In 1986 more than 8 million liters of crude oil spilled into a complex region of mangroves, seagrasses, and coral reefs just east of the Caribbean entrance to the Panama Canal. This was the largest recorded spill into coastal habitats in the tropical Americas. Many populations of plants and animals in both oiled and unoiled sites had been studied previously, thereby providing an unprecedented measure of ecological variation before the spill. Documentation of the spread of oil and its biological effects begun immediately. Intertidal mangroves, seagrasses, algae, and associated invertebrates were covered by oil and died soon after. More surprisingly, there was also extensive mortality of shallow subtidal reef corals and infauna of seagrass beds. After 1.5 5 years only some organisms in areas exposed to the open sea have recovered.

Our study is unique for two reasons. First, many of the habitats damaged by oil have been investigated since the Witwater spill 18 years before. Results of this research have been published in more than 130 articles, and the flora and fauna are well known taxonomically (7). Detailed time-series data on the physical environment and biota (8, 9) provide a measure of natural variation before the more recent oil spill against which subsequent events can be compared. Second, observations of the effects of the spill began as oil was coming ashore. Of additional importance, the coastal environments (including seagrass beds, mangroves, algal flats, and coral reefs) and most of the species affected are similar to those throughout the

O

IL POLLUTION IN THE SEA HAS BEEN A MAJOR ENVIRON

mental problem for several decades, but we know remarkably little about the effects of oil on natural populations and communities (1). Uptake of oil and physiological responses of organisms have been investigated in the laboratory or field, sometimes coupled with short-term monitoring of communities and of amounts of oil in water, sediments, and organisms (2, 3). Investigations of oil spills generally commence after any initial damage has occurred, and baseline ecological data are usually lacking. This approach precludes measurement of the effect of oil spills because there usually is no knowledge of natural ecological variation (2). Such was the case in 1968 following the release of 3.2 million liters of oil from the wreck of the Witwater near the Galeta Marine Laboratory (Smithsonian Tropical Research Institute) in Panama (4, 5). Another even larger spill occurred in the same region in 1986 (6) (Fig. 1). In this article, we describe the types and extent of damage to coastal populations and communities in the first 1.5 years after the 1986 spill, and contrast our findings with earlier work and widely held views regarding the effects of oil on tropical coastal

communities.

The authors are resident or visiting investigators at the Smithsonian Tropical Research Institute (STRI), Apartado 2072 Balboa, Republic of Panama. Institutional affiliations of authors not employed by STRI are as follows: K. Burns and A. H. Knap, Bermuda Biological Station for Research, Bermuda GEO1; R. L. Caldwell, Department of Zoology, University of California, Berkeley, CA 94720; C. D. Getter, 915 Academy Street, N.E., Salem, OR 97303; R. Steger, University of California, Berkeley, Richard Gump Biological Research Station, B.P. 244 Temac, Moorea, French Polynesia; and E. Weil. Department of Zoology, University of Texas, Austin, TX 78712, and Fundacion Cantifica Los Roques, Apartado 1, Caracas 1010A, Venezuela.

6 JANUARY 1989

Fig. 1. Crude oil washing onto Galeta Reef on the Caribbean coast of Panama, site of a 15-year-old environmental monitoring program at the Galeta Marine Laboratory, Smithsonian Tropical Research Institute (white building on reef platform). (Photo by C. Hansen]

ARTICLES 37

Caribbean, Gulf of Mexico, and southeastern United States (10). Thus our observations are relevant to assessment of potential biological effects of pollution in several areas where extraction or refining of oil is ongoing on planned (11).

On 27 April 1986 at least 8 million liters of medium-weight crude oil (12) spilled from a ruptured storage tank into the sea on the Caribbean coast of Panama (6) (Fig. 2). This is the greatest amount of oil spilled directly into a sheltered coastal habitat in the tropical Americas (3, 13). For 6 days, onshore winds held the oil within Bahia Cativa adjacent to the refinery (Fig. 2), but runoff from rains and shifting winds then flushed the oil out to sea. At this time dispersant was sprayed on slicks in the mouth of Bahia las Minas and offshore (14). By 15 May, the oil had swept across fringing reefs and entered mangrove forests, small estuaries, and sand beaches within 10 km of the refinery.

Table 1. Comparison of visual assessment of amount of oiling on four coral reefs (ranked heavy (H), moderate (M), and none (N)] with results of gas chromatographic (GC) analyses of saturated hydrocarbons in coral tissues and by ultraviolet fluorescence (UVF) analyses of aromatic hydrocarbon fractions. Oil content by GC was conservatively estimated as the unresolved hydrocarbons in the elution range for alkanes containing 12 to 36 carbon atoms. Oil units by UVF were determined by comparison with the spilled oil. Values are expressed as micrograms per milligram of coral tissue (lipid or protein) and are the mean and standard deviation of triplicate analyses. Reefs are listed from west to cast (Fig. 2), Pavardi West is adjacent to the refinery and Palina West is just west of Isla Grande.

Plants and animals died wherever they came in contact with oil However, the types and magnitude of effects varied greatly with coastal topography and location, and among habitats and taxa. Such complex effects are similar to those of powerful hurricanes on Jamaican and Australian coral reefs (15). We first consider the spacial pattern of the spill and its effects in different environments, inclading variations in effects on different taxa. We then describe the patterns of responses of different organisms that discuss prospecta for recovery.

