A Rapid Assessment Protocol (RAP) for non-canopy spiders was used to collect replicate samples from four lowland rainforest sites for a proof-of-concept comparison of spider assemblages from the Democratic Republic of the Congo (hereafter, Congo) and Panama. Collecting was done at two 0.25 ha sites in Panama and two 0.25 ha sites in Congo. At each site, three 0.01 ha plots were randomly located, and within each we did 1 person-hour of aerial sampling (sweeping and beating/brushing) and two person-hours of ground sampling (field sieving of leaf litter). The samples yielded 350 adult spiders belonging to 29 spider families. The Panama samples yielded more adult spiders (235 vs. 115) and more spider families (24 vs. 14) than the Congo samples. Overall, the dominant five spider families in these non-canopy samples were Theridiidae (24%), Salticidae (15%), Linyphiidae (11%), Oonopidae (10%), and Pholcidae (7%), with the 20 remaining families each making up less than 5% of the total adults. The three most abundant families in Congo were Theridiidae, Oonopidae, and Thomisidae, while the top three in Panama were Salticidae, Theridiidae, and Linyphiidae. An NMDS ordination analysis of the four plots failed to show significant differences between any of the four assemblages, but when the plots were analyzed by region, there was a significant difference in the family-level assemblages between the continents. This paper shows proof-of-concept that this RAP can produce statistically valid data from brief sampling trips by teams with inexperienced collectors and simple, inexpensive sampling equipment.

Afrotropical, family-level analysis, Neotropical, NMDS, Sampling

Comparing invertebrate faunas from different regions has historically proven to be problematic. Firstly, there is the significant taxonomic barrier of the necessity of knowing two disparate faunas, and the fact that there is often very little species overlap between very distant locations, and what overlap there is may be mainly due to the occurrence of widespread invasive or cosmopolitan species. These problems are largely alleviated by considering faunas at a higher taxonomic level, as many spider families have very broad geographic distributions (Platnick et al. 2020; World Spider Catalog 2024). With spiders (Araneae), recent advances in molecular phylogenetic techniques (Coddington 2017) have increased our confidence that higher level taxa are monophyletic, allowing us to study disparate faunas at family level. The over 52,000 described spider species are currently placed within 134 spider families (World Spider Catalog 2024), and so the presence and abundance of these give us a way to compare the structure of spider assemblages in similar habitats across the planet. This approach is particularly valuable in studying the faunas of tropical rainforests, where the species tend to be poorly known and difficult to determine, and where many specimens within ecological samples may belong to undescribed species.

Even when this simplifying approach is used, our ability to compare faunas is very limited by the different ways in which the spiders are sampled. Whether examining carefully annotated regional species lists, or comparing faunas of particular sites with one another, the area sampled and the way the samples have been obtained can have significant effects on the resulting data. The species-area relationship, for example, is one of the best-known ecological principles (Connor and McCoy 1979), stating that larger areas have a very marked tendency to contain more taxa than small areas, and so two areas of very different size can be expected to vary in diversity even if they are very similar ecologically. The effect of sampling intensity on diversity is also very well known. Regions that have been intensively sampled, yielding vast numbers of individuals, will seemingly be more diverse than undersampled areas that have so far yielded only small numbers of specimens (Magurran 2004). Lastly, the methods used to sample animals obviously have a major effect on the species that are recorded (Samways et al. 2010), and so differences in two faunas could be due to different sizes of areas sampled, different intensity of sampling, and/or different methods of generating samples. If these factors are not kept constant, it makes comparison of even fairly well studied areas very difficult.

One solution to this problem that has emerged among community ecologists is the development of standardized sampling protocols. Several teams (for example, Scharff et al. 2003; Cardoso et al. 2008) have developed very robust sampling protocols to produce statistically comparable samples among sites, but the problem with these protocols generally is that they rely on very intensive sampling to produce a very thorough sampling of the site’s fauna. They tend to use 1 ha as a plot size, which is a very large area to fully sample within a tropical rainforest (Draney personal observations). Unfortunately, most workers in tropical regions lack time and other resources needed to complete these very valuable surveys; they usually lack resources needed to access the canopy; and often it is unsafe to do fieldwork in the locations of interest at night, as recommended by many of these protocols. Therefore, most tropical sites have not (and probably will not) be sampled in this way, and yet environmental degradation of many sorts continues apace, and conservationists are in need of a way to quickly and cheaply obtain statistically sound, comparable data from which to base conservation prioritization and other management decisions.

