INTRODUCTION

Climate change during the late Anthropocene is well documented (e.g., ^{Hartmann et al. 2013}; ^{IPCC 2013}; ^{Alexander 2015}; ^{Hoegh-Guldberg et al. 2018}). Numerous predictions have been published regarding the short-to long-term consequences of climate change on human populations (e.g., ^{Costello et al. 2009}; ^{Anderson et al. 2019}; ^{Hoegh-Guldberg et al. 2018}). Similarly, many studies have documented or modeled the extirpation of populations or loss of wildlife species diversity due to climate change (e.g., ^{Sheldon 2011}; ^{Ashrafzadeh et al. 2018}; ^{Hidasi-Neto et al. 2019}; ^{Wan et al. 2019}).

As emergent qualities arising from the combination of shifting distributional limits and populational viability, community composition and structure may also experience medium- or long-term directional changes (^{Wake & Vredenburg 2008}; ^{Sheldon 2011}; ^{Torres et al. 2015}). Several studies have focused on changes in mammalian community structure or composition resulting from climate change. An exemplary series of papers resulted from the Grinnell Resurvey Project in California (e.g., ^{Moritz et al. 2008}; ^{Rubidge et al. 2011}; ^{Santos et al. 2014}, ²⁰¹⁷), and others have also documented changes in range and relative abundance of small mammals (e.g., ^{Myers et al. 2009}). At the Sevilleta Long-term Ecological Research site in New Mexico USA, ^{Morgan Ernest et al. (2000)} found that precipitation, plant cover, and rodent dynamics were temporally correlated, and that rodent dynamics depend on both biotic and abiotic factors in their environment.

In South America, Meserve and colleagues have conducted long-term studies in temperate forests and a subtropical semiarid scrubland (e.g., ^{Meserve et al. 2009}; ^{Gutiérrez et al. 2010}; ^{Meserve et al. 2011}; ^{Meserve 2016}). Two studies in the Argentine Pampas have examined changes in rodent species abundances and community composition in agroecosystems (^{Fraschina et al. 2011}; ^{Polop et al. 2012}). However, these have been limited geographically and climatologically to subtropical and mostly temperate regions, and no study has investigated long-term effects of climate change on small mammals in tropical South American forests.

Medium- or long-term changes in community structure are inherently difficult to assess with shorter-term sampling, because observed patterns may be confounded by short-term climatic and other extrinsic variation in their environment (^{Naxara et al. 2009}; ^{Owen 2013}; ^{Ribeiro et al. 2019}; ^{Barreto Cáceres & Owen 2019}), or changes in land use or land cover (^{Bonvicino et al. 2002}; ^{Fraschina et al. 2011}; ^{Delciellos et al. 2016}; ^{Barreto Cáceres & Owen 2019}; ^{Owen et al. 2019}). Thus, long-term studies in areas without land use change are needed to evaluate long-term changes in community structure due to climate change. This paper documents the increasing relative abundance of a tropical rodent in a diverse sigmodontine community in the Upper Paraná Atlantic Forest of Paraguay, over a period of 22 years.

MATERIALS AND METHODS

The Reserva Natural del Bosque Mbaracayú (RNBM, hereafter) was created in 1991 from a large forested area in Canindeyú department, northeastern Paraguay. With subsequent additions to the property, the RNBM currently includes 64405 ha and lies between 24°00’ and 24°15’ S latitude and 55°00’ and 55°32’ W longitude. The main body of the RNBM is covered by virtually pristine Upper Paraná Atlantic Forest (UPAF, hereafter), although there was minor selective logging in the 1970s (^{FMB/BM 2005}). A combination of remote imaging and extensive ground-truthing have resulted in a detailed vegetation map of the reserve, including eight general forest types and several small isolated natural grasslands within the forest matrix (^{Naidoo & Hill 2006}) (Fig. 1).

