Effect of plant species and fraction (location: understory, litter, humus) on the magnitude of temperature fluctuations during the study period. Anovas were run using daily maximum, minimum and temperature range. There were 435 days (from September 8, 2004 to November 16, 2005).
Plant diversity is a key factor influencing belowground dynamics including microclimate and decomposer arthropod communities. This study addresses the effect of individual plant species on belowground arthropods by focusing on seasonal variations in precipitation, temperature and arthropods along the vertical organic matter profile. In the Guanica Dry Forest, Puerto Rico, microclimate was described and 5 plant species and 10 trees/species were selected. Under each tree, for one year, temperature was measured and samples collected along the organic matter fractions. Collected arthropods were standardized to ind/m2, identified to Order/Family and assigned to morphotypes. The annual temperature pattern was similar for all species and OM fractions. Arthropod abundance was similar among plant species and higher in humus than in litter fractions. Richness and species composition were different among plant species and OM fractions. All plant species and OM fractions showed low arthropod abundance and richness, and similar arthropod species composition in the dry season, while in the wet season abundance and richness were higher and species composition varied across plant species and OM fractions. These data suggest that arthropods form specific assemblages under plant species and stages of decomposition that, during the dry season, represent a subgroup adapted to extreme environmental conditions.
- Berlese funnels
- extreme conditions
- plant structure
- seasonal dynamics
Individual trees modify the below microclimate, resources and associated biodiversity . Examples of trees modifying the below microclimate include trees serve as wind shelters , canopies intercept rainfall  and solar radiation [2, 4, 5, 6] resulting in calm drier and warmer below microclimate in comparison to nearby open areas. Through leaf fall, trees influence belowground microclimate and litter quantity and quality, for example dark litter warms more than light-colored litter , deciduous species show a pulse in litter production  and litter C: N and C:P ratios directly influence soil N and P  and N transformation rates . In addition, physicochemical changes that happen during litter decomposition  are paralleled in the vertical stratification of the organic matter (OM) . As a consequence, trees produce spatial variation in microenvironments and a patchy distribution of litter that is vertically stratified into progressive decomposition stages .
The distribution of trees results in a patchy distribution of litter and associated organisms [14, 15, 16]. For example, microarthropods were more abundant in aspen than in pine  earthworms were abundant under
With the overwhelming impact of human activities on biodiversity [25, 26], an enormous amount of studies (e.g. [25, 27, 28, 29, 30] are only few examples) have addressed the relationship between diversity and ecosystem processes at the stand and local scales, but few happen at the individual tree species scale [23, 24, 31]. Therefore, we evaluated how aboveground diversity, specifically individual tree species, modify microclimate and litter, and the relation to the dynamics and diversity of belowground arthropods in the vertical organic matter profile. For this, we selected isolated individual trees, such as those found in the Guanica dwarf forest, that provide the ideal setting to study tree species in complete isolation .
2.1 Study site
The study was conducted in the Guánica dry forest, a Biosphere Reserve established in 80’s (for the specific location, please see
Mesoclimate was characterized by using data obtained from the NRCS  and RAWS sites . The first site is a SCAN (Soil Climate Analysis Network) weather station that is located in an open area in the forest and up 165 m from the coastal plateau while the second site is a RAWS (Remote Automated Weather Stations) weather station located near the coast but the data for the study period is too fragmentary. Temperature and precipitation were taken from the SCAN site, except for October 2005, that is missing and was then was calculated by difference between the previous and next month. This datum was corroborated with the RAWS site. This information is presented in Figure 1.
Microclimate data was obtained from data loggers. For each of the study species, three (out of five) trees were selected to ensure trees were interspersed within the study area and therefore represented the local variation. Data loggers were placed under each tree as follows: one HOBO temperature/humidity data logger was placed at 25 cm above ground, and will be referred as understory temperature throughout the chapter, one TidBit temperature data logger was placed in the old litter fraction (∼2.5 cm depth) and one TidBit temperature data logger was placed in the humus fraction (∼4.5 cm depth). Data loggers collected temperature data every hour from September 2004 to November 2005. Although the HOBO data logger collected temperature and humidity, only temperature is presented because the humidity sensor shorted out as it was exposed to a salty environment and produced unreliable data.
