RESULTS

There are many ways to use data from controlled surveys to explore patterns across space or through time in the fossil assemblages themselves, or to test hypotheses concerning relationships of paleontological trends to geological or geochemical evidence for paleoenvironmental characteristics of the ancient landscapes and faunas. The examples below are primarily exploratory in nature and are used to demonstrate the potential benefits of standardized sampling. The results of these initial analyses raise many questions that can be pursued in subsequent studies.

In the following sections, we use tallies and proportions of the basic data collected for the biostratigraphic surveys, which consist of identified specimens and tallies of turtle fragments and unidentifiable scraps. Because specimens that were found in several or many recently broken fragments were counted as one, the total number of specimens should be a good approximation of the actual number of separate fossils for each survey block or interval. We refer to this as NISP (number of identifiable specimens; Badgley 1986a). The sample that was identifiable to taxon is NISP_V (identifiable at least to major vertebrate class) or NISP_F (for family or order), and the sample identifiable to skeletal region is NISP_SK, or NISP_SKM for mammals only. In both cases, NISP probably is fairly close to MNI, minimum number of individuals, or MNE, minimum number of elements, respectively, given the wide dispersal of most of the specimens across the surveyed outcrops (Badgley 1986a). However, we retain NISP as our basic unit of analysis since we cannot test for MNI and MNE using data recorded on the survey cards. A total of 121 survey blocks combined into 24 separate numbered surveys (levels) were used in this analysis.

Fossil Productivity

The number of fossil bones that were identified at least to major vertebrate class (mammal, reptile, fish, bird) and/or to skeletal element provides the basic data used for analysis of overall fossil productivity. This combines the numbered specimens on the survey cards and the “turtle tally,” which was used as a quick way to keep track of small fragments of fossil turtle shell. The number of identified specimens (NISP_V) divided by the total number of search hours for each survey level (i.e., the total for all surveyors who searched that level) gives a standardized measure of its fossil productivity (P_f; Table 1, Figure 6A), with the mean value for all survey levels of about 10 identifiable fossils per hour. Alternatively, we could have used the area of outcrop covered in each survey to standardize search effort; this was recorded on air photographs, but digitized information for outcrop area is not yet available.

We can make the assumption that the NISP_V /Hour (P_f) accurately represents the underlying fossil productivity of each interval, but other variables may also affect the pattern of temporal variation in productivity shown in Figure 6A. One of these is the thickness (duration) of the stratigraphic interval being surveyed, which was variable depending on search conditions. We tended to range vertically through thicker intervals for surveys that were relatively unproductive but followed productive strata laterally as far as possible, typically remaining within a relatively thin stratigraphic interval. Dividing P_f by interval duration gives a measure of productivity per 100 kyr (Table 1, Figure 6B), which highlights the narrow zone of exceptionally high productivity at KL01 and also the marked drop-off in productivity upward in the sequence, after 7.6 Ma.

Unidentifiable scrap was tallied for each survey interval, partly as a measure of the preservational state of surface fossils, and partly to encourage surveyors to pick up and examine every fossil they encountered. Not surprisingly, there is a high correlation between the number of identifiable bones (NISP_V) and scrap (Figure 7, Table 1B). The ratio is remarkably consistent throughout the survey samples, and we assume that this reflects a combination of taphonomic processes operating prior to deposition as well as fragmentation on the modern outcrop surfaces. In future analyses it should be possible to test the role of modern outcrop topography on the proportion of unidentifiable scrap using notes on the terrain and photographs for each of the surveys. On average, for the portion of the Siwalik sequence sampled using the surveys, one can expect to find a minimally identifiable fossil for every 1.6 unidentifiable scraps, and a mammal specimen identifiable at least to family for every 5 unidentifiable scraps. This metric is a good indicator of the abundance of information for higher taxonomic levels that is available in the eroded surface fossil assemblages of this fluvial sequence. The proportion of museum-quality, collectible specimens found on these surveys is much lower, compared to the high-density patches that constitute formal localities.

Fossil productivity (P_f) based on biostratigraphic survey data can be compared with productivity based on number of localities for approximately the same intervals (Figure 8, Table 2). The regression coefficient is positive but insignificant (R² = 0.11), and when the two obvious outliers are removed, it is also insignificant (R² = 0.35). The productivity of a biostratigraphic survey thus is not a good predictor of whether the interval will have rich concentrations of fossils, indicating a partial disconnect between the presence of such patches and the scatter of vertebrate remains between them. Interestingly, this suggests some degree of continuity through time in the background of isolated fossil vertebrate occurrences in the Siwalik deposits, contrasting with strong fluvial and/or taphonomic controls on the presence or absence of notable bone concentrations.