A solution developed by Coddington et al. (1991) was an attempt to get comparable data from the existing practices of museum-based collectors, who often visit remote sites for short periods and do very intensive collecting. These methods standardize the collecting effort by constraining sampling time, so that one sample would be the result of one collector using one method for one hour, for example. This allows for very quick collection of samples, and these samples are indeed statistically replicable, but the method does not specify sampling intensity (since museum-based collectors tend to collect as much as time will allow), which makes comparisons from different sites problematic. Also, the area sampled is not specified or controlled for, so researchers following the same protocol in very different sized areas may thus still end up with significant differences in diversity among sites that may or may not be due to faunal differences. Some researchers have responded by specifying a very large sample area (1 ha or more; Scharff et al. 2003; Cardoso et al. 2008) such that “complete” sampling isn’t possible, and sampling intensity thus has a big effect on the resulting data. Dobyns (1997) showed that constraining sampling to a smaller plot size allows for more efficient collection of rare species and better estimation of richness on a smaller scale.

We have developed and are beginning to use a Rapid Assessment Protocol (RAP) that attempts to “thread the needle” between these two approaches (time and effort intensive, but thorough; versus efficient, but unconstrained with regard to plot size and sampling intensity), allowing researchers without sophisticated equipment to rapidly collect comparable samples of the ground-accessible portion of terrestrial ecosystems from randomly selected areas. These areas are always the same size and are sampled with the same amount of effort using the same sampling equipment, and so statistical comparison of samples should be possible. This paper represents our first attempt to compare the spider fauna of ecologically similar sites (lowland tropical rainforest) from two distant biogeographic regions, the Afrotropics (represented by two sites in the Democratic Republic of Congo) and the Neotropics (represented by two sites in Panama). The Panama sites were collected by a team of university students from the United States, visiting Panama on a travel course. They were mostly inexperienced collectors in an unfamiliar area, but supervised by an experienced field arachnologist. The Congo sites were also sampled by a team of mostly inexperienced Congolese university students, also supervised by an experienced field arachnologist. Both teams were limited in time (each site was sampled in 1–2 days) and equipment available: They each had to bring sampling equipment with them on arduous trips (transcontinental commercial air travel for the US students, and rented canoes and motorcycles for the Congolese students). In addition to giving us an interesting first comparison of the family-level spider communities of lowland rainforests of these two continents, we hope this paper also demonstrates the potential for using a RAP like ours to quickly obtain useful site-specific spider assemblage data, even with inexperienced and under-resourced research teams.

The RAP Samples from the four rainforest sites yielded a total of 350 adult spiders (Table 1), so an average of about 88 adult spiders per RAP sample set. However, about 2/3 of the adult spiders came from the Panama sites and 1/3 from the Congo sites (235 vs. 115). The Congo sites yielded 50 and 65 adult spiders, and the Panama sites yielded 103 and 132 spiders. From all four sites, spiders from 29 spider families (all Araneomorphae) were collected. The Congo sites yielded a total of 14 spider families (9 from the monotypic site and 14 from the mixed forest site) and the Panama sites yielded a total of 24 spider families (13 from the La Grua site, and 20 from the Rio Frijolitos site). Spiders in the family Theridiidae were most frequent in the samples, making up ~24% of adult spiders. Five families constituted over 2/3 of the adult spiders captured. After Theridiidae, these were Salticidae (~15%), Linyphiidae (~11%), Oonopidae (~10%), and Pholcidae (~7%), with the 24 remaining families comprising about 33.1% of the adults, and all of the remaining families each made up less than 5% of the total adult spiders sampled.

