) across the North American Southwest. Our results indicate that QV2M itself, the third most important contributory variable overall, has decreased in importance as an environmental driver over the years as SPEED, EVPSOIL, and EVPINTR have become the dominant influences. Taken together, however, the higher surface wind speeds, reduced bare soil evaporation, drying air, and lower precipitation efficiency reflected in the M2 time series’ top four variables suggest that microclimatic drying may be the particular aspect of varying monsoon precipitation of importance to Cassin’s Sparrow (Tran et al. 2016, Lehmann et al. 2018, Cheng et al. 2021, Lian et al. 2022, Peevey 2022).

Implications for future status assessments

The observed connection between top contributing variables, monsoon rainfall, and ground-level drying conditions is consistent with what is known about Cassin’s Sparrow’s ground-dwelling habit and the importance of microclimatic conditions to almost all aspects of the species’ life (Schnase and Maxwell 1989, Schnase et al. 1991). Our results are also consistent with research showing that increasing temperatures, increased temperature and precipitation variability, and drying soils are potent drivers of environmental suitability for many species found across the arid grasslands of the southwestern U.S. (Varner and Dearing 2014, Lortie et al. 2022). All told, it appears that spatiotemporal attributes of North American monsoon precipitation; variables related to the surface moisture conditions of the soil, air, and vegetation; and surface wind speed should be considered variables of particular relevance to Cassin’s Sparrow and perhaps key to understanding the species’ conservation status. It is important to note, however, that large-scale inferences are often complicated by changing properties in the underlying processes that drive phenomena, such as the spatially and temporally varying trends we see in the M2 and MC variables (Rollinson et al. 2021). As a result, Cassin’s Sparrow’s response to these non-constant trends tends to be complex and localized, which limits our ability to generalize about widespread trends in overall conditions that may be influencing the species.

CONCLUSION

We have used an automated and reproducible approach to retrospective ENM to better understand historical patterns of changing climatic suitability for Cassin’s Sparrow, the key drivers underlying the species’ response to a changing climate, and whether that response might contribute to the variability we see across various status assessments for the species. To do this, we built two multi-decadal, time- and variable-specific MaxEnt time series based on NASA’s MERRA-2 reanalysis. Trend analysis of the MC time series provided a seasonal, macroclimatic perspective that confirmed the importance of monsoon rainfall in Cassin’s Sparrow’s response to changing environmental conditions. Trend analysis of the M2 time series provided a microclimatic view that revealed the importance of increasing wind speed and drying land surface conditions to the species’ response. Estimates of climatic suitability for Cassin’s Sparrow in the MC time series were generally more favorable than what we observed in the M2 time series. The differences between the two time series were even more pronounced when considered at the state level. These two findings could have important implications for conservation practice, given that the vast majority of conservation policy decisions and management actions play out at the state level and often rely heavily on climatic suitability models based solely on bioclimatic predictors. Furthermore, our results suggest that modeled estimates of climatic suitability for Cassin’s Sparrow can vary widely depending on the temporal frame, spatial extent, and environmental drivers considered. This likely reflects Cassin’s Sparrow’s localized response to non-constant, spatiotemporal trends in the drivers themselves. This variability mirrors the inconsistencies we see in the current literature regarding this species’ status and points to the need for an updated, state-by-state examination of Cassin’s Sparrow’s regional vulnerabilities. Given the variability we observed in climatic suitability across areas occupied by Cassin’s Sparrow during the breeding season, a closer look at the nature and scope of historical, inter-annual range changes and local population trends would also be helpful in clarifying the relationship between climatic suitability and the demographic status of the species. Retrospective ENM that incorporates both time and variable specificity, as we have done here, appears to be a promising approach to understanding the complexities of a species’ response to changing climatic conditions and a useful adjunct to assessments of a species’ conservation status.