Skeletal Parts

The proportions of different skeletal parts in a fossil assemblage can be used to infer the impact of taphonomic processes such as transport and density-dependent destruction on the remains prior to burial (Voorhies 1969; Behrensmeyer 1991). The biostratigraphic surveys provide standardized data for examining patterns through time in the representation of different skeletal elements. Here we focus on teeth and axial elements (vertebrae plus ribs), which represent the most and least dense elements in a complete skeleton, respectively (Voorhies 1969; Behrensmeyer 1975, 1988). Teeth are generally regarded as the most “preservable” elements in the vertebrate body, based on their density and mineralogy, which are particularly resistant to chemical and biological break-down. Teeth average 37% and axial parts 19% of the total sample of 1282 mammalian records identifiable to body part (Figure 9, Table 3), whereas they are 27% and 39%, respectively (excluding caudal vertebrae), in the skeleton of a living ungulate (combination of bovid and equid). Relative to this standard, the Siwalik fossil assemblage is shifted toward the denser, more preservable (and identifiable) elements. However, variations through time show that some survey intervals preserved a much higher proportion of teeth than others, suggesting differences in the taphonomic filter(s) that controlled the preservation of vertebrae and ribs versus teeth.

Based on studies in modern ecosystems and laboratories, non-random variations through time in axial versus tooth frequencies shown in Figure 9 could result from changes in: 1) levels of pre-burial biotic processing of skeletons, i.e., carnivore and scavenger pressure (Behrensmeyer 1993, 2002); 2) degrees of fluvial reworking of the original bone assemblages, with increased reworking resulting in proportionately fewer axial elements (Voorhies 1969; Behrensmeyer 1991); 3) contributions of channel versus floodplain deposits to the surface fossil assemblages recorded in the biostratigraphic surveys; more durable body parts, especially teeth, would be expected if channel deposits are the primary source of the fossils for any given level. In the biostratigraphic survey data, teeth are consistently dominant through the sequence, except for three intervals where they drop close to a 1:1 ratio relative to axial elements. There is an unusual dominance of teeth at about 8.8 Ma (survey ML06), followed by a drop to an unusually low proportion at 8.7 Ma (ML05). Both of these extremes are in the Malhuwala Kas area, ~15 km southwest of Kaulial and Ratha Kas where most of the surveys were done. It is possible that variations in search conditions or original position on the alluvial plain contribute to the differences in the ML samples. If we ignore these two points, the ratio in Figure 9B shows a slight trend toward increased tooth dominance upward in time, which corresponds to the sedimentological shift toward more mountain-proximal (buff), higher energy fluvial systems in the Dhok Pathan Fm. of the Kaulial Kas section (Behrensmeyer and Tauxe 1982). This suggests that the overall tooth versus axial pattern reflects degree of fluvial reworking rather than other possible causes listed above, but further work is needed to test this hypothesis.

Major Vertebrate Groups

Most paleontological collecting efforts focus on one vertebrate class or size category (e.g., macro-mammals) and pay less attention to associated fossils from other groups such as fish and reptiles. Therefore, the proportions of major vertebrate groups in most catalogued inventories are biased by collecting practices and cannot be used to examine the proportions of these groups in the source assemblages. Such information can be important, however, for instance as a general indicator of aquatic versus terrestrial habitats in fossil-preserving environments and overall taphonomic (and potentially ecological) dominance of the different types of vertebrates. Standardized sampling also provides a means of examining and comparing these variables at different times and places in the vertebrate record.

In the Siwalik biostratigraphic survey data, mammal remains average 71% and reptiles 28% of the recorded sample, whereas fish are very rare (0.4%; Figure 10, Table 1, and Table 4). The near absence of fish is unexpected, since many of the depositional environments were clearly aquatic and occasional beds of abundant fish remains occur throughout the sequence. It is probable that this pattern represents a taphonomic bias against the preservation of fish remains in the Siwalik fluvial system. Apparently there were few robust forms, such as armored catfish, whose remains would likely survive as fossils and also be recognized on the biostratigraphic surveys. Of the documented reptilian remains, 88% are chelonian, 8% crocodyloid, and the remainder snake, lizard, and unidentifiable reptile. Most of the chelonian remains are ornamented shell fragments from the family Trionychidae, which are the common “soft-shelled” aquatic turtles, but tortoise and other smooth-shelled fragments also occur.

The relative abundance of reptiles versus mammals through time (Figure 10) shows an initial decline from RH02 through KL04, which coincides with the transition from the channel-dominated blue-gray fluvial system of the Nagri Fm. to the more floodplain-dominated buff fluvial system of the Dhok Pathan Fm. (Behrensmeyer and Tauxe 1982; Barry et al. 2002). The anomalous peak in reptile versus mammal in KL03 is followed by a fairly constant reptile abundance of around 20%. In both RK02 and KL03, the high relative abundance of reptiles is accompanied by fish remains, suggesting that these two levels sample more aquatic environments than the other levels, and also that the decline in the reptiles in the early part of the sequence reflects a shift to less aquatic conditions in the source deposits of the surface fossil assumblages.

Only one bird was recorded in the entire biostratigraphic survey sample (on ML06). Since it is unlikely that we would have missed many identifiable avian remains in the nearly 5000 bones examined during the surveys, this indicates a strong taphonomic bias against the preservation of such remains in the Siwalik fluvial system.