The RNBM was extensively sampled for small mammals in three periods during 22 years, beginning in 1996. All captures of sigmodontine rodents were recorded, including those collected and those encountered in mark-recapture sampling. Although exact information is not available from all sessions, this author led all sampling sessions, and sampling efforts were always general (i.e., targeted toward the broadest taxonomic and habitat representation). Sampling could be expected to capture the various species in similar proportions across sampling sessions, so that individual species capture probabilities did not vary due to specific type of sampling. Because capture effort varied among sampling sessions, proportional abundances of the species were used for comparisons, rather than raw numbers of captures in each year.

Captures of sigmodontines were recorded in 11 years, but years with <75 total sigmodontine captures were excluded from analyses, leaving eight years with data to be an alyzed. Simpson’s Diversity Index λ = i ni (ni −1) was calculated for each of the eight years. This is a dominance index which is relatively insensitive to species richness and heavily weighted toward the more abundant N (N −1)species. Thus the presence or absence of rarer species in the capture record should not affect this index.

Importar imagen Importar imagen Monthly temperature averages and precipitation totals were downloaded from <https://crudata.uea.ac.uk/cru/data/hrg/cruts4.02/ge/>; (^{Harris et al. 2014}), and summarized as annual averages and totals, respectively (Table S1). Both annual temperature and precipitation were regressed against year (1995-2017) to evaluate recent climatic trends at the RNBM. Regression analyses of percentage representation (for species represented by >10% of the records overall, along with “other”—the remaining species) were conducted against year, average temperature and total precipitation. Simpson diversity indices were also regressed against year, average temperature and total precipitation. All statistical procedures were conducted with SigmaPlot 12.3 (^{Systat Software, Inc. 2011}).

RESULTS

The dataset includes 4503 captures of sigmodontine rodents (Table S2). At least 16 species of sigmodontine rodents were recorded. Three years had <75 total captures and were dropped from further analyses. Of the remaining 4473 records, species with >10% overall percentage were Akodon montensis Thomas, 1913 (63.5%), Hylaeamys megacephalus (G. Fischer, 1814) (15.6%) and Oligoryzomys nigripes (Olfers, 1818) (10.1%). Each of the other species represented <5% of the total captures, and the remaining species combined (“other”, hereafter) represented 10.8% of the total captures (Table 1).

The 4473 captures occurred during the eight years being evaluated. From one to nine months of each year included sampling and captures. Every month is represented by at least one capture during the eight years. The months of highest captures for each of the eight years were scattered across seven different months, and the months that had the highest numbers of captures (> 500) across the eight years included February, June – August and November. Thus it appears unlikely that any observed long-term trend would be an artifact of differential sampling among years (Table 2).

Linear regressions indicated that both temperature and precipitation increased during the period 1995-2017. Although neither regression coefficient was statistically significant, and the power of both tests was relatively low, the data indicate that during the 22 years of the study, average temperature increased by nearly 0.4°C, and average annual precipitation by about 140 mm (Fig. 2, Table S1).

None of the three predominant rodent species varied significantly with either annual temperature or precipitation, nor did A. montensis or O. nigripes vary significantly with year (Table 3). In contrast, H. megacephalus increased significantly with year, registering an average overall increase of 14.2% of their proportion of the sigmodontine population (Fig. 3). As H. megacephalus increased and A. montensis and nigripes remained relatively constant, “other” sigmodontines registered a marginally non-significant proportional decrease of 1.33% per year, or 29.3% of their proportion of the overall population during the 22 years (Table 3). Although the proportional decrease in the less common sigmodontine species, along with the increase in H. megacephalus, would suggest that species diversity was decreasing during the study period, Simpson’s λ was not linearly associated with year (F1,6=0.0449, p=0.839).

A. montensis has been consistently and by far the most abundant species in the forests of the Mbaracayú Reserve. O. nigripes was initially the second most abundant species, followed by H. megacephalus (Table 1, Fig. 4). However in the more recent sampling period (2014-17), populations of H. megacephalus have consistently been higher than those of O. nigripes (Fig. 4). O. nigripes exhibits considerable variation in abundance, with no individuals encountered in 2003 or 2006, but representing 18.3% of the sigmodontine captures in 2005 (Table 1).

