Integration of individual-based and population-level studies is essential to understanding effects of pollution on populations and ecosystems. Here we provide an example of such integration from our exploration of effects of the Exxon Valdez oil spill (EVOS) on river otters (Lontra canadensis) inhabiting the terrestrial-marine interface in Prince William Sound, Alaska, USA. Our research was divided into 2 phases: an early phase (1989-92) immediately following the oil spill; and a late phase (1996-99), which focused on potential chronic effects of oil contamination in the Sound. We used a variety of measurements that considered the physiological status and health of individual river otters, as well as aspects of their ecology, behavior, and demography. We then conducted meta-analysis to explore interactions between individual-based and population-level data in demonstrating injury and subsequent recovery of otters from ill effects of EVOS. During both phases of our studies, we first conducted intensive research at 2 study sites, which we believed to be oiled and nonoiled, and then expanded our investigations throughout similar areas of Prince William Sound. Nonetheless, our data are best interpreted as differences between heavily oiled areas and lightly oiled sites because later information indicated that our reference sites were lightly oiled. Thus we refer to heavily oiled sites as oiled and lightly oiled sites as "nonoiled." In the later phase, we were part of a broader ecosystem-based project (Nearshore Vertebrate Predators) designed to assess long-term effects of EVOS on a suite of key organisms, and to determine whether those species had recovered from that catastrophic accident. We used radiotelemetry to locate carcasses of animals that died from natural causes, and documented that searching beaches immediately following the spill was not a reliable method for locating dead river otters. Our early research (1989-92) demonstrated that river otters living in oiled areas had lower body mass (P < 0.04) and elevated biomarkers (P < 0.05) in their blood (e.g., haptoglobin [Hp], interleukin-6 immunoreactive [IL-6 ir], aspartate aminotransferase [AST]) than otters inhabiting "nonoiled" areas. Likewise, otters from oiled areas had higher levels of fecal porphyrins (P < 0.001), ate a less-diverse diet (P < 0.001), had larger home ranges (P < 0.05), and selected habitats differently (P < 0.01) than otters living in areas that were not heavily oiled. A mark-recapture analysis based on radiotracers in otter feces during 1990 indicated no difference (P > 0.10) between density of otters in Herring Bay (oiled) or Esther Passage ("nonoiled"), but no prespill data were available. Likewise, by 1992, biomarkers (Hp, IL-6 ir, AST) did not differ (P > 0.05) between oiled and "nonoiled" areas. During the later phase of research, hydrocarbons on the pelage of river otters and the elevation of endothelial P450-1A, a biomarker sensitive to hydrocarbon exposure, indicated that river otters were exposed to oil still present in Prince William Sound. Nonetheless, body mass of otters continued to increase on oiled areas over tune (P < 0.05), and eventually did not differ from otters living in "nonoiled" sites (P > 0.05). All blood biomarkers (Hp, IL-6 ir, AST) were markedly reduced from the early phase of our research, and no longer differed (P > 0.10) between oiled and "nonoiled" sites. We used principal component analysis (PCA) to determine that few differences existed in an array of blood characteristics for otters inhabiting oiled and "nonoiled" sites, and those differences that did exist likely were related to diet. Corpoporphyrin III, a key biomarker in home synthesis, was reduced (P = 0.008) from post-spill collections made in 1990 in the oiled area, and no longer differed (P > 0.05) between oiled and "nonoiled" areas in 1996. We used stable isotope analysis to investigate differences in diet of river otters inhabiting oiled and "nonoiled" areas in 1996-97. When we controlled for otters inhabiting extensive freshwater habitats (which did not occur in our early studies), no differences in diet or the trophic level of otters were identified (P > 0.20) for otters living in oiled versus "nonoiled" sites. Similarly, density of marine fishes (greater than or equal to8 con in total length) on underwater transects did not differ (P = 0.97) between oiled and "nonoiled" areas, although an area by year interaction occurred (P = 0.01). Habitat selection by otters also was altered from the early phase; river otters on both study areas selected vegetated slopes that were not steep, and selected sites with more understory (brush) and greater exposure; selection for those characteristics was more pronounced in the oiled area. Otters on both sites avoided (use < availability) gravel and small rocks. Although selected variables differed between oiled and "nonoiled" sites (P < 0.001), the direction of selection did not differ between areas. Moreover, tidal slope did not enter any of the models, in contrast to our early studies, indicating that differences in selection were not related to avoidance of oiled shores. Hone-range size declined (P < 0.05) for otters living in oiled areas, and no longer differed (P > 0.7) from animals inhabiting "nonoiled" sites. We enumerated populations from oiled and "nonoiled" areas using a combination of live-captured individuals and DNA fingerprinting using microsatellite from otter feces at latrines. We also performed a conventional reconstruction based on age structure to calculate population size in 1997. Those methods indicated that most animals in the population were recruited following the oil spill and both methods characterized slowly (lambda = 1.03-1.06) growing or stable population in the oiled area. Age structure of river otters in the Sound differed neither between oiled and "nonoiled" areas (P > 0.36), nor from a harvested population of river otters in Maine (P > 0.49). Finally, survivorship of river otters did not differ (P > 0.2) between oiled and "nonoiled" areas of Prince William Sound and was high compared with data on other otter populations in North America. Our data indicate that although river otters continued to be exposed to low levels of crude oil, effects of that exposure were no longer sufficient to cause obvious injury. We cautiously conclude that river otters have recovered from the more pernicious effects of EVOS. Based on our experiences in this research, we provide theoretical considerations for use of biomarkers in wildlife studies and describe statistical approaches, including principal component analysis of blood variables, which may assist researchers with interpreting complicated results of multiple variables and datasets. Likewise, we describe how dose-response curves should be used in understanding population-level responses to pollutants. We hope that this monograph will provide valuable insights for other wildlife biologists on the process of integration of toxicological data with that of ecological data useful for studying effects of pollution on wildlife populations and their habitats.