Table 1 also shows the pattern of occurrence of the 29 sampled families in each of the four sampled sites. Five families were found in all four RAP sample sets; these included four of the most abundant families (Theridiidae, Salticidae, Linyphiidae, and Oonopidae) plus one additional family, the Tetragnathidae. Four families were found only in the Congo samples: Thomisidae, Mimetidae, Palpimanidae, and Zodariidae. One family, Thomisidae, was found at both Congo sites, but not at Panama sites. Fifteen spider families were found within the Panama RAP samples, but not the Congo samples (see Table 1). Of these, two families (Theridiosomatidae and Deinopidae) were found at both Panama sites but not in the Congo samples.

As shown in Table 1, Theridiidae dominated Congo samples, with over 36% of all adult spiders; second most abundant in these samples was Oonopidae at ~14%. Interestingly, the third most abundant spider family in the Congo samples was Thomisidae (10.4%), a family absent from the Panama samples. The top five families were the same as for the pooled RAP rainforest data (Table 1), but in Panama, Salticidae were the most abundant adult spiders in the samples, making up 18.7% of the total. Theridiidae were second (~18%), followed by Linyphiidae (12.3%).

In Table 2, we show the iChao1 estimates and 95% confidence intervals for family-level richness at each of the sampled sites (all RAP data pooled). These results show there is no overlap between either the lower or upper confidence bounds of these estimates, implying higher family-level richness in Panama than in Congo. Note that the lower bound estimates are more accurate than the upper bounds (Chiu et al. 2014), and there is only a modest difference in them (Congo sites varied from 9–12, and Panama sites from 14–21 estimated families). The upper and lower confidence intervals together produced only a small overlap between the less family-rich Congo sites and the richer Panama sites.

Fig. 9 shows the results of the NMDS ordination (of adult spider distribution among spider families) of the four RAP sites; in other words, the ordination of spider assemblage composition. The figure shows that plots within a site always clustered near each other, and the plots within continents likewise ended up near each other. Note that the low number of replicates per site (three) precluded the production of 95% confidence interval ellipses surrounding the data from each site. The overall analysis was found to be significant (p < 0.0001), indicating significant differences among sites in total. However, a pairwise comparison of the sites revealed that no site was significantly different from any other site; p values were all equal to or greater than 0.1.

Figure 9. 

NMDS plot of among-site comparison of each of four sites separately. The two Panama sites are La Grua and Rio Frijolitos, and the two DR Congo sites are Yoko G. dewevrei monotypic and Yoko mixed species forest. Each data point represents distribution of adult spiders by family in one of three replicate 0.01 ha plots from each site, with ground and aerial data combined. The overall analysis was found to be significant (p < 0.0001), but no site was significantly different from any other site via pairwise comparisons (p = 0.1 was the lowest significance). A low number of replicates precludes the creation of 95% confidence interval ellipses.

Because the sites were not found to be significantly different in this analysis, we pooled the data from each continent to examine the family-level composition of lowland rainforest spider assemblages from Congo versus Panama (Fig. 10). Note that this two-region comparison involved six replicate plots per region, enough to construct 95% confidence interval ellipses around the data from each region. The family-level spider assemblages of the two regions were found to be significantly different from each other (P = 0.0021). Fig. 10 shows that the plots within each region clustered in fairly linear sequences in the ordination space, with the Congo plots spread principally along the X axis and the Panama plots spread in a tight sequence along the Y axis.

Figure 10. 

NMDS Comparison of plots from two biogeographic regions (Panama, Congo). Each region is represented by data from six replicate 0.01 ha plots. Each replicate consists of the pooled aerial + ground data, representing distribution of the adult spiders by family. The two regions were significantly different from each other (p = 0.0021) with little overlap in the predicted data ellipses, which represent 95% confidence intervals.

Table 3 shows the number of adult spiders and distinct spider families recovered from the aerial (sweepnet, beating sheet) and the ground (litter sieving) efforts in each plot of the dataset. The total Congo and Panama data show that both aerial and ground sampling yielded more spiders and families in the Panama sites than the same methods in the Congo sites. There may be a difference in effectiveness between the two regions, since both methods yielded similar spider numbers and families in Panama, yet in Congo the aerial samples yielded ~4 times more individuals and ~3 times more families than the ground samples did. The method totals (all sites pooled) showed that 1 person hour aerial samples yielded slightly more adult spiders than the ground samples (184 from aerial vs. 151 from ground), and that both recovered a similar number of spider families (20 from aerial and 21 from ground).

