Influence of hydrology on temporal variability of fish abundance in the onchocerciasis control programme area in west Africa

aONCHOCERCIASIS CONTROL PROGRAMME IN WEST AFRTCA PROGRAMME DE LUTTE CONTRE L'ONCHOCERCOSE EN AFRIQUE DE L'OUEST ECOLOGICAL GROUP Fifteenth session Bamako. 3l Januarv - 4 Februarv 1994 ocP/vcu /HYBrol94.s ORIGINAL: French INFLIJENCE OF HYDROLOGY ON TEMPORAL VARHBILITY OF FISH ABIJNDANCE IN THE ONCHOCERCIASIS CONTROL PROGRAMME AREA IN WEST AFRICA Gilles Poizat Report to the Ecological Group Onchocerciasis Control Programme in West Africa World Health Organization December 1993 SUMMARY The analysis was centred on five stations in Cote d'Ivoire for which OCP's monitoring covers periods ranging from 10 to 20 years. . To begin with, the effect of discharge on sampling has been revealed: the numbers caught are generally less when the discharges are high. Next, what was related to the seasonal scale, and to the annual scale in the temporal fish abundance variations, was examined. The annual component proved to represent a greater part of the temporal variations than the seasonal component. The seasonal variations are partly explained by the influence of discharge on the sampling and most likely also by seasonal migrations. According to the stations, the minimal abundances are observed at the beginning or at the height of the spate. The annual species abundance variations have been described separately in each station, using the principal component analysis. Contrary to the results of previous analyses by lrveque et al. (1988) and Hugueny (1990), clear trends towards a decrease in the abundances of most of the species appear from 1975 to 1992 at the stations of Border Bridge on the Irraba and Ganse on the Comoe. At Niaka on the White Bandama, the trend exists only from 1987 to 1992. In the first two stations mentioned, the latter period corresponds to an increase in the trend. The increase in the fishing pressure exerted by the riverine populations, which has been expanding markedly in recent years, could explain the trend observed from 1987 to 7992. A long-term trend has not been detected in the hydrology of these stations. It therefore seems unlikely that this environmental factor is the cause of the fish community trends. The influence of hydrology on annual fish abundance variations has been analysed by taking into account only the juveniles of the most abundant species. The correlations between the abundances per catch and the previous discharges show the existence of relations that are often strong but very variable within one and the same species between the stations. The positive relations between the abundances of juveniles and the discharge of the high-water months of the previous year which were expected appeared only rarely. Many correlations are negative, whereas a generally positive effect of the previous discharges was expected. Finally, strong correlations have often been noted between the abundances of juveniles and the discharge of many preceding years, with a shift far above their presumed age. TABLE OF CONTENTS INTRODUCTION a' DATA INFLTJENCE OF DISCHARGE ON SAMPLING L. Introduction 2. Method 3. Results SEASONAL AND ANNUAL COMPONENTS OF TEMPORAL VARIATIONS IN SPECIES ABTJNDANCES 1. Introduction 2. Definition of scales 3. Partition of temporal variations 4. Seasonal variations 5. Annual variations CORREI.ATIONS BETWEEN SPECIES ABUNDANCES AND PREVIOUS DISCHARGES 1. Introduction 2. Method 3. Results 4. Discussion CONCLUSION REFERENCES INTRODUCTION As part of the WHO Onchocerciasis Control Programme in West Africa (OCP), rivers are treated with insecticides in order to control the populations of the blackflies, the vectors of this disease. At the same time, the non-target fauna in these rivers (invertebrates and fish) is monitored before and after larviciding in order* to check in situ the impact of the insecticides. As regards fish, two approaches have been used for the analysis of the long-term impacts of the insecticides on the communities. I-eveque g[ al. (1988) studied the temporal trajectories of variables describing the communities, in order to observe whether insecticide spraying makes these variables leave their field of "natural" variations and whether a long- term trend could be detected. It appears that the "natural" variations are of a very wide range, notably within each year, and no general long-term trend has been detected. In order to detect possible insecticide effects within this considerable "natural" variability, Hugueny (1990) used a more refined statistical method. The time-lag between the start of the insecticide spraying in the neighbouring stations makes it possible to place oneself a posteriori in a pseudo-experimental situation in which one has a control station and a test station. When the communities of nvo geographically neighbouring stations have concordant temporal trends, the station treated first is considered after the start of its treatment as the control. The station treated secondly constitutes the test station. By comparing the faunistic differences between those two stations, before the treatment of the test station and after the treatment of the test station, the effect of the insecticide can be tested within the considerable "natural" variability. Despite the fact that the method is statistically refined, Hugueny concluded, like [rveque et al., that the "natural" variations are preponderant and that if any insecticide effect exists it is minimal. However, the two studies detected a long-term downward trend in the abundance of some species (particularly Petrocephalus bovei). The conclusions of these two studies make one to be interested in the nature of these "natural" variations and look for their causes. The hydrological regime is suspected of having a considerable effect on these variations in fish abundance on a seasonal scale on the one hand and on an annual scale on the other hand. On a seasonal scale, lrveque et al. mentioned that the variations in catches are partly related to the influence of discharge on the sampling, the nets being more effective at low water. On an annual scale, the influence of hydrology on variations in fish abundances has been revealed in many intertropical watercourses (see Welcomme, 1985), recruitment being generally favoured by a considerable spate. According to Hugueny, the similarity between annual variations in communities in geographically neighbouring stations indicates that these variations are governed by climatic factors, notably hydrology. 