Fig. 1 Map of the Reserva Natural del Bosque Mbaracayú, Canindeyú Department, northeastern Paraguay, showing matrix of forest types and other habitats. Inset map of Paraguay, showing location of Canindeyú Department and the Reserve. 

Table 1 Number of captures and percentage of total captures for each of eight years, for Akodon montensis (Akmo), Hylaeamys megacephalus (Hyme), Oligoryzomys nigripes (Olni) and all other sigmodontine species combined (OtherS). Simpson Diversity Index λ, average temperature (°C) and total precipitation (mm) are also listed. 

Table 2 Total number of captures during each month of each of the eight years evaluated. Numbers in bold face indicate the month with the highest number of captures for each year, and the highest numbers among the totals for each month across the eight years. 

Fig. 2 Linear regressions of (A) average annual temperature on year, and (B) total annual precipitation on year. During the 22 years of the study in Canindeyú Department, northeastern Paraguay, average annual temperature has risen 0.018o C per year (0.396°C overall) and total annual precipitation by 6 371 mm per year (140.2 mm overall). 

Table 3 Linear regression results from regressions of Akodon montensis, Hylaeamys megacephalus, Oligoryzomys nigripes and “Other” (all other sigmodontine species lumped) against year, average annual temperature and total annual precipitation. In each case, coefficient indicates slope of regression line, p is two-tailed probability, and 1 − β is the power of the performed test. 

Fig. 3 Regression of Hylaeamys megacephalus proportion of population on year, showing average increase through 22-year study period in the Reserva Natural del Bosque Mbaracayú, Canindeyú Department, northeastern Paraguay. 

DISCUSSION

Several studies from North America have conclusively demonstrated long-term changes in rodent community composition and structure, as well as species distributional limits. ^{Myers et al. (2009)} analyzed 30 years of data from the northern Great Lakes region and found that species reaching their northern limit were increasing and those reaching their southern limit were decreasing in abundance. Moreover, two southern species had extended their ranges 225 km northward during that time.

The Grinnell Resurvey Project produced an important series of articles documenting changes in elevational limits in small mammal species (^{Moritz et al. 2008}), differing in extent of the effect of climate change on different species (^{Rubidge et al. 2011}), greater synchronicity of elevational range shifts within particular trophic niches (^{Santos et al. 2014}) and the interaction of climate change and habitat on small mammal elevational ranges (^{Santos et al. 2017}), among others. Long-term studies have been conducted for 30 years in a subtropical semiarid site in Chile, documenting the diverse and interactive effects of predator exclusion, food supplementation, and ENSO events on the small mammal community (^{Gutiérrez et al. 2010}; ^{Meserve 2016}), and more recently focusing on climate change (^{Meserve et al.2009}, 2011).

Fig. 4 Proportions of sigmodontine population of three species and “other” sigmodontines, plotted for the 22-year period of the study in the Reserva Natural del Bosque Mbaracayú, Canindeyú Department, northeastern Paraguay. 

Two studies in Argentine Pampean agroecosystems have evaluated long-term changes in sigmodontine species populations and community composition, with both studies finding differential responses among the species present. ^{Fraschina et al. (2011)} found that total rodent abundance during autumn and winter decreased over time in crop field borders, but not in the crop fields. Moreover, different species responded differently over time, with two Calomys species decreasing during the study period and Akodon azarae remaining stable. ^{Polop et al. (2012)} reported that A. azarae abundance increased over a 12-year period, while that of Akodon dolores decreased. However the two species responded differently to agricultural expansion. In both of these studies, climatic variables (temperature and precipitation) were found to be important determinants, although not consistently among the species examined.