2.4 Tree species characterization
Each tree was characterized by measuring tree height, canopy area, organic matter dry mass and depth. Tree height was measured as the distance between the ground and the highest canopy point in vertical orientation, and canopy length and width were measured in horizontal orientation. Canopy area was calculated from length and width. Organic matter dry mass and fraction depth were measured separately for each tree and fraction in November 2004, February, April, June, September and November 2005.
Arthropod collections were performed under each tree on November 2004, February, April, June, and September 2005. During each sampling, one 10 cm x 10 cm sample/tree/species was collected, and the sample was separated into three fractions: loose litter, old litter and humus. Each fraction was kept separately and placed in a Berlese funnel for one week for arthropod extraction using light . This sampling design gave 5 species x 10 trees x 3 fractions x 5 samplings = 750 samples. Collected arthropods were taxonomically identified to the lowest category possible, either class, subclass, order or suborder, and classified as adult or immature, and assigned to a morphospecies. Collembolans were not assigned to morphotypes since variation in the morphology can only be seen in mounted slides and by a specialist. The abundance of each morphospecies was recorded and standardized to number of individuals per square meter. Morphospecies composition was used as a surrogate for species composition.
2.6 Statistical analyses
Anovas were used to evaluate differences in tree structure and organic matter among plant species. Also, a two-way Anova was used to establish the effect of plant species and fraction on the modulation of temperature. Modulation was estimated as the maximum, minimum and daily range in temperature. To assess the effect of tree structure on temperature modulation we used spearman rank correlations. To assess the effect of time (sampling dates), plant species and fraction on the abundance and richness of arthropods three-way AOV were used. The abundance of arthropod morphotypes was used in a Multi-Response Permutation Procedure (MRPP) to evaluate the effect of time, plant species and fraction on the species composition of adult arthropods. MRPP is a non-parametric test that calculates a distance matrix of average observed distance within predefined groups and compares it to an average distance expected by chance. Within group distance is used to calculate A, a homogeneity parameter of within group variability that ranges between zero and one. In community ecology, values close to zero are common and suggest a heterogenous community that can still be different from other communities [24, 43, 44, 45].
3. Results and discussion
During the study, temperature decreased from September to February, and then it increased again (Figure 1) and total precipitation was 1575 mm, and was distributed as follows: 480 mm from September to October 2004 producing wet conditions at the end of 2004, 120 mm between November 2004 and April 2005 producing dry conditions at the beginning of 2005, and 975 mm between May and October 2005 producing wet conditions in the second half of 2005 that were interrupted by a short water deficit in June 2005 (Figure 1). In general, during the study, the pattern was similar to the one historically established for the Guanica forest [36, 46] with some variations. Historically, the dry season runs from January to July interrupted by a small pulse in precipitation in May, and then follows a wet season from August to December with water surplus during September. This study encompassed two wet periods and one dry period, there was water surplus in September and October 2004, followed by dry months up to April 2005 indicating 6 consecutive months of water deficit. Then, there were wet conditions after April 2005 with three pulses of water surplus in May, July and October 2005. This indicates that during the study, precipitation was atypical because dry conditions lasted 6 consecutive months while Lugo et al.  found that, for this forest, dry conditions usually last 3–4 consecutive months. In addition, historically the wet season usually presents one pulse of water surplus but during wet 2005 there were three pulses in precipitation. These data confirm that precipitation in Guánica is highly erratic  and show that the 2004–2005 months encompassed in this study represent extreme dry conditions followed by extreme wet conditions.