2f. a GERIA Lo0o. SE LI Nlomry lo I o--- 8 E N IN IE co RE E AL M B I I G E a Tofilol.a naa a 2 6 o t-M AL I ER IL- I t I I \I ougou -- ^, N I I I I rT I \o 'l *' a I El. I H ANA\, ( I I I ,/ Lt R AE Cooo lry Fraalorn Moarorlo ccro Ab{dlon Figure l: Geographical location of the five stations selected. River Station No OCP Fish Monitoring Hydrological data Leraba Comoe Bandama Blanc N'Zi Sassandra Pont Frontirire Niaka Pont, route de Dabakala 0r 02 03 74 - 92 (fishing) 75 - 92 (fishing) 75 - 92 (fishing) 80-92 78-92 75 - 92 Pont de Semien 05 06 76 - 85 (fishine) 54-83 s4-8675 - 86 (fishins) Table l: Description of the extent of the data available in the five stations slected -t I I I I I I t \ I 3In this work, there will be a more in-depth examination of the relations between the hydrological regime and the variations in fish abundances. Firstly, the effect of discharge on sampling will be tested. Next, seasonal and annual variations in species abundances will be analysed, laying more emphasis on the influence of hydrqlogy on an annual scale. DATA Only data collected as part of OCP activities in Cdte d'Ivoire in stations where the monitoring has been sufficiently long and for which there are adequate hydrological data will be analysed. Thus, five stations (Figure 1), for which the extent of the fish monitoring ranges from 10 to 20 years (Table 1), have been selected. It has been chosen to concentrate on only one geographical area whose fish communities are quite homogeneous rather than including as many stations as possible, which would have led to an additional source of variability. The sampling was made using a battery of gill nets comprising five nets whose mesh sizes were 15,20,25,30 and 40 mm respectively. Intermediate mesh sizes were added to them during the monitoring but they will not be taken into account. The results of each fishing have been reduced to a standard sampling effort corresponding to catches per 100 m2 of fishing area per night. Each station is sampled several times per year (generally four times), but the fishing months are not necessarily the same from one year to another. More details on the fishing method can be found in the article of I-eveque et al. (1988). The abundance measure that will be used is the number caught per unit effort transformed into Napierian logarithm. The logarithmic transformation normalizes the distribution of the measure and thus allows the use of parametric statistical tests. INFLUENCE OF DISCHARGE ON SAMPLING L. Introduction The sampling conditions were not similar, according to the discharge. The depth at the place where nets are placed depends on it and generally they touch the bottom during the low-water period while at high water they sample only the upper part of the waterspout (L. Yameogo, pers. comm.). Furthermore, assuming there is a dispersion of fish in the whole mass of water, an increase in discharge "dilutes" the fish in a greater volume of water, which leads to a decrease in densities. In order to evaluate the influence of discharge on sampling, the relations between the numbers caught per fishing and the mean discharge for the same month will be examined. The total number (all species included) per unit effort will be considered and then, on the one hand, the number of individuals belonging to the benthic species and, on the other hand, the number of other species which would be termed "pelagic". A species is considered to be benthic if its mouth is in an inferior position. This morphological characteristic is noted according to the fauna of lrveque et al. (1992). 4The use of this classification should make it possible to examine, on the one hand, the effect of depth on benthic species catches alone and on the other hand, the effect of dilution on all the species. 2. Method The statistical distribution of the discharge values per month is very asymmetrical, with many low values and a few high values. Accordingly, discharge classes with a logarithmic scale will be defined in each station. Next, the numbers caught (in log) in the different discharge classes witl be compared using a variance analysis to a factor (ANOVA) in each of the stations. The number of discharge classes has been set at four. The range of classes is defined as follows: u = (Ln (ma*. discharge) - Ln (min. discharge)/4 If the minimum discharge is equal to zeto then the transformation Ln (discharge + 1) is used. The F test of ANOVA assumes the equality of the variances of the variable studied (here the log of the number) in the different classes. This assumption will be verified by Bartlett's test and when the variances are significantly different, the classes will be grouped together so as to put us under the conditions for the application of ANOVA. In each station, the test will be applied to the log of the total number (LN-NT), the log of the number of benthic species (LN-N Benthic) of and the log of the number of other species (LN-N Pelagic) caught per unit effort. In order to increase the power of the test, the data for the five stations will then be cumulated after having normalized the abundance measures in each station (i.e., to reduce the mean to zero and the variance to one by subtracting the mean from it and then dividing it by the standard deviation), in order to eliminate differences between the stations. The transformation of the discharge into classes constitutes a normalization in itself. The classes defined separately in each station will therefore be maintained in the cumulative analysis. 