Spatial Distribution of Oil Among Habitats

The distribution of oil within 2 months of the spill was visUNİY assessed by aerial surveys between Rio Chagres, 27 km west of the refinery, and Punta San Blas, 98 km to the east, and by boat and foor from Rio Chagres to Nombre de Dios (16) (Fig. 2). No al was observed on shorelines cast of Isla Grande, and only a few patches were observed cast of Maria Chiquita and west of the Panama Canal entrance. Heavy oiling occurred along most of the coast between Isla Margarita and Islas Naranjos (Fig. 2). The straight-line distance between these two points is 11 km, but the labyrinthine shoreline is at least 82 km long and contains about 16 km2 of mangroves and 8 km2 of intertidal reef flats and subtidal reefs (17). Within this area. the only large unoiled areas 2 months after the spill were two mangrove lagoons that are isolated from open water by narrow channels (hatched areas in Fig. 2). Oil slicks ranging from a metallic sheen to brown patches are still common between Punta Muerto and Galeta, especially after heavy rains flush oil from mangroves and from the landfill beneath the refinery (18).

Within similar habitats, distance from the refinery, direction of al movement, and water depth apparently caused considerable vana tion in the degree of oiling even in the most polluted region Amounts of oil observed on or in sediments of seagrass beds. mangroves, and over reef flats were highest within a few kilometen of the refinery (Punta Muerto to Largo Remo and Punta Gaieta)

38

SCIENCE, VOL 241

Fig. 4. Percent cover of formerly abundant taxa and oil on mangrove roots in riverine, channel, and open coast habitats before and after the oil spill. O, Prespill data; O, unoiled sites; and, oiled sites. Means are plotted ±1 SE, converted from arcsine transformations. Some error bars lie within the plotting symbols, except for sampling time 3, when 1. ND, no data. "Leafy algae" include Polysiphoma, Acanthophora, and Ceramium as common genera; the most abundant "sessile animals" include hydroids and sponges. Sampling dates: 1 September to October 1981; 2 = January 1982; 3 June 1982; 4 = July to August 1986; 5 October to November 1986; 6 February 1987; and 7 May 1987. Results of repeated measures analysis of variance for oiled and unoiled sites after the spill are shown on each graph, NS, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

sis of aromatic hydrocarbon fractions (22). Preliminary resis generally parallel classification of sites based on visual inspection, a shown here for the concentration of saturated hydrocarbons tissues of the coral Siderastrea sidera (Table 1).

Biological Effects

Consequences of the spill were assessed differently, depending on the types of data available from before the oil spill (23). Ideally, biological parameters should have been measured at oiled and unoiled sites before and after the spill. This condition was sansbed for biota of mangrove roots and subtidal corals. Extensive prespul data are available for the reef flat at Galeta, but there are no cons sites and effects must be inferred from temporal change and from spatial distribution of oiling on the reef flat. In contrast, there are little or no appropriate prespill data for subtidal seagrass comman ties. In this case, comparisons were made between oiled and unqued sites after the spill, confounding the treatment (oiling) and geogra phy (Bahia las Minas verus the region between Portobelo and Isla Grande) (Fig. 2). The value of this approach is strengthened by information suggesting that faunas in seagrass beds were similar in the two regions before the spill (24), but confidence in assessment of oiling effects in this habitat is still more limited than in other habeats studied.

Mangrove communities. Red mangrove, Rhizophora mangle, forms nearly all of the fringing forest along this coast (17). By September 1986, a band of dead or dying trees marked the zone where ou washed ashore between Punta Galeta and Islas Naranjos (Fig. 3), no such band appears in photographs taken just as the oil was coming ashore. By November 1987 dead mangroves occurred along an estimated 27 km of the coast (25). Seedlings transplanted to heavi oiled sites did not produce new leaves, in contrast to transplants an unoiled site (26).

The prop roots of R. mangle are overgrown by algae and invertebrates that vary in species composition with exposure to the sea (Fig. 3) (6, 27). The epibiota on roots in three different habitas were sampled before and after the spill (Fig. 2) (28). Before the spell (Fig. 4), roots of trees directly facing the open ocean were covered with foliose algae and sessile invertebrates such as sponges, tvdroids, and ascidians. In mangrove channels, the edible over Crassostrea rhizophorae, and a barnacle, Balanus improvisus, were most abundant on roots. Roots in small rivers were dominated by the false mussel Mytilopsis domingensis and B. improvisus. Certain groups were more abundant a few years before the oil spill (1981 to 1987 than in 1986 to 1987 at unoiled sites (Fig. 4), possibly because of natural fluctuations in abundance. After the spill, the cover of al major groups was very greatly reduced in each oiled habitat (Fig. 41 There has been patchy recovery in the open habitat of foliose algae and sessile invertebrates, although not of the same relative abun dance of species. In the channels, cover of both the oyster and

Fig. 5. Abundance or biomass per sample of eight major infaunal taxa in three oiled () and four control (O) seagrass beds 5, 7, and 9 months after the oil spill (30). Means are plotted 1 SE, backtransformed from in (x + 1); some error bars lie within the plotting symbols. "Burrowing shrimp" include alpheids (most abundant), processids, callianassids, and upogebids. NS, P>0.05; *, P < 0.05 by repeated measures analysis of variance. Number per sample is shown at two scales. Data for polychaetes are shown as grams per sample.

100

80

60

40

20

SCIENCE. VOL 144