3.2.1 Variation among tree species
There was a significant effect of plant species and fraction on maximum, minimum and temperature range (Table 1, Figure 2). Understory maxima temperatures ranged between 35.9°C and 34.9°C and followed the pattern
|Species x fraction||8||149||<0.001||27||<0.001||142||<0.001|
Throughfall and stem flow are affected by tree size and differentially moisten soil under canopies  while surrounding rock similarly warms the OM underneath trees . In this study temperature range was largest in the litter suggesting that the buffer capacity of humidity was absent in litter (upper OM) and present in humus (deeper OM). On the other hand, among the five species,
3.2.2 Variations among seasons
As expected, we found that the pattern of temperature fluctuations under all species (Figure 2) was similar to the pattern of mesoclimate temperature fluctuations (Figure 1) (Kolmogorov Smirnov, p = 1.00), e.g. lower temperatures in January and February in comparison to the remaining months. Also, understory air temperature under the five species was consistently higher than temperature from the mesoclimate site by 2.13°C on average. This is due to differences in altitude, the mesoclimate SCAN site was located 165 masl while trees were located at 0 masl, and since temperature decreases by ∼1°C for every 100 m in altitude, therefore this resulted in the consistent higher temperatures under the trees. Across months, understory, litter and humus temperature fluctuations were similar for all species, and the largest temperature increase occurred from February to March, an increase of 2.98°C (Figure 2). Our data is consistent with the historical pattern described for this forest  and with other studies that established that different vegetation associations within a geographic area follow similar fluctuations  because fluctuations at the regional scale drive the seasonal pattern at smaller scales, i.e. tree understory temperature.
In all species, through time, daily range significantly varied in all fractions and was larger from December 2004 to March 2005 (Figure 3). In
Therefore, in Guanica, months of water deficit (Figure 1) coincide with cool months suggesting that mesoclimate drives the decrease in rainfall and minima temperatures while plant species modulate maxima temperatures. In addition, temperature fluctuations in the soil are buffered by soil moisture  and isothermal karst rocks warm OM  therefore water deficit  in combination with an increase in temperature better explain the largest daily variations found in March 2005.
3.3 Tree species characterization
Across months, litter fractions varied significantly (Figure 5). For example, litter depth under
3.4 Tree species and microclimate
In summary, we have shown that microclimate fluctuations follow the seasonal pattern of mesoclimate, that maxima and temperature range, tree height, canopy area and litter are different among plant species. Different tree species characteristics correlated significantly with temperature, and these correlations varied seasonally (Table 2). In general, understory air temperature correlated with tree characteristics in the dry season but not in the wet season, while litter and humus temperatures correlated with tree characteristics in both seasons and with litter mass only in the wet season. Canopy size correlated significantly with understory maximum temperature in February 2005 when deciduous species drop the leaves  suggesting that more open canopies allowed for higher maximum temperatures in the dry season. On the other hand, litter mass and depth correlated with litter temperature suggesting that the quantity of organic matter and its water holding capacity is important for buffering litter temperature during the wet season.
|February 2005||September 2005|
Canopy area significantly correlated with temperature, for example maximum temperature was highest in
3.5.1 Litter fractions
A total of 8702 arthropods representing 22 orders and 301 morphotypes were collected. Arthropod abundance and richness were significantly different among fractions (Table 3). Overall
|Date x species||16||0.90||0.57||1.52||0.09|
|Date x fraction||8||2.92||0.00||7.23||<0.001|
|Species x fraction||8||2.59||0.01||1.93||0.05|
|Date x species x fraction||32||0.79||0.79||0.93||0.58|
Hansen  found that large litter dwellers were positively associated with particular litter substrates, Berg et al.  similarly found that litter in different decomposition stages had sets of associated arthropods, and Prinzing et al.  found that even sexes within a mite species differentially used organic matter strata. At the same time, Hansen  proposed that a higher abundance and richness of arthropods would be found in strata with higher diversity of resources (i.e. litter). Our results support that specific groups are associated to specific litter strata, being more abundant where more resources are found, but not in the litter, instead in the humus fraction. Kardol et al.  suggest that plant–soil feedbacks are mediated by microbes; specifically, they found that soil microbial pathogens affected the development of the future plant community because soils have a biotic legacy of past vegetation cover. Similarly, soils may have a biotic legacy of beneficial microbes that, as Wardle states, maximizes decomposition, and can have a positive historical feedback on the soil community. In addition, Bezemer et al.  found that plant–soil feedbacks are dependent on plant species. In our case, these data suggest that trees in the coastal plateau have been there for long enough to result in a biotic legacy that, under certain plant species favors some arthropods, while under other plant species other types of arthropods are favored.