3. Results In the separate analyses by station, a significant effect (threshold of 5%o) of discharge on the total number caught is observed solely at station 02 (Table 2.). This effect concerns only the benthic species. This suggests that this effect is related more to the variation in the depth of the place where the nets were placed than to the dispersion of the fish in a greater volume of water. In stations 03 and 05, the effect of the discharge on the total number caught is close to the threshold of statistical significance but, unlike the previous stations, it is observed that these effects mainly concern the pelagic species. The numbers of benthic species caught at low water in these two stations are relatively (compared to the other 5species) less than those caught in station 02. Since this effect of the discharge does not concern the benthic species it can be thought to be due to dilution. This dilution would affect the pelagic species more, maybe because of a more marked tendency to go and exploit the flooded areas. As regards the few stations where the effect of discharge is statistically significant, it is probably related to the number of fishing per discharge class which is often quite low. The analysis of the cumulation of the normalized data for the five stations confirms this, revealing a significant decrease in the numbers caught when the discharge increases. This effect is shared by the benthic and pelagic species. However, it is observed, particularly for the pelagic species but also for the benthic species of several stations, that the minimum of the numbers caught is in class 3 of the discharge and not for the ma:<imum discharges. The interpretation of these results requires a specific knowledge of the fishing methods, notably changes in the location of the nets according to the discharge. In conclusion, it has turned out that discharge influences sampling significantly in a way which may depend on the stations and species. This should therefore be taken into account when temporal variations in fish abundances are analysed. SEASONAL AND ANNUAL COMPONENTS OF TEMPORAL VARIA'TIONS IN SPECIES ABUNDANCES 1. Introduction The cyclicity of the discharge leads classically to the distinguishing of two time scales for the description of its variations: the seasonal scale and the annual scale. These two observation scales will therefore be used for the study of the influence of hydrology on fish abundance variations. After defining these scales more specifically, the seasonal and annual components of the fish abundance variations will be analysed in relation to discharge variations on these same scales. 2. Definition of scales Seasonal scale The two-monthly division of the year employed by Hugueny (1990) will be used. The year is divided into six periods of two months (January-February, March-April, etc.). Annual scale In order to be able to compare the annual variations with the hydrolory of the previous years, the year has to be defined according to the hydrological cycle. Considering the potential role of spate, for each of the catches in a particular year the previous year's spate should be the same. In the calendar year, this is not respected for 6the months of November and December. Therefore, the hydrological year will start in November (flood-subsidence period) and finish in October (maximum spate or beginning of flood-subsidence period). 3. Partition of temporal variations Model The temporal variations in species abundances will be partitioned according to the additive model of variance analysis to 2 factors (season and year) without interaction (similar to that used by Hugueny): V(T) = V(S) + V(A/S) + V(R) where V(T) corresponds to total temporal variations, V(S) to seasonal variations, V(A/S) to annual variations, knowing the seasonal variations, and V(R) to residual variations. The annual variations are determined "knowing the seasonal variations" because the observation plan according to the factors season and year is not complete (i.e., there is no fishing during all the periods in each year). The simple calculation of the annual means would therefore be inlluenced by the periods represented in each of the years. To correct that, the annual means will be calculated after having eliminated the seasonal abundance variations. The systematic effect of one season is estimated by the difference between the season's mean and the general mean. This correction will be applied separately to the numbers of each species caught in each station. Only the years having more than one fishing are taken into account. The variations are measured by the sum of the variances between the log- transformed abundance of each species. The statistical importance of the seasonal and annual components of the temporal variations is measured by the ratio between the variations related to a component and the total variations. It is therefore expressed as a percentage of the total variations (Sabatier et al., 1989, Brocard et al., L992). Results Station Seasonal