LITERATURE CITED

Adams, D. K., and A. C. Comrie. 1997. The North American monsoon. Bulletin of the American Meteorological Society 78:2197-2213. https://doi.org/10.1175/1520-0477(1997)078<2197:TNAM>2.0.CO;2

Aiello-Lammens, M. E., R. A. Boria, A. Radosavljevic, B. Vilela, and R. P. Anderson. 2015. spThin: an R package for spatial thinning of species occurrence records for use in ecological niche models. Ecography 38(5):541-545. https://doi.org/10.1111/ecog.01132

Akaike, H. 1974. A new look at the statistical model identification. IEEE Transactions on Automatic Control 19(6):716-723. https://doi.org/10.1109/TAC.1974.1100705

Allouche, O., A. Tsoar, and R. Kadmon. 2006. Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS). Journal of Applied Ecology 43(6):1223-1232. https://doi.org/10.1111/j.1365-2664.2006.01214.x

Arenas-Castro, S., J. Gonçalves, P. Alves, D. Alcaraz-Segura, and J. P. Honrado. 2018. Assessing the multi-scale predictive ability of ecosystem functional attributes for species distribution modelling. PLoS ONE 13(6):e0199292. https://doi.org/10.1371/journal.pone.0199292

Arenas-Castro, S., and N. Sillero. 2021. Cross-scale monitoring of habitat suitability changes using satellite time series and ecological niche models. Science of The Total Environment 784:147172. https://doi.org/10.1016/j.scitotenv.2021.147172

Bateman, B. L., A. M. Pidgeon, V. C. Radeloff, J. VanDerWal, W. E. Thogmartin, S. J. Vavrus, and P. J. Heglund. 2016. The pace of past climate change vs. potential bird distributions and land use in the United States. Global Change Biology 22:1130-1144. https://doi.org/10.1111/gcb.13154

Becker, E. 2021. The North American Monsoon. ENSO Blog. https://www.climate.gov/news-features/blogs/enso/north-american-monsoon

Bede-Fazekas, Á., and I. Somodi. 2020. The way bioclimatic variables are calculated has impact on potential distribution models. Methods in Ecology and Evolution 11(2):1559-1570. https://doi.org/10.1111/2041-210X.13488

Boria, R. A., L. E. Olson, S. M. Goodman, and R. P. Anderson. 2014. Spatial filtering to reduce sampling bias can improve the performance of ecological niche models. Ecological Modelling 275:73-77. https://doi.org/10.1016/j.ecolmodel.2013.12.012

Bosilovich, M. G., R. Lucchesi, and M. Suarez. 2016. MERRA-2: file specification. NASA Global Modeling and Assimilation Office Note No. 9 (Version 1.1). NASA Goddard Space Flight Center, Greenbelt, Maryland, USA. https://gmao.gsfc.nasa.gov/pubs/docs/Bosilovich785.pdf

Cabello, J., N. Fernández, D. Alcaraz-Segura, C. Oyonarte, G. Piñeiro, A. Altesor, M. Delibes, and J. M. Paruelo. 2012. The ecosystem functioning dimension in conservation: insights from remote sensing. Biodiversity and Conservation 21:3287-3305. https://doi.org/10.1007/s10531-012-0370-7

Cavalcante, T., M. M. Weber, and A. A. Barnett. 2022. Combining geospatial abundance and ecological niche models to identify high-priority areas for conservation: the neglected role of broadscale interspecific competition. Frontiers in Ecology and Evolution 10:915325. https://doi.org/10.3389/fevo.2022.915325

Cheng, Y., P. W. Chan, X. Wei, Z. Hu, Z. Kuang, and K. A. McColl. 2021. Soil moisture control of precipitation reevaporation over a heterogeneous land surface. Journal of the Atmospheric Sciences 78:3369-3383. https://doi.org/10.1175/JAS-D-21-0059.1

Dunning, J. B., R. K. Bowers, Jr., S. J. Suter, and C. E. Bock. 2020. Cassin’s Sparrow (Peucaea cassinii) version 1.0. In P. G. Rodewald, editor. Birds of the world. Cornell Lab of Ornithology, Ithaca, New York, USA. https://doi.org/10.2173/bow.casspa.01

Edwards, P. N. 2010. A vast machine: computer models, climate data, and the politics of global warming. MIT Press, Cambridge, Massachusetts, USA.

