Since the end of last century, oil pollution of the open seas and coastal waters has become obvious mainly because of the frequent mass strandings of heavily oiled seabirds. In contrast to what is generally believed, oil incidents play a rather insignificant role in this form of pollution. Operational discharges by ships and frequent leakages of oil by ships and offshore installations are the main sources of oil washing ashore beached and found on oiled, beached birds. The oil pollution problem has been recognized as a significant threat to the marine environment, and several measures were taken to reduce the amount of oil which is released into the sea. Beached bird surveys (BBS) have always been used as an aid to demonstrate the impact of oil pollution on the marine environment, hut BBS results have played only a minor role in the assessment of the scale of and trends in marine oil pollution. Weather and wind ar normally said to influence the data so much, that the outcome is of limited value or very difficult to interpret at best. However, the use of an oil rate (the fraction of birds oiled out of the total number of birds washing ashore) to demonstrate the level of oil pollution in different sea areas is relatively new. Total numbers of birds washing ashore, usually expressed as densities (number per km surveyed), are now considered of secondary importance and these figures may only be used to examine the (local) impact of a given oil incident. Oil rates were found consistent in different species and in different areas. It is now believed that BBS results are quite useful indicators of the occurrence of marine oil pollution. On the third North Sea Ministers Conference in 1990 it was concluded that the possible use of Beached Bird Surveys was to be investigated, as an indicator of the effectiveness of actions taken to reduce oil pollution of the seas. Following a report on ’The Value of Beached Bird Surveys in monitoring oil pollution’, published in 1992, it was concluded on the interim Ministers Conference in Copenhagen in December 1993 that "In 1995 it should he possible to assess the effectiveness of the measures already agreed, and an assessment should he made available to the Fourth International Conference on the Protection of the North Sea. The Monitoring of oiled seabirds should continue as a useful indicator of the effectiviness of these measures In The Netherlands, BBS were an activity of volunteers during the last three decades. Now that BBS results were considered of interest to monitor trends in oil pollution rather than the effect of oil on (sea-) birds, the Directorate-General of Shipping and Maritime Affairs (DGSM) initiated the continuation of Beached Bird surveys in the Netherlands in the form of a research project to evaluate its own, national 'Milieubeleidsplan voor de Scheepvaart’ (environmental policy plan for shipping). In this project, (1) 10 years of BBS data were computerized and analysed, (2) the statitistical validity of the information collected during beached bird surveys was evaluated by means of a power analysis and (3) the surveys were continued in 1994/95. The Institute of Forestry and Nature Research was ordered to produce a report on these matters, based on data collected by the Dutch Seabird Group, and CSR Consultancy acted as a sub-contractor to perform the project. In this report, the results of beached bird surveys over 1986-95 are summarized (chapter 2), it provides the results of a power analysis (chapter 3) and discusses the use of BBS results for policy makers (chapter 4). BBS results 1986-95 In 1986-1995, the highest oil rates were found in divers, grebes, Gannet, scoters, Kittiwake and auks (table 4). Oil rates were significantly higher in winter (November-April) than in summer and it was concluded that these data sets should not be mixed in further analysis. In this report, ’winter oil rates’ were provided, unless otherwise stated. A clear exception is the comparison of oil rates found in 1969-85 and 1986-95 (tables 8 & 14), because in the former period ’winter surveys’ could not easily be separated from summer surveys. The oil rate found in 1986-95 was lower than the oil rate found in 1969-85 and this was concluded for all species and species groups of birds. Compared to other North Sea countries, the oil rate in The Netherlands is still very high. Most of the oil found on Dutch beaches and stranded birds in The Netherlands originated from operational discharges by ships (bilge oil and engineroom residues); crude oil was rarely encountered. Numbers of seabirds washing ashore are subject to massive fluctuations from year to year and month to month, caused by a variety of factors including variable bird densities at sea, residual currents, prevailing winds, and several mortality factors. The variation in oil rates, specific for species, groups of birds and certain areas, is minimal compared to the variations in overall numbers. The oil pollution of beaches showed the same seasonal pattern as oil rates in stranded birds (figure 2) and the frequency by which polluted beaches were reported has not changed since registrations began in the early 1980s. Recording trends in marine oil pollution: using oil rates One objective of the Beached Bird Survey (BBS) is monitoring the amount of oil pollution of the sea by assessing the fraction of oiled objects on a beach. BBS results are a derivative of a direct census of the occurrence of oil, with some very strong points because of its scale (all Europe), cost (with partly volunteer schemes rather low budgets are possible) and the length of its time series. In most countries, data are available over the last two or three decades, with unchanged methods, forming a unique data set which can readily be explored and which may form an additional source of information to other, perhaps more direct measurements. Ideally, an experiment would be set up in which clean pieces of cloth or whatever were released into the sea in huge numbers, to be recovered on the beach. The fraction (%) of oiled objects, the oil rate, would represent the chance for the pieces of cloth to become oil contaminated in that particular sea area. The same experiment in The Netherlands and in Shetland would result into a totally different oil rate (very low in Shetland, very high in The Netherlands). It has been suggested, that the recovery of beached birds is in fact such an experiment because the frequency of oiling of stranded seabirds is a reflection of the chance to become oil contaminated. However, if birds would only die at sea becaused of oil, the oil rate on the beach would be meaningless. If birds would never die because of oil, but get oil in their feathers while dead and afloat, the oil rate would he precisely what was wanted. Assuming