comp. Vo of V(T) Annual comp. Vo of. V(T) 43.87o 38.3%o 43.9Vo 32.5Vo 45.5Vo Residual %o of V(T) 37.4Vo 50.0Vo 50.7Vo 58.7Vo 40.9Vo 01 02 03 05 06 V(T) 15.10 1,5.46 13.88 23.57 11.55 18.8Vo ll.7Vo 5.4Vo 8.8Vo 13.5% Table 3: Partition of temporal species abundance variations according to the additive model of effects of the factors season and year. The results (cf. Table 3) show that the annual component represents a very considerable part of the temporal species abundance variations (32 to 45Vo according to the stations). Comparatively, the seasonal component is less marked (5 to l9Vo). The residual variations, i.e., not explained by the simple additive model of seasonal and annual effects, represents 3l to 59Vo of the temporalvariations. This residual component 7is due both to the interaction between the two factors and to variations not related to these factors. Sampling fluctuations are probably preponderant within this second source of residual variability. In the light of this partition of the temporal variations, each of these components will be analysed in relation to hydrology, laying more emphasis on annual variations. 4. Seasonal variations Considering the effect of discharge on the numbers caught, seasonal variations in catches related to the annual hydrological cycle could be expected. Furthermore, the biological cycle of the species sometimes implies changes in habitat and activity, related mainly to reproduction, likely to lead also to seasonal variations in catches. Only seasonal variations in the total numbers caught in each station will be presented. The maximum numbers caught are generally observed during the dry season, but also during the flood-subsidence period (Fig. 2). The minima are observed during the period of rise in water level. At the height of the spate, the status of the numbers caught varies according to the stations. Thus, for similar discharge values, the numbers caught depend on the period of the year. This cannot therefore be explained by the sole effect of discharge on the sampling. The minimum observed in the catches at that period could correspond to the reproduction migration to habitats different from those where the nets had been placed because the reproduction period of most of the species is at the beginning of the spate (Paugy, 1988). Another possible explanation is the year's recruitment which would appear in the catches during the flood-subsidence period. This second assumption is not confirmed when the seasonal variations in catches per mesh size are observed (Fig. 3). In actual fact, the increase in catches during the flood-subsidence period is common to all the mesh sizes, including the big ones in which the year's juveniles cannot be found. It therefore seems more plausible that these seasonal variations not explained by the direct effect of discharge on the sampling are due to seasonal changes in the habitat or activity of the fishes. 5. Annual variations lntroduction On an annual scale, variations in the abundances of the populations are due mainly to demographic processes (natality, mortality, immigration, emigration). Hydrology is likely to influence the recruitment of fish populations (i.e., natality and juvenile mortality because many species use zones flooded by the spate as reproduction and nursery habitats (Welcomme 1985). The availability of these habitats depends on the volume and duration of the spate. It is therefore directly related to hydrology. The annual component of the variability of species abundances will be explored. If the effect of the hydrology were preponderant and if all the species responded to it in a similar way then annual variations in the abundances of the different species should be positively correlated among themselves. Deviations from this assumption will be 8studied by means of principal component analysis (PCA). Principal components summarize, optimally, variations in species abundances. They will therefore make it possible to describe the annual variations synthetically. Methods As was done above, during the partition of the temporal variations, the seasonal abundance variations will be eliminated before studying their annual variations. This method is similar to that used by Hugueny (1990). The demographic processes, particularly recruitment, are expressed quantitatively on the abundances of each species. The method of analysis of annual species abundance variations should therefore (1) be multivariate and (2) deal with the absolute abundances and not the relative abundances of the different species. These criteria lead to a principal component analysis (PCA). A PCA will be applied to the table of mean abundances per hydrological year and per species, separately in each of the five stations. In order not to encumber the analysis with rare species which, in any case, would not participate actively in the definition of the principal components, only species present in more thanA\Vo of the fish catches will be taken into account. Only the years for which the fish catching was more than one will be included in the analyses. Results Stations 01, 02, 03 In stations 01, 02 and 03 (Fig. 4, 6, 8), the first principal component bears a size effect, i.e., the abundances of most of the species are positively correlated and the years are distinguished between themselves mainly by their total abundances. In these three stations, only Marcusenius ussheri (1109) and Chrysichthys nigrodigitatus (4502) do not follow this general trend. The second principal component, being independent of the size effect borne by the first, accounts for annual variations in the species composition. The relations between species are not very similar between