Elith, J., S. J. Phillips, T. Hastie, M. Dudík, Y. E. Chee, and C. J. Yates. 2011. A statistical explanation of MaxEnt for ecologists. Diversity and Distributions 17:43-57. https://doi.org/10.1111/j.1472-4642.2010.00725.x

Fielding, A. H., and J. F. Bell. 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environmental Conservation 24(1):38-49. https://doi.org/10.1017/S0376892997000088

Gelaro, R., W. McCarty, M. J. Suárez, R. Todling, A. Molod, L. Takacs, C. A. Randles, A. Darmenov, M. G. Bosilovich, R. Reichle, et al. 2017. The modern-era retrospective analysis for research and applications, version 2 (MERRA-2). Journal of Climate 30:5419-5454. https://doi.org/10.1175/JCLI-D-16-0758.1

Goberville, E., G. Beaugrand, N.-C. Hautekèete, Y. Piquot, and C. Luczak. 2015. Uncertainties in the projection of species distributions related to general circulation models. Ecology and Evolution 5(5):1100-1116. https://doi.org/10.1002/ece3.1411

Gonçalves, J., P. Alves, I. Pôças, B. Marcos, R. Sousa-Silva, Â. Lomba, and J. P. Honrado. 2016. Exploring the spatiotemporal dynamics of habitat suitability to improve conservation management of a vulnerable plant species. Biodiversity and Conservation 25:2867-2888. https://doi.org/10.1007/s10531-016-1206-7

Guida, R. J., S. R. Abella, C. L. Roberts, C. M. Norman, and W. J. Smith, Jr. 2019. Assessing historical and future habitat models for four conservation-priority Mojave Desert species. Journal of Biogeography 46:2081-2097. https://doi.org/10.1111/jbi.13645

Guisan, A., R. Tingley, J. B. Baumgartner, I. Naujokaitis‐Lewis, P. R. Sutcliffe, A. I. T. Tulloch, T. J. Regan, L. Brotons, E. McDonald‐Madden, C. Mantyka‐Pringle, et al. 2013. Predicting species distributions for conservation decisions. Ecology Letters 16:1424-1435. https://doi.org/10.1111/ele.12189

Heenan, C. B., and R. S. Seymour. 2012. The effect of wind on the rate of heat loss from avian cup-shaped nests. PLoS ONE 7(2):e32252. https://doi.org/10.1371/journal.pone.0032252

Hijmans, R. J., S. Phillips, J. Leathwick, and J. Elith. 2023. dismo: species distribution modeling. R package version 1.3-1.4. https://cran.r-project.org/web/packages/dismo/index.html

Huang, Q., B. L. Bateman, N. L. Michel, A. M. Pidgeon, V. C. Radeloff, P. Heglund, A. J. Allstadt, A. J. Nowakowski, J. Wong, and J. R. Sauer. 2023. Modeled distribution shifts of North American birds over four decades based on suitable climate alone do not predict observed shifts. Science of The Total Environment 857(3):159603. https://doi.org/10.1016/j.scitotenv.2022.159603

Hubbard, J. P. 1974. Avian evolution in the aridlands of North America. Living Bird 12:155-196.

Huntley, B., P. Barnard, R. Altwegg, L. Chambers, B. W. T. Coetzee, L. Gibson, P. A. R. Hockey, D. G. Hole, G. F. Midgley, L. G. Underhill, and S. G. Willis. 2010. Beyond bioclimatic envelopes: dynamic species’ range and abundance modelling in the context of climatic change. Ecography 33:621-626. https://doi.org/10.1111/j.1600-0587.2009.06023.x

Iknayan, K. J., and S. R. Beissinger. 2018. Collapse of a desert bird community over the past century driven by climate change. Proceedings of the National Academy of Sciences 115(34):8597-8602. https://doi.org/10.1073/pnas.1805123115

Ingenloff, K. 2020. Enhancing the correlative ecological niche modeling framework to incorporate the temporal dimension of species’ distributions. Dissertation. University of Kansas, Lawrence, Kansas, USA. https://www.proquest.com/openview/56e595f00e55b6ffa8cf92c702a3dc96/1?pq-origsite=gscholar&cbl=18750&diss=y

Ingenloff, K., and A. T. Peterson. 2021. Incorporating time into the traditional correlational distributional modelling framework: a proof-of-concept using the Wood Thrush Hylocichla mustelina. Methods in Ecology and Evolution 12:311-321. https://doi.org/10.1111/2041-210X.13523