The RNBM lies in the tropical-subtropical interface of South America (<1° south of the Tropic of Capricorn), and near the western limit of the Upper Parana Atlantic Forest. Of the three most abundant sigmodontine species encountered there, A. montensis and O. nigripes have extensive tropical and subtropical distributions, extending to approximately 30° and 35° S latitude, respectively. Only H. megacephalus is almost completely limited to the tropics, occurring in lowland tropical rainforests and gallery forests from the northern coast of Venezuela through northeastern Paraguay (Fig. 5). Thus, the RNBM is near the extreme southwestern limit of the known distribution of H. megacephalus.

It is noteworthy that H. megacephalus is an important component of the sigmodontine community at a location this near its distributional limit. That it is experiencing an overall increase in proportional abundance in that rodent community is even more noteworthy. Substantial short-term fluctuations in the proportional abundances of the three predominant species occurred during the eight sampling years being reported here. In another study in the RNBM, ^{Barreto Cáceres & Owen (2019)} found that H. megacephalus populations (as well as those of other sigmodontine rodents) generally declined during the two years of their study (2015–2016), although H. megacephalus briefly increased following an extreme El Niño event in 2015. Another potential source of short-term variation might be a “ratada” following a masting event of the prominent understory bamboo species Chusquea ramossisima. However, such an event has not occurred within the memory of several people familiar with the flora of the RNBM (S. Fernández pers. comm.). Regardless, it bears reiterating that long-term studies are necessary to document long-term trends.

^{Wiens (2011)} defined the niche as “the combination of abiotic and biotic conditions where a species can persist”. Most studies reporting changing distributional limits, and/or community composition in small mammal communities as a result of climate change have reported that one or more environmental (abiotic) variables are associated with these changes in the small mammal community (e.g., ^{Parra & Monahan 2008}; ^{Garroway et al. 2010}; ^{Rubidge et al. 2011}; ^{Santos et al. 2017}; ^{Hannibal et al. 2020}). Although there have been long-term increases in both average temperature and annual precipitation in the RNBM, the yearly proportional abundances of the predominant species in this study are not significantly associated with these environmental measures, and that of H. megacephalus is significantly associated only with time, but not with temperature or precipitation, although if longer-term data were available these associations might prove also to be significant. Among other things, this suggests that although temperature and precipitation changes are components of climate change, climate change is a complex phenomenon that cannot be reduced to simple environmental measures.

Fig. 5 Distribution maps for (A) Akodon montensis, (B) Hylaeamys megacephalus and (C) Oligoryzomys nigripes. Red star on each map indicates location of the Reserva Natural del Bosque Mbaracayú, Canindeyú department, northeastern Paraguay. Shading in inset maps indicates distribution of the entire genus. Maps adapted from ^{Pardiñas et al. (2015)}, ^{Percequillo (2015)} and ^{Weksler & Bonvicino (2015)}, respectively. 

Also in the RNBM, ^{Eastwood et al. (2018)} encountered H. megacephalus in all Atlantic Forest habitats except patches dominated by large bamboo, and the species was common in most of the same habitats as A. montensis and O. nigripes. However, in a more detailed study of habitat and behavior, ^{Owen et al. (2020)} found H. megacephalus to be preferentially associated with bamboo understory, A. montensis with high forest, and O. nigripes to have no preferential habitat association. In the same study, each of these three species was found behaviorally to exhibit microsympatry with conspecifics, and to avoid trap stations frequented by the other species. In sum, these studies underline the complexity of the “combination of abiotic and biotic conditions where [H. megacephalus] can persist”, to paraphrase ^{Wiens (2011)}.

This is the first documentation of increased proportional abundance over time of a tropical rodent in a tropical-subtropical sigmodontine community. These results strongly highlight the need for long-term studies such as this one, to begin to understand the complexity of faunal response to climate change. This is especially urgent in tropical and subtropical areas of transition between ecoregions, where changes in small mammal community structure may result from the complex effects of climate change.

Supplementary materials