these three stations. The representation of the coordinates of the years according to time (Fig. 5, 7, 9) shows a marked long-term trend on the first principal component, corresponding to a decrease in the total abundances. For stations 01 and 02, this trend is marked on the whole duration of the monitoring, i.e., from 1975 to 1992. For station 03, it is marked only from 1987 to 7992. The latter period corresponds also to a strengthening of the trend in the first two stations. Stations 05 and 06 In these two stations (Fig. 10 and 12), some species greatly contrast with all the others on the first component. They are Alestes baremoze (1702), Schilbe mandibularis (3902) and Schilbe mystus (3801) in station 05 and Brycinus nurse (1709) in station 06. For that reason, the first principal component does not correspond to a total abundance gradient but reveals clearly contrasting annual variations among the community's species. I 9In station 06, the second principal component, whose statistical significance is similar to tht of the first (3lVo of the annual variations as against 35Vo for PCl), bears a size effect since only a few species contrast slightly with all the others. The temporal trend of species abundances summarizedby the first nvo principal components (Fig. 11 and 13) shows significant variations between consecutive years withbut it being possible to detect a clear long-term trend during the monitoring. However, in station 06, a change in sign is noted on the first component as from 1"981, corresponding to an increase in the abundances of Brycinus nurse (1709) to the detriment of most of the other species. Discussion The two previous studies on long-term variations in fish communities in the OCP area (I-eveque et al. 1988 and Hugueny 1990) concluded that there was no long-term trend. The present analysis shows clearly the existence of a long-term trend towards a decrease in ihe abundance of most of the species at the Border Bridge on the Irraba (station 01), Ganse on the Comoe (Station 02) and Niaka on the White Bandama (station 03). These conclusion differences can be explained by the additional monitoring years available today. As a matter of fact, it is observed that the trend is particularly marked as from 1987. In the previous years, the "natural" variations between consecutive years were preponderant and masked the trend which appears in stations 01 and 02. At Niaka (station 03), the trend is noticeable only from 1987 to 1992. On the other hand, at the Border Bridge just as at Ganse (stations 01 and 02) the trend is clear from 1975 although it was'Jumbled" by an abundance peak in 1980 (and 1981 and 1987 in station 01) and a hollow in 1983 at the Border Bridge. It is possible that the cause of this trend was different between 1975 and 1987 on the one hand and between 1987 and 1992 on the other hand. The riverine populations have increased markedly by in recent years and maybe the fishing pressure they exert on the fish populations is responsible for the trend observed since 1987 (L. Yameogo, pers. comm.). A responsibility of the insecticides in the abundance decreases observed since 1975 is not impossible. Hugueny's analyses (1990) had concluded that there were no major effects of ttre insecticides. However, the method used could detect only rapid reactions of the populations to the insecticides and it is possible that the reactions were slow or delayed, particularly if they involved demographic processes (B. Hugueny, pers. comm.). Howev-er, only two stations show a trend from 1975 to 1987 while the five stations studied have been under treatment since 1975; this would run counter to the assumption of an effect of the insecticides. These questions will not be delved into further here because this is not the subject of the study. Could hydrology explain (1) the long-term trends? (2) the annual variations around the trend? (3) the total annual variations in the stations where there is no trend? In each of the five stations, the parallel representation of the annual variations in discharge for the high-water months and of the coordinates on the principal components (Fig. 5,'1,9,11 and 13) does not reveal clear systematic relations between the species abundances of the community and the discharges for the same year or 10 previous years. No long-term trend is distinguished in the high-water discharges. Thus, the similarity of the annual variations between species cannot be explained simply by the same reaction to hydrological variations. In the following chapter, the possible relations between fish abundances and hydrolory will be looked into in a more in-depth way. CORRETATIONS BETWEEN SPECIES ABUNDANCES AND PREVIOUS DISCHARGES 1. Introduction By influencing reproduction and the survival of young stages in particular, hydrology is likely to act on fish abundances with a delayed effect. According to their adaptive strategy, particularly the reproduction period and the mode of use of the habitat, it can be expected that the species would react in different ways to hydrolory. The key periods, during which the influence of discharge could be considerable, vary probably according to species. The correlations between the abundances of each species and the previous monthly discharges will be examined systematically. This constitutes a classic method for the study of the influence of hydrology on variations in fish abundances (Merona & Gascuel 1992). The correlations obtained do not constitute proofs of causality and could result from fortuitous relations. However, in making the correlation analyses for each station separately, it is correlations which recur on the abundance of one and the same species in the different