Johnston, A., T. Auer, D. Fink, M. Strimas‐Mackey, M. Iliff, K. V. Rosenberg, S. Brown, R. Lanctot, A. D. Rodewald, and S. Kelling. 2020. Comparing abundance distributions and range maps in spatial conservation planning for migratory species. Ecological Applications 30(3):02058. https://doi.org/10.1002/eap.2058

Kalinski, C. E. 2019. Building better species distribution models with machine learning: assessing the role of covariate scale and tuning in Maxent models. Dissertation. University of Southern California, Los Angeles, California, USA.

Kass, J. M., R. Muscarella, P. J. Galante, C. Bohl, G. E. Buitrago-Pinilla, R. A. Boria, M. Soley-Guardia, and R. P. Anderson. 2023. ENMeval: automated tuning and evaluations of ecological niche models. R package version 2.0.4. https://cran.r-project.org/web/packages/ENMeval/index.html

Lehmann, P., O. Merlin, P. Gentine, and D. Or. 2018. Soil texture effects on surface resistance to bare‐soil evaporation. Geophysical Research Letters 45(19):10398-10405. https://doi.org/10.1029/2018GL078803

Leitão, P. J., and M. J. Santos. 2019. Improving models of species ecological niches: a remote sensing overview. Frontiers in Ecology and Evolution 7:9. https://doi.org/10.3389/fevo.2019.00009

Lian, X., W. Zhao, and P. Gentine. 2022. Recent global decline in rainfall interception loss due to altered rainfall regimes. Nature Communications 13:7642. https://doi.org/10.1038/s41467-022-35414-y

Lipschutz, M. L. 2016. Effects of drought and grazing on land bird populations in South Texas. Dissertation. Texas A&M University-Kingsville, Kingsville, Texas, USA. https://www.proquest.com/dissertations-theses/effects-drought-grazing-on-land-bird-populations/docview/1814764436/se-2

Lortie, C. J., A. Filazzola, M. Westphal, and H. S. Butterfield. 2022. Foundation plant species provide resilience and microclimatic heterogeneity in drylands. Scientific Reports 12:18005. https://doi.org/10.1038/s41598-022-22579-1

Lynn, J. 2006. Cassin’s Sparrow (Aimophila cassinii): a technical conservation assessment. U.S. Department of Agriculture Forest Service, Rocky Mountain Region, Lakewood, Colorado, USA. https://www.fs.usda.gov/Internet/FSE_DOCUMENTS/stelprdb5182053.pdf

Maguire, B., Jr. 1973. Niche response structure and the analytical potentials of its relationship to the habitat. American Naturalist 107:213-246. https://doi.org/10.1086/282827

Merow, C., M. J. Smith, and J. A. Silander, Jr. 2013. A practical guide to MaxEnt for modeling species’ distributions: what it does, and why inputs and settings matter. Ecography 36(10):1058-1069. https://doi.org/10.1111/j.1600-0587.2013.07872.x

Muscarella, R., P. J. Galante, M. Soley-Guardia, R. A. Boria, J. M. Kass, M. Uriarte, and R. P. Anderson. 2014. ENMeval: an R package for conducting spatially independent evaluations and estimating optimal model complexity for Maxent ecological niche models. Methods in Ecology and Evolution 5(11):1198-1205. https://doi.org/10.1111/2041-210X.12261

New Mexico Department of Game and Fish. 2016. State wildlife action plan for New Mexico. New Mexico Department of Game and Fish, Santa Fe, New Mexico, USA. https://www.wildlife.state.nm.us/wpfb-file/new-mexico-state-wildlife-action-plan-swap-final-2019-pdf

Norman, J. A., and L. Christidis. 2016. Ecological opportunity and the evolution of habitat preferences in an arid-zone bird: implications for speciation in a climate-modified landscape. Scientific Reports 6:19613. https://doi.org/10.1038/srep19613

O’Donnell, M. S., and D. A. Ignizio. 2012. Bioclimatic predictors for supporting ecological applications in the conterminous United States. U.S. Geological Survey Data Series 691. U.S. Geological Survey, Reston, Virginia, USA. https://doi.org/10.3133/ds691

Ohmart, R. D. 1969. Dual breeding ranges in Cassin’s Sparrow (Aimophila cassinii). Page 105 in C. C. Hoff and M. L. Riedesel, editors. Physiological systems in semiarid environments. University of New Mexico Press, Albuquerque, New Mexico, USA.