that, generally speaking, a minority of the birds recorded on beaches died because of oil and considering that there is a linear relationship between the desired oil rate V ’ and the oil rate recorded on the beach 's’ (figure 12), the BBS will serve as an accurate tool to measure trends in oil pollution, but a less accurate tool to work out ’true’ levels of oil at sea. If methods within countries remain unchanged also in the future, results of trends in different schemes can readily be compared in space and time. Results of the power analysis The assumption is made that the fraction of all beached birds that is oil contaminated is in someway related to oil-pollution. This leads to the question: is there a significant trend over years in the fraction of oiled birds (and hence in oil pollution). This note is concerned with the statistical power of appropriate trend tests. The power (l-(3) is the probability that a trend, if present, will be detected as statistically significant. It depends on the size of the trend, the error variance, the number of years (n), and the size of the test (a). Presumably the fraction of oiled birds (y) has some s-shaped relation with some index of oil-pollution (x) (figure 11). A widely used mathematical representation of such sshaped curve is the logit function: ex y= 1+<‘ (i). The analysis focuses on this index of oil pollution, which equals (as follows from (1)): (2). Figures 14 (Guillemot) and 15 (Razorbill) show time series of the observed index x and the fitted linear trends (by least-squares estimation) for several countries (h The Netherlands, d Denmark, g Germany, n Norway, s Shetland). Table 10 gives the residual mean squares, which can be used as estimates of the error variances. These residual mean squares are in the same order of magnitude for the various countries and do not show any relationship with the size of the average index. This ‘homogeneity of variances ’ is a desirable property as it is one of the assumptions of the underlying regression model. The untransformed data, i.e. the fraction of oiled birds do not show this property. For the Guillemot the error variance is about 0.49, i.e. an error standard devaition of about 0.7. Table 10 also gives the estimated slopes and the accompanying P-values. If, as a side-step, we consider the case that the true x=0, which implies that the true y= 0.5. Then, an error standard deviation of 0.7 for the observed x is equivalent to an error standard deviation of 0.175 (0.7/4) for the observed y, as m =1 4 (3). If the error variance would be solely due to a binomial sampling error (which equals n (4), where it is the independent probability that a bird is oiled, i.e. the true y), then such error variance would be obtained by sampling only 8 birds (which follows from 2\- -2/=0.175 N 8 (5). In practice the number of birds that have been observed is much larger. Hence, this little excercise showed that the observed error is probably not due to sampling error hut to ‘real’ deviations of the ‘true’ yearly means from the linear trend. It supports our choice for the use of a least-squares approach. As the test of the regression slope is, in fact, a one-sample t-test, the power can be relatively simply calculated by using the cumulative Students t-distribution Junction (tcf with n-2 degrees of freedom), where the effect size d is expressed as the size of the trend (the slope of the regression) divided by its standard error (which follows from the estimated error variance and the number of years that will be sampled). Hence the power equals l~V~tCf{tal2,n-2~d)+{l~tCS[ 1 a 12.n-2 <*)) (6). Figure 16 gives as an example the power as a function of the number of years for slope=-.11 (h) and slope =-.24 (g) with an error variance of 0.49 (as is about true for the Guillemot). It says that a decrease in oil-pollution as observed in Germany (- .24) will be detected with a probability of 90% after 12 years. The same procedure was followed using data collected in The Netherlands during 1986-95 (figures 17-20, tables 11-12) and using a slightly longer set of data which was available for Noord-Holland, a small part of the country (figures 21-22, table 13). The results showed declines in oil rates all over, and the probahlity to find significant results with a certainty of ca. 75% within 13-17 years. The longer data set used illustrated that this was indeed the case: all delines were significant trends. The conclusion from the analysis was that BBS results are sensitive and useful to detect even minor trends in the frequency of occurrence of oil on the corpses. Conclusions and recommendations for further research Oil rates in beached birds in the Netherlands have consistently declined over the last 10 years and are now lower than before (table 14). The trends found over the last decade were quite weak and not significant, but can be expected to be so over a slightly longer period. The trends in different groups of birds (estuarine, coastal and offshore species; figure 27) run more or less parallel. If the oil rates found represent the chance for (corpses of) birds to become oil contaminated, and if this chance is mainly affected by the amount of oil at sea (number of slicks, densities, quantity of oil released), than a decline in oil rates on the beach would imply a decline in the amount of oil at sea. If we assume that other factors influencing the chance for birds to become oiled are (on average) constant, than, on the basis of beached birds, the amount of oil released into the southern North Sea would have declined by at least 20% since 1986. Future research will have to focus on several species and/or groups of species simultaneously to avoid problems caused by certain mortality incidents in individual species. Densities will have to be measured to enable a fair judgement of drops or jumps in oil rates. At the same time, background information needs to be collected for all species used in the monitoring programme, again to make sure that the oil rate found is not influenced by circumstances which are particular for any of the individual species, key species in future monitoring in The Netherlands would be Guillemot, Razorbill, Kittiwake, Fulmar, Gannet, scoters and Larus-gulls. It is concluded that more historical data will need to be computerized to enable further analysis of trends in oil rates, including information collected prior to the date when MARPOL Annex I was effectuated (October 1983). A continuation of the monitoring programme will focus on the winter period (November-April). It is strongly recommended to include a sampling programme to assess the different types of oil on beaches and beached birds. Such a programme would also provide information on the occurrence of other chemical substances and non-mineral oils.