This study suggests that the ground-accessible spider assemblages of lowland rainforests from Congo and Panama may vary in several important aspects. The two Panama spider sites yielded more adult spiders, and more families of spiders were represented. Moreover, the family level structure of the assemblages (the families present and distribution of individuals among those families) was statistically different between the Panama and the Congo sites. Whether these differences are biologically significant, and whether they say something about the history and ecology of the respective regions, depends on the reasons for the differences of these assemblages.

The RAP controls for many variables that often cloud our ability to interpret differences between sites. The size of the sampled site, the size of the plots where the samples were collected, and the number of plots sampled were the same for all compared sample sets, and the sampling intensity was exactly equal at three person-hours per plot, with 2/3 of the time devoted to sampling the ground layer and 1/3 of the time devoted to sampling the vegetation layer which is inherently easier and more efficient to sample, and yields more spiders and taxa per hour of sampling. Yet, the litter samples seem to access different taxa from the aerial samples, and so contribute significantly to the total numbers (M. Draney, personal observations). Table 3 shows that this built-in bias toward litter sampling was justified, in that twice as much litter sampling yielded a similar number of adult spiders (184 from aerial samples, 151 from ground samples), and a similar family level richness (20 families recovered from aerial samples, 21 families from ground samples). It seems that spending twice as much time on litter sieving may indeed optimize the information gained per unit of sampling time.

Of the between-site differences observed, adult spider abundance is likely the most variable and the least likely to be due to ecological differences between the sites. Although all four sites were sampled in relatively dry seasons, they were collected in different months, and so seasonality can impact the catch. Although all samples were collected when no rain was falling, the weather during the days and weeks before sampling can alter spider abundance. Spiders are more likely to be actively foraging during humid periods, for example (Draney personal observations), and this can impact the number of spiders encountered. We do not consider the differences in abundance to be strong evidence of actual differences in spider density between the regions, although that hypothesis remains to be tested.

Family richness, the second variable examined in this study, seems less prone to the vagaries of weather than abundance, but here the problem is the well-documented relationship between taxon richness and sample size (Magurran 2004). Both Panama sites yielded more adult spiders and more spider families than both Congo sites, but the higher number of individuals collected from Panama sites could at least partly account for the difference in family richness. Still, the large difference in Family-level richness (nearly twice as many families in Panama samples: 24 versus 14) did surprise us, and these data yielded iChao1 richness estimates that showed little overlap, giving additional support to the hypothesis that the Panama sites have higher family-level richness.

The composition of the spider assemblages (the identity of sampled families) is likely to reflect real differences among the sites, at least when comparing the most commonly encountered families. The greater dominance of Theridiidae and Thomisidae in Congo sites, and of Salticidae in Panama sites, is hard to explain as reflecting anything other than real ecological differences between the sites. However, it has to be kept in mind that our relatively small sample size (350 adult spiders among four sites) means that interpreting less frequently collected taxa is inherently problematic. Many spider families that are important parts of rainforest spider assemblages were not recovered in our samples. For example, Nentwig et al. (1993) lists Panamanian spiders from 33 families from habitats very similar to our sampling locations, so our samples failed to turn up representatives from at least 9 families, including all the Mygalomorph taxa. Likewise, a pitfall study of the Masako Forest Reserve near Kisangani (Juakaly and Jocqué 2021) yielded 918 adult spiders from 18 families from young and old successional forests. This includes 11 families not found in our Yoko samples (although our Yoko RAP samples turned up 7 families not recovered in the pitfall study). Many of the spider families missing from sites in our study undoubtedly occur there but were not collected due to relative rarity, or to undersampling of microhabitats such as the forest canopy, or Mygalomorph burrows. Five spider families in this study are represented by single individuals (singletons; Table 1), and it might have been just as likely to have collected zero individuals of these families as one. Most of the spider families in our study have very wide distributions and have been recorded in both regions. Of the 29 spider families sampled in this study, only two are likely not found in both regions (Central Africa and Central America): Anyphaenidae and Mysmenidae (neither recorded from D.R. Congo; Mysmenidae known from one species from Rwanda: Dippenaar-Schoeman and Jocqué 1997; World Spider Catalog 2024).