Pearson, R. G., C. J. Raxworthy, M. Nakamura, and A. Townsend Peterson. 2007. Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar. Journal of Biogeography 34:102-117. https://doi.org/10.1111/j.1365-2699.2006.01594.x

Pecl, G. T., M. B. Araújo, J. D. Bell, J. Blanchard, T. C. Bonebrake, I.-C. Chen, T. D. Clark, R. K. Colwell, F. Danielsen, B. Evengård, et al. 2017. Biodiversity redistribution under climate change: impacts on ecosystems and human well-being. Science 355:eaai9214. https://doi.org/10.1126/science.aai9214

Peevey, T. 2022. Changes in US wind speeds since 2000 point to a dynamic landscape. https://www.spglobal.com/marketintelligence/en/news-insights/research/changes-in-us-wind-speeds-since-2000-point-to-a-dynamic-landscape

Pérez‐Navarro, M. A., O. Broennimann, M. A. Esteve, J. M. Moya‐Perez, M. F. Carreño, A. Guisan, and F. Lloret. 2021. Temporal variability is key to modelling the climatic niche. Diversity and Distributions 27:473-484. https://doi.org/10.1111/ddi.13207

Peterson, A. T., J. Soberón, R. G. Pearson, R. P. Anderson, E. Martínez-Meyer, M. Nakamura, and M. B. Araújo. 2011. Ecological niches and geographic distributions (MPB-49). Princeton University Press, Princeton, New Jersey, USA. https://doi.org/10.23943/princeton/9780691136868.001.0001

Petitpierre, B., O. Broennimann, C. Kueffer, C. Daehler, and A. Guisan. 2017. Selecting predictors to maximize the transferability of species distribution models: lessons from cross-continental plant invasions. Global Ecology and Biogeography 26:275-287. https://doi.org/10.1111/geb.12530

Pettorelli, N., H. Schulte to Bühne, A. Tulloch, G. Dubois, C. Macinnis-Ng, A. M. Queirós, D. A. Keith, M. Wegmann, F. Schrodt, M. Stellmes, et al. 2018. Satellite remote sensing of ecosystem functions: opportunities, challenges and way forward. Remote Sensing in Ecology and Conservation 4(2):71-93. https://doi.org/10.1002/rse2.59

Phillips, S. J., R. P. Anderson, M. Dudík, R. E. Schapire, and M. E. Blair. 2017. Opening the black box: an open-source release of Maxent. Ecography 40(7):887-893. https://doi.org/10.1111/ecog.03049

Phillips, S. J., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling 190(3-4):231-259. https://doi.org/10.1016/j.ecolmodel.2005.03.026

Porzig, E. L., N. E. Seavy, T. Gardali, G. R. Geupel, M. Holyoak, and J. M. Eadie. 2014. Habitat suitability through time: using time series and habitat models to understand changes in bird density. Ecosphere 5:12. https://doi.org/10.1890/ES13-00166.1

Pradhan, P. 2016. Strengthening MaxEnt modelling through screening of redundant explanatory bioclimatic variables with variance inflation factor analysis. Researcher 8(5):29-34. https://www.academia.edu/28906048/Strengthening_MaxEnt_modelling_through_screening_of_redundant_explanatory_Bioclimatic_Variables_with_Variance_Inflation_Factor_analysis

Regos, A., P. Gómez-Rodríguez, S. Arenas-Castro, L. Tapia, M. Vidal, and J. Domínguez. 2020. Model-assisted bird monitoring based on remotely sensed ecosystem functioning and atlas data. Remote Sensing 12:2549. https://doi.org/10.3390/rs12162549

Regos, A., J. Gonçalves, S. Arenas‐Castro, D. Alcaraz‐Segura, A. Guisan, and J. P. Honrado. 2022a. Mainstreaming remotely sensed ecosystem functioning in ecological niche models. Remote Sensing in Ecology and Conservation 8(4):431-447. https://doi.org/10.1002/rse2.255