stations that will be looked for. Such relations would constitute a sign in favour of causality links. In order to better isolate the effect of hydrology on the recruitment of the populations, the correlations will be analysed by considering only catches in the smallest mesh sizes: 15 and 20 mm, for the frequent species whose maximal sizes or body heights are the greatest (this characteristic is noted according to the fauna of [rveque et al. L992). Thus, the youngest individuals of these populations will be selected. The species chosen for this analysis are: Alestes baremoze, Brycinus nurse, Synodontis schall, Schilbe mandibularis, Chrysichthys maurus, Hydrocynus forskalii and Brycinus macrolepidotus. According to Paugy (1988) it is known that A. baremoze and S. schall reproduce at the beginning of the spate, while H. forskalii and B. macrolepidotus reproduce both before and after the spate. These reproduction period differences will be observed to see whether they lead to differences in the reactions to hydrology. 2. Method In each station and for each species selected, the temporal autocorrelation of the mean annual abundances will be examined. The abundance-discharge correlations will be calculated only if the autocorrelation is not significant (risk l%o), in order to place onself under the sample independence conditions imposed by the correlation test. I 11 The correlations will only be calculated if the species is present in at least 70Vo of the catches for the mesh sizes considered. A too great proportion of zero abundances corresponds to a very asymmetrical distribution of the abundances per catch, which does not agree with the assumption of normality of the vapiables which underlies the correlation test. Each of the discharge value distributions of a particular month for different years has a distribution close to the normality. The correlation analyses will be made on the species abundances after having eliminated the seasonal component from them. From a statistical viewpoint, it is preferable to consider the catches separately rather than treating the annual means. In fact, the trust that can be put in the means depends, on the one hand, on the number of catches in each year and, on the other hand, on the variability around the mean (non- seasonal intra-annual variability). In considering the catches individually, each couple (abundance of the species in a catch, discharge) will have the same statistical weight. The correlations will be calculated with the discharge of each month of the year, with a shift towards the past varying from one to six years. The linear model used for the measurement of the reactions of fish abundances to previous discharges has the disadvantage of not detecting bell-shaped reactions in which the ma.:rimal reaction is observed in the middle of the gradient and not at one of its ends. To get round this disadvantage, the discharges should have been converted into classes and treated without taking their order into account (Rose et al. 1986). For practical reasons, this type of method could not be used; linear adjustments, which constitute a classic method, will be made. The non-significant correlations will sometimes be examined graphically to see whether they are due to bell-shaped reactions. 3. Results The correlations are calculated on series having at least 26 catches and more often than not between 40 and 50 catches. The signs and statistical meanings of the correlations are noted in Figures 14 and 15. It will be observed that some correlations are very highly significant, in positive and negative terms. On the contrary, in station 05, the abundances of B. nurse and S. schall have no significant correlation with the previous discharges. The comparison of the correlations obtained in different stations for one and the same species shows that there are no stable relations which could be found systematically. In many cases, significant relations are observed for great shifts in years. The graphic examination of some non-significant relations, particularly on the high-water discharges of the year preceding the catch, does not show a bell-shaped reaction. 4. Discussion The results obtained are quite surprising. The relations expected are of the type observed for A. baremoze at the Semien Bridge (station 06). The high positive correlations with the high-water discharges of the previous year suggest that the 15 and 20 mm mesh sizes catch the individuals in their first year and that they are all the more abundant if the previous spate had been considerable. For B. nurse at Ganse (station 02), the correlations are positive also with the discharges of the month of September of t2 two years earlier which suggests that the 15 and 20 mm mesh sizes catch the first two cohorts of the population. This type of relations which was thought to be the rule appears, in fact, to be an exception. l Just as a stability is not detected in the correlations obtained for one and the same species in different stations, no systematic differences are observed between the species, which are related to their reproduction period. The frequent negative correlations with the discharge are observed for some months and with varied shifts. These relations can be understood when they are around the reproduction period of the parents of the cohort caught, as is the case of B. nurse at Ganse and Semien (stations 02 and 06) in the months of May and June of the previous year. Great discharges would have a negative effect on the survival of the eggs by transporting them or getting them stuck in the mud. It is quite difficult to interpret, biologically, the positive or negative correlations observed with time intervals