Equidae versus Bovidae

An initial motivation for doing biostratigraphic surveys was to increase the temporal resolution on important biostratigraphic events, such as the appearance of “Hipparion” and the shift from equid to bovid dominance through the Siwalik sequence. Biostratigraphic surveys in the northern Potwar Plateau record the regional “Hipparion” appearance datum as shown in Figure 11. About two-thirds of the remains consist of teeth or tooth fragments (Table 5), which should be similar in terms of the impact of fluvial processes on their taphonomic histories. These remains also should be equally identifiable to family. Equid molars are generally larger than bovid molars, however, thus their abundance may be somewhat inflated in the preserved remains and recorded samples. Overall, however, we regard the survey data for equids and bovids as more or less isotaphonomic. Any biases in relative abundance should be equivalent from level to level, and changes through time are likely to reflect underlying ecological shifts in the diversity and/or abundance of these two groups.

The biostratigraphic survey data begin close to the “Hipparion” datum. There is an estimated 85 kyr between RK01, which has no equid (i.e., species of the genus “Hipparion”) specimens, and RK02+DH01+DH02 with five equid specimens. Biostratigraphic surveys in other regions plus the locality data provide further support for a first appearance datum (FAD) at 10.3 Ma (Barry et al. 2002). Based on their frequency in the sample identifiable to mammalian family, the abundance of equid remains rises while bovid abundance falls sharply between 10.3 and 9.8 Ma. Equids continue to dominate the mammalian macro-fauna until shortly before 8.5 Ma (Figure 11A, Figure 12A), when bovids become more abundant. The ratio of equids to bovids shows that equids reached their peak relative to bovids between 9.5 and 9.0 Ma (Figure 12A). The same overall pattern is preserved in the teeth-only analysis (Figure 11B, Figure 12B), except that the two lines are farther apart and equids are more common than bovids until 7.7 Ma. We suggest that this results primarily from higher survival and visibility of equid teeth on outcrop surface. Using all documented skeletal remains (primarily appendicular) helps to boost tallies of bovids relative to equids, perhaps because of more equivalent survival and visibility levels for these post-cranial parts. Our working hypothesis, therefore, is that the differences between Figure 11A-Figure 12A and Figure 11B-Figure 12B are a measure of durability and collecting bias between these two families rather than a pre-burial taphonomic or ecological signal.

The overall pattern through time in Equidae versus Bovidae, plus some of the shorter-term fluctuations in the sampled abundances not related to teeth versus all identifiable parts, may indicate shifts in the ecology of the alluvial plain favoring greater original abundance of one or the other. There is no obvious environmental event at the “Hipparion” appearance datum, and Barry et al. (2002) suggest that the faunal turnover at around that time reflects biotic processes (e.g., competition). Our data support this hypothesis, because partial competitive exclusion could explain the reciprocal relationship of bovids versus equids shortly after 10.3 Ma, as well as the low levels of bovid abundance for several million years thereafter. The switch in abundance around 8.5 Ma also is not closely correlated with environmental change, although there is evidence that patches of C₄ vegetation may have been present at this time (Morgan 1994; Barry et al. 2002). Turnover events at 7.8 Ma and 7.3-7.0 Ma, which are based on the overall Siwalik faunal record and linked to environmental changes, are not obviously correlated with the equid versus bovid trends in Figure 11-Figure 12.

Mammalian Families

Eight major groups of macro-mammals dominate the Siwalik paleocommunity. Their relative abundances in the biostratigraphic survey sample are represented in Figure 13 (Table 6 and Table 7). Although most of these taxa have been identified based on teeth, they are not necessarily as isotaphonomic as bovids and equids. For example, a single proboscidean or rhinoceros tooth can produce a large number of identifiable fragments, especially compared with smaller artiodactyl teeth. Thus, the proportions of the different groups in the survey samples are not a fair representation of their original relative abundances. As in the case of the plots of equid versus bovid abundance, however, these biases should be relatively constant from interval to interval. The rare mammalian taxa found on the surveys include aardvark, primate, chalicothere, carnivore (including hyena) and rodent, which are grouped as “other” in Figure 13.

Overall there is moderate consistency in the proportions of the eight major groups, and nearly all continue in the paleocommunity through a time span of 3 Ma. Giraffes disappear from the sample between 8.0 and 7.7 Ma, and equids become dominant, mostly at the expense of bovids, shortly after their appearance (see also Figure 11). There is an interesting peak in giraffe abundance at 9.3 Ma (KL03), which coincides with the period of maximum equid dominance, as well as unusual numbers of turtles (Figure 11, Table 1). There also are a large number of fossil localities at this level (Table 2), including the Sivapithecus face site. This suggests that KL03 had somewhat different fluvial conditions and perhaps less seasonally dry habitats than other intervals. Another intriguing pattern is the increase of tragulids and suids in the youngest intervals (after 7.9 Ma), coinciding with the decline of giraffes and equids. Stable isotopes indicate an important transition toward more intensely monsoonal climate and C₄ vegetation starting around 7.3 Ma (Quade et al. 1989), and tragulid extinctions were part of the major faunal turnover between 7.3 and 7.0 Ma (Barry et al. 2002). It is interesting that shortly before that time, tragulids were still prominent members of the Siwalik paleocommunity.