Regos, A., L. Tapia, S. Arenas-Castro, A. Gil-Carrera, and J. Domínguez. 2022b. Ecosystem functioning influences species fitness at upper trophic levels. Ecosystems 25:1037-1051. https://doi.org/10.1007/s10021-021-00699-5

Reside, A. E., J. J. VanDerWal, A. S. Kutt, and G. C. Perkins. 2010. Weather, not climate, defines distributions of vagile bird species. PLoS ONE 5(10):e13569. https://doi.org/10.1371/journal.pone.0013569

Rollinson, C. R., A. O. Finley, M. R. Alexander, S. Banerjee, K.-A. Dixon Hamil, L. E. Koenig, D. H. Locke, M. L. DeMarche, M. W. Tingley, K. Wheeler, et al. 2021. Working across space and time: nonstationarity in ecological research and application. Frontiers in Ecology and the Environment 19(1):66-72. https://doi.org/10.1002/fee.2298

Rosenberg, K. V., A. M. Dokter, P. J. Blancher, J. R. Sauer, A. C. Smith, P. A. Smith, J. C. Stanton, A. Panjabi, L. Helft, M. Parr, and P. P. Marra. 2019. Decline of the North American avifauna. Science 366(6461):120-124. https://doi.org/10.1126/science.aaw1313

Rosenberg, K. V., J. A. Kennedy, R. Dettmers, R. P. Ford, D. Reynolds, J. D. Alexander, C. J. Beardmore, P. J. Blancher, R. E. Bogart, G. S. Butcher, et al. 2016. Partners in Flight landbird conservation plan: 2016 revision for Canada and continental United States. https://partnersinflight.org/resources/the-plan/

Roubicek, A. J., J. VanDerWal, L. J. Beaumont, A. J. Pitman, P. Wilson, and L. Hughes. 2010. Does the choice of climate baseline matter in ecological niche modelling? Ecological Modelling 221(19):2280-2286. https://doi.org/10.1016/j.ecolmodel.2010.06.021

Ruth, J. M. 2000. Cassin’s Sparrow (Aimophila cassinii) status assessment and conservation plan. U.S. Fish and Wildlife Service Biological Technical Publication BTP-R6002-2000. U.S. Geological Survey Midcontinent Ecological Science Center, Fort Collins, Colorado, USA. https://digitalmedia.fws.gov/digital/collection/document/id/1235/rec/4

Salas, E. A. L., V. A. Seamster, K. G. Boykin, N. M. Harings, and K. W. Dixon. 2017. Modeling the impacts of climate change on Species of Concern (birds) in South Central U.S. based on bioclimatic variables. AIMS Environmental Science 4(2):358-385. https://doi.org/10.3934/environsci.2017.2.358

Schnase, J. L., and M. L. Carroll. 2022. Automatic variable selection in ecological niche modeling: a case study using Cassin’s Sparrow (Peucaea cassinii). PLoS ONE 17(1):e0257502. https://doi.org/10.1371/journal.pone.0257502

Schnase, J. L., M. L. Carroll, R. L. Gill, G. S. Tamkin, J. Li, S. L. Strong, T. P. Maxwell, M. E. Aronne, and C. S. Spradlin. 2021. Toward a Monte Carlo approach to selecting climate variables in MaxEnt. PLoS ONE 16(3):e0237208. https://doi.org/10.1371/journal.pone.0237208

Schnase, J. L., W. E. Grant, T. C. Maxwell, and J. J. Leggett. 1991. Time and energy budgets of Cassin’s Sparrow (Aimophila cassinii) during the breeding season: evaluation through modelling. Ecological Modelling 55(3-4):285-319. https://doi.org/10.1016/0304-3800(91)90091-E

Schnase, J. L., and T. C. Maxwell. 1989. Use of song patterns to identify individual male Cassin’s Sparrows. Journal of Field Ornithology 60(1):12-19. https://sora.unm.edu/sites/default/files/journals/jfo/v060n01/p0012-p0019.pdf