The aspect of our data that is hardest to explain away and is most likely to reflect actual differences among the spider assemblages is the distribution of spiders among the sampled families, which was found to be significantly different between the two regions in our NMDS pooled analysis. The most dominant taxa in an assemblage have the biggest effect on the placement of plots in the NMDS ordination space, and proportional representation of dominant taxa is not likely to be much affected by sample size or any of the other factors (weather, etc.) discussed above. Thus, one result of our study that we take seriously is that the family-level spider distribution of Central African rainforest sites is considerably different in composition to that of Central American sites. The differences in the composition of the most abundant families (Table 1) gives an idea of the specific differences, which remain to be clearly delineated and explained by further research. As far as we know, this is the first time that assemblages (as opposed to national or regional species lists) of African and Central American spiders have been compared, and standardized assessment protocols such as the one used in this paper make such comparisons possible. Clearly, this study is too small in scope to make a definitive comparison of lowland tropical rainforests of the two regions. Many more sites would need to be included in a more definitive analysis, and ideally more would be known about the vegetation and history of the sites. For rainforest sites, information on the spiders of the canopy would ultimately need to be included; they are excluded from our RAP because of the difficulty and cost of accessing them. This preliminary study serves as a proof of concept that a carefully thought-out Rapid Assessment Protocol can enable statistically rigorous comparisons of spider assemblages using data collected in a short period of time by teams with inexperienced collectors using inexpensive supplies and equipment.

The authors have many happy memories of interacting with Dr Stefan H. Foord at the 2016 International Congress of Arachnology in Golden, CO. He is very much missed among our group interested in diversity and ecology of ground-dwelling spiders.

The Panama samples were collected during a Panama Travel Course trip in 2007–2008. Collectors included M. Draney, A. Choudhury, B. Howe, V. Medland, A. Wolf and a number of students from UW-Green Bay and from St. Norbert College, De Pere, Wisconsin: B. Butterfield, K. Corio, J. Goyette, B. Grasse, M. Harvey, C. Osman, A. Rick, J. Sundance, J. Watson, and C. Wepking. Andrew McKenna-Foster helped Draney to initially develop the Rapid Assessment Protocol. We thank Raineldo Urriola, Adriana Bilgray, Lil Camacho and the rest of the logistics support team at Smithsonian Tropical Research Institute-Panama for logistical help with the trip and sampling, and R. Shuman Baquiran for help in importing the specimens to the USA for study. The Panama material was collected under permit SE/A-113-07 to Michael Draney, and exported out of Panama under scientific permit No. SEX/A-2-08 to Michael Draney. We thank the Field Museum of Natural History, Chicago, and especially museum volunteer James Case and Collections Assistant James Louderman, for initial family-level spider identification.

The Congo samples were collected by Mbumba Jean-Louis Juakaly and his students, mainly André Lofanga, Pascal Baelo, Steve Ngoie, Sylvie Kasienene, John Ngbangi, Vinny Kitenge, Jules Esongolo, Héritier Golo, and Drangozo. The Congo sampling and shipping were supported by a 2014 UW-Green Bay Grant in Aid of Research to Draney and Juakaly, and a grant from the Field Museum Africa Council to Sierwald and Juakaly, which also supported a trip to the FMNH for Dr Juakaly. The Congo material was exported out of DR Congo under Attestation de Transport de Materiel Scientifique No FS/VDR/M8/2014, FS/VDR/M6/2014, and FS/VDR/o4/2015 to Mbumba Jean-Louis Juakaly.

We thank Subject Editor Caswell Munyai, reviewer Charles Haddad, and an anonymous reviewer, whose comments greatly improved this article.