3.5.2 Variations through time
Arthropod abundance and richness were significantly different among sampling dates (Table 3). Through time, arthropod abundance varied with a trend for a decreased abundance in the dry period (April 2005) and higher abundance in the wet period (Figure 7). In all species and all samplings, arthropod abundance was consistently low in loose litter while variations in old litter and humus explained significant differences. For example, in
During the study, precipitation varied producing a wet 2004 period, a dry 2005 period and a wet 2005 period (Figure 1). Herrera  found that the abundance of soil fauna fluctuated seasonally influenced by the precipitation regime. Specifically, groups like ants and beetles were present during year-round but isopods and diplopods were more abundant at the end of the wet season while chilopods were present all the year but during the dry season they migrated vertically and thus were completely absent in the upper litter. On the other hand, Prather et al.  found that high soil moisture promoted arthropods while high temperature decreased arthropods. Our results show that arthropod abundance and richness follow a seasonal pattern tight to precipitation, and that arthropods tightly linked to litter are able to overcome drought although in such low abundance that does not support upper trophic levels .
Arthropod species composition was dynamic through time as it varied among plant species (Figure 9). Under wet conditions such as in November 2004, arthropod species composition was similar between
From these data, we conclude that two major arthropod communities were formed, one with arthropods common to
Both, arthropod abundance and richness (Figures 7 and 8) decreased at the onset of the dry period (February 2005) and were at a minimum in the middle of the period (April 2005), on the other hand species composition was similar at the onset but different in the middle of the dry season (Figure 9). Furthermore, litter depth was high at the onset of the dry season (Figure 5) because of new litter produced by plants as a response of water deficit (Figure 1). These results suggest that during the dry season, conditions are such that plants drop leaves and belowground arthropods associated to plants represent a shared subgroup common to all plant species, while further into the dry season, arthropod abundance and richness are at the lowest such that only arthropods specific to plant species are present. Abundance and richness are not linearly related, nevertheless an increase in abundance results in more species, for example Prather et al.  found that for every 16 individuals a new species appears. From these data we can expect that a decrease in abundance results in loss of species, therefore, in our study the decrease in abundance at the beginning of the dry season may have resulted in a decrease in species such that only stress-tolerant arthropod species  were present in all plant species producing a homogenization of the arthropod community. A further decrease in abundance led to the minimum found in mid-dry season which might have produced further species loss so that only species adapted to the specific plant microhabitat remained resulting in a differentiation in arthropod community.
Within each plant species, arthropod species composition significantly varied through time (Table 4). In
|Species||Nov 04 – Feb 05||Feb – Apr 05||Apr – June 05||June – Sept 05|
For most plant species, these data show a shared pattern for a shift in arthropod communities between wet and dry seasons (November 2004 to February 2005), then arthropods in
3.5.3 Variations across fractions
Abundance and richness of arthropods were significantly affected by sampling time and fraction (interaction term Date x Fraction in Table 3), suggesting that through time there is a dynamic movement of arthropods among fractions (Figures 7 and 8). In addition, for each plant species the composition of arthropods varied across the vertical organic matter profile (Figure 10).
We found abundance and richness to be highest in humus and species composition to be different among fractions suggesting that arthropod communities are segregated among plant species and are further stratified by decomposition stage. Also, abundance and richness were lowest in the dry months when arthropod communities homogenize while in the wet season there was a pattern for arthropods to form two distinctive groups, one formed by arthropods common to
We found for 4 out of 5 plant species, that arthropod communities shift at the end of the wet season when water becomes scarce, and again at the onset of next wet season when temperature daily range was largest. In addition, we identified three patterns of arthropod dynamics across fractions.
This research was partially funded by CREST-Center for Applied Tropical Ecology and Conservation of the University of Puerto Rico at Rio Piedras Campus, grant NSF-HRD-0206200 through a fellowship to MFBA. Additional funding and logistic support was provided by the CREST-Center for Applied Tropical Ecology and Conservation of the University of Puerto Rico at Rio Piedras Campus, USDA Forest Service-International Institute of Tropical Forestry and the Guanica Dry Forest, a Biosphere Reserve staff.