Searcy, C. A., and H. B. Shaffer. 2016. Do ecological niche models accurately identify climatic determinants of species ranges? American Naturalist 187(4):423-435. https://doi.org/10.1086/685387

Sen, P. K. 1968. Estimates of the regression coefficient based on Kendall’s tau. Journal of the American Statistical Association 63(324):1379-1389. https://doi.org/10.1080/01621459.1968.10480934

Smith, A. B., and M. J. Santos. 2020. Testing the ability of species distribution models to infer variable importance. Ecography 43(12):1801-1813. https://doi.org/10.1111/ecog.05317

Sohl, T. L. 2014. The relative impacts of climate and land-use change on conterminous United States bird species from 2001 to 2075. PLoS ONE 9(11):e112251. https://doi.org/10.1371/journal.pone.0112251

Stahle, D. W. 2020. Anthropogenic megadrought. Science 368(6488):238-239. https://doi.org/10.1126/science.abb6902

Theil, H. 1950. A rank-invariant method of linear and polynomial regression analysis. Indagationes Mathematicae 12:386-392. https://dwc.knaw.nl/DL/publications/PU00018789.pdf

Tran, D. T. Q., D. G. Fredlund, and D. H. Chan. 2016. Improvements to the calculation of actual evaporation from bare soil surfaces. Canadian Geotechnical Journal 53(1):118-133. https://doi.org/10.1139/cgj-2014-0512

VanDerWal, J., H. T. Murphy, A. S. Kutt, G. C. Perkins, B. L. Bateman, J. J. Perry, and A. E. Reside. 2013. Focus on poleward shifts in species’ distribution underestimates the fingerprint of climate change. Nature Climate Change 3:239-243. https://doi.org/10.1038/nclimate1688

Varner, J., and M. D. Dearing. 2014. The importance of biologically relevant microclimates in habitat suitability assessments. PLoS ONE 9(8):e104648. https://doi.org/10.1371/journal.pone.0104648

Vega, G. C., L. R. Pertierra, and M. Á. Olalla-Tárraga. 2017. MERRAclim, a high-resolution global dataset of remotely sensed bioclimatic variables for ecological modelling. Scientific Data 4:170078. https://doi.org/10.1038/sdata.2017.78

Ward, E. J., L. A. K. Barnett, S. C. Anderson, C. J. C. Commander, and T. E. Essington. 2022. Incorporating non-stationary spatial variability into dynamic species distribution models. ICES Journal of Marine Science 79(9):2422-2429. https://doi.org/10.1093/icesjms/fsac179

Warren, D. L., and S. N. Seifert. 2011. Ecological niche modeling in Maxent: the importance of model complexity and the performance of model selection criteria. Ecological Applications 21(2):335-342. https://doi.org/10.1890/10-1171.1

Weber, M. M., R. D. Stevens, J. A. F. Diniz-Filho, and C. E. V. Grelle. 2017. Is there a correlation between abundance and environmental suitability derived from ecological niche modelling? A meta-analysis. Ecography 40(7):817-828. https://doi.org/10.1111/ecog.02125

Williams, A. P., E. R. Cook, J. E. Smerdon, B. I. Cook, J. T. Abatzoglou, K. Bolles, S. H. Baek, A. M. Badger, and B. Livneh. 2020. Large contribution from anthropogenic warming to an emerging North American megadrought. Science 368(6488):314-318. https://doi.org/10.1126/science.aaz9600

Williams, F. C., and A. L. LeSassier. 1968. Cassin’s Sparrow. Pages 981-990 in O. L. Austin, editor. Life histories of North American cardinals, grosbeaks, buntings, towhees, finches, sparrows, and allies, order Passeriformes, family Fringillidae: part 2, genera Pipilo through Spizella. Dover, New York, New York, USA.

Wilsey, C., L. Taylor, B. Bateman, C. Jensen, N. Michel, A. Panjabi, and G. Langham. 2019. Climate policy action needed to reduce vulnerability of conservation‐reliant grassland birds in North America. Conservation Science and Practice 1(4):e21. https://doi.org/10.1111/csp2.21

Woodhouse, S. W. 1852. Zonotrichia Cassinii, nobis. Proceedings of the Academy of Natural Science of Philadelphia 60-61.