above the presumed age of the fishes caught. However, correlations greatly significant at shifts of three to six years are frequent. Two explanations appear to be plausible: either these relations are fortuitous and not due to the discharge, or the effects of the discharge are indirect and act through the previous cohorts. In this second explanation, a competition with the previous cohorts and a positive relation to the abundance of parents could be expected. Thus, discharges favouring previous cohorts would be harmful unless this concerns the cohort of the parents. The latter assumption is supported by the fact that often an inversion of correlation signs is observed over the years (particularly for B. nurse, S. mandibularis and H. forskalii). However, it would be quite surprising that these indirect effects are stronger than the direct effect of discharge on cohort(s) caught. The assumption of indirect effects through previous cohorts could be verified by including their abundances alongside the discharge, as explanatory variables of a more general model. The possibly fortuitous character of the relations would be due to a similarity between the variations in the abundance of a species and some previous discharges, without there being a causality link. In principle, this type of relations is a[[ the more probable because the number of years represented in the data is low. The minimum number of years represented in each of the correlation calculations is five in station 0L, eight in stations 02 and 06 and nine in stations 03 and 05. However, in each station, more significant correlations are not observed in situations in which the numbers of years represented are the lowest. It is possible that the trends observed, particularly from 1987 to 1992 in stations 07, 02 and 03, have jumbled the relations with hydrolory. However, the results for stations 05 and 06, which are devoid of a clear trend, are not much clearer. These results suggest that hydrology does not act in a simple way on annual fish abundance variations. The environmental characteristics of each station and the past history of the community probably interact on the effects of hydrology. 13 CONCLUSION The objective of this work was to statistically study the influence of hydrolory on temporal fish abundance variations. The analysis was centred on five stations in C6te d'Ivoire for which OCP's monitoring covers periods ranging from L0 to 20 years. To begin with, the effect of discharge on sampling has been revealed: the numbers caught are generally less when the discharges are high. This effect is related both to the variations in the depth at which the nets are placed and to the decrease in densities due to a dispersion of the fish in a greater volume of water. Next, what was related to the seasonal scale and to the annual scale, in the temporal fish abundance variations, was examined. The annual component proved to represent a greater part of the temporal variations than the seasonal component. The seasonal variations are partly explained by the influence of discharge on the sampling and most likely also by seasonal migrations. According to the stations, the minimal abundances are observed at the beginning or at the height of the spate. The annual species abundance variations have been described separately in each station, using the principal component analysis. Contrary to the results of previous analyses by I-eveque et al. (1988) and Hugueny (1990), clear trends towards a decrease in the abundances of most of the species appear from 1975 to L992 at the stations of Border Bridge on the [rraba and Ganse on the Comoe. At Niaka on the White Bandama, the trend exists only from 1987 to 1992. In the first two stations mentioned, the latter period corresponds to an increase in the trend. The increase in the fishing pressure exerted by the riverine populations, which has been expanding markedly in recent years, could explain the trend observed from 1987 to 7992. A long-term trend has not been detected in the hydrology of these stations. It therefore seems unlikely that this environmental factor is the cause of the fish community trends. The influence of hydrology on annual fish abundance variations has been analysed by taking into account only the juveniles of the most abundant species. This choice was motivated by the abundant literature showing the effects of hydrology on the recruitment of populations. The correlations between the abundances per catch and the previous discharges show the existence of relations that are often strong but very variable within one and the same species between the stations. The positive relations between the abundances of juveniles and the discharge of the high-water months of the previous year which were expected appeared only rarely. Many correlations are negative, whereas a generally positive effect of the previous discharges was expected. Finally, strong correlations have often been noted between the abundances of juveniles and the discharge of many preceding years, with a shift far above their presumed age. Two explanations have been envisaged: either these relations are fortuitous and therefore do not correspond to a causality relationship between the discharge and the fish abundance variations or the discharge acts indirectly through previous cohorts. These results show that the influence of hydrology on annual fish abundance variations is far from being simple. Further analyses, taking into account more hydrological data and more information on the biology of the species and on the 14 in the abundances of most of the species which were detected in the three stations having the longest monitoring deserve to be looked into in an in-depth way. a' -a 15 REFERE,NCES Brocard, P. L,egendre & P. Drapeau, 1992. Partialing out the spatial component of ecological variation. Ecolory 73 (3): 1045-1055. Hugueny B., 1990. Analyse des donndes i long terme r6colt6es dans le cadre de la surveillance des peuplements de poissons des cours d'eau trait6s par I'OCP. Rapport OCP-OMS. l7 pp.+figures. Ilvdque C., C.P. Fairhurst, K. Abban, D. Paugy, M.S. Curtis & IC Traor6, 1988. Onchocerciasis Control Programme in West Africa: ten years monitoring of fish populations. Chemosphere. 17 (2) : 427-440. llvdque C., D. Paugy & G.G. Teugels (Eds.), 1992. Faune des poissons d'eaux douces et saumdtres d'Afrique de l'Ouest. Tomes I & 2. Editions ORSTOM/MRAC. 902 pp. M6rona B. de & D. Gascuel,1992. The effect of flood regime and fishing effort on the overall abundance of an exploited fish community in the Amazon floodplain. Aquat. Living Resour. 6 : 97-108. Paugy D., 1988. l^a reproduction des poissons du Baoul6 au Mali (haut-bassin du S6n6gal) : taille de premidre maturit6 sexuelle et p6riode de ponte. Rapport OMS-OCP. 24 pp. Rose K.A., J.K. Summers, R.A. Cummins & D.G. Heimbuch, 1986. Analysis of long-term ecological data using categorical time series regression. Can. J. Fish. Aquat. Sci. 43 : 2418-2426. Sabatier R., [rbreton J.D. & D. Chessel, 1989. Principal component analysis with instrumental variables as a tool for modelling composition data. in Multiway data analysis. Coppi & Balasco (Eds.) Elservier Science. North Holland. 351-352. Welcomme R.L. 1985 River fisheries. FAO Fisheries Technical Paper No.262.330 pp. oGERIA LoOor -,-SE LI co EENI I I I lr t AL tM -tEE I I G 't N E oIE E MALI ER L- I N uoI , \ I I o o Olouto ER 2 a t c l{lomty hnnr E lao \ 'l \tr H AII ERI \ I I I LI Coaolry Fr tatoun Monrovio fo on Figuro 1 : Localieation g6ographiquo dee 5 etatione choieice. Tablcau I : Dcscription de l'6tendue des donn6es disponiblee dane loe 5 stationa choiricc. Evilo 6trtion to Saryi Dornao L5raba Como6 Bandama Blanc N'Z Sarandrc Pont frontilrc Gane6 Niaka Pont, route dc Dabakala Pont dc S6micn 80-92 78-92 76-92 54-83 54-86 ol o2 o3 o5 o6 74-92 76-92 75-92 76-85 75-86 (70 pechcr) (84 pechcrl (33 p6choc) (66 ^6P'6 .!i o -65o- -zPIdz -o,!ro .o9465'F iE6 oa_o tE; T !-aOer*:3 {3.8sigl! c!a- P 3I E E e'3 "*'EBFP6a-9i:€E - o.= -. €:€E9ii sE€e!<89?i9 t Ed-l G.s E=-e n =OF=E96E c.g6 Et€ E > .o e; o -a_o =oE!dB .PE,i Ecr pJ €a!: TIF ..2 c.t z' -J ooEg66Fo C' o @ € o E oao! o oCCso c o G o o 6 a0 ot, o oF t ,33" t F.' -Nao.,-stN +d t l@qq't- FF(?Nc) o' 6 Ec ^oF@zz Q 3 zrzl?€JJgrD o o ^ OE EE€ii s€ss o @ to ({ ul o(o ro o @ if. (oI o n oo I lot @ (D- NNa)c{ddltI ao oG)Fu, UIq,to(l) c, cr.r,It tr, c.lNrolfdd lI o.a -stO, -c)- o. N?o:tt loFtNa)(, u) to to' lo -aD+o- 'F-(r @'N o'Noulul 'F+(otl tri +t uiorooo , -dc;'+o (, aE c ^oFOzztt 13 t z'z'?€=rJ ,5.o o o .oFEE€iiSfSS N lo u loo oI o I t \tl\!toooddil[ oo o)o Nt[il @(ottt G' CI' ll. lJ. qr@ F{ddllIgo o-.ul@.FttN(Y, N{'T.q: t--oN o'N o'N(\|-tIr ,N-ctri .. lo dooo orNdci o' EcPg ^oF@zz Q t z'zl?€JJ .9.o O o;OFFt€sgsfss u) st(o N rf' @ lt () o I o c) lt oN F (\to r. c!otdd II ocl !tiftd) NO il[ o(o(r) o) lJ- l! u)@ FNddlt IOCL t1\ ..Glsis o NC{ stOr- '-tc.rtt @ + q, dcr:ul , FlOG, o o c ^oFOzzlt{tzz'?€JJg.D o o .dFFE€i9 ,5*SS il tl('' tNtt N u) o o lo ul o u) No @ NN I (t, N (n IL o o il @ t\(, FN FNdd ll il o.o or\ -qa{F II c)G)NN(,(') Elt loo F.r,ddilIoo o .alN ,Gtq- cll(r) (t aN.i (N o6tt ,F++(o c)G,(' -.: -:qroo 6 E c ^oFr0zztl .tl B z'z'?5JJgiD o o "dFFE€ig sfss 10 oo(, (o o o) rt + G) o N + @oNG'dd Iltoo. NOtIONOq, --OIIil oootrtt l! t! lJ- tr\c.qooIuoo (o .f' o ul C{N lo ciN o + {t :.: N$tt++ ocroOTFNOto, fo, ooNd)Nto, i s g. g.i;oo+ : o I (t I oNN N ll. q o I ot oz z o e o o o ro oo! o o €c oc l(0lole LCc o! o o!, f EJo 8.8. ooIlt oo ttFoodd @at a\ a) rolo[il 00q,NNNN c, aa l! lt <Oe ooIil eCl oE c ^oFtazzltzzJJ ooccccoo oo L. station 01 't 2 3 4 5 6 7 I 9 10 11 12 mois station 03 ,H H 1 2 3 4 5 6 7 I910 1't 12 station 06 1 2 3 4 5 6 7 I I 10 1'l 't2 station 05 123456789101112 station 0? 5.0H H 120 100 .ts 80 €60!40 20 0 4.5 4.0 3.5 4.3 3.8 z I z 500 400 300 200 100 0 4.5 4.0 200 150 100 50 0 200 150 100 50 0 4.8 200 150 100 50 0 5.0 4.0 5.0 5.0 5.0 ).-l \ r-r,,\-l-r 401 2 3 4 5 6 7 I 9 10'11 12 Figure 2 : Variations saisonnidres du d6bit (moyennes mensuelles en m3ls . trait simple) et des effectils captur6s par unit6 d'effort (moyennes bimensuelles des log. traiB + carr6s) rr/ 0,8 0,6 o,4 o,2 0,0 4,2 o,4{,6 F2 I2 0,6 0,4 o,2 o,o 4,2 o,4 o,6 F = Iz o,8 0,6 L o,42 o,2 z' o.oJ e'.2 4,4 o,6 0,5 o,4 0,3 o,2 0,1 o,o{,1 4,2 o,s F - I 3 52652 station Ol 3 4 p6riode stauon 03 ttadon 02 3 4 p&lodc staUon 05 ,X 5262 1,0 0,5 0,o o,5 -t,o -1,5 -2,O 6 6 52 Fz I 3 3 4 5 3 4 p6riode p6riode station 06 --c- maille 15 rnaillc 2O X- mailles 25-3040 3 4 6 p6rlode Figure 3 : variatiors saisonnilres dcs cffectifs totaux moyens capturds par rnaille (en log) La p6riode I correspond A janvicr-fevrier, la p6riode 2 I mars-avril, etc. t- 1 1 1 1 Code Nom de I' 203 Polypterus endlicheri 501 gor Noto-pterus afer Mormyrus rume 9O2 Morm 1OO2 Mormyrops deliciosus 1 103 Marcusenius furcidens 1 104 Marcusenius sanegalensis 1 108 r io5 Marcusenius deboensis Marcusenius ussheri 1203 1 501 Petrocephalus bovei Hepsetus odoe 1 604 17O2 Alestes baremoze 1704 Brycinus 1 706 1709 Erycinus nurse 2402 Distichodus rostratus 2804 3001 C i t h a i n u s e_b y rg e_e 1 sj s Labeo 3003 Labeo coubie 3005 Labeo parvus 3203 Earbus waldroni 3403 Raiamas senegalensis 3605 Clarias 3702 Heterobranchus longifilis 3801 3902 Schilbe intermedius Scnnoi mandibuiaris 4502 4503 C l ry s ic hq! u 9 q i g rqt lg i t alu 9 chrysichthys maurus 4901 4vqllg_toe!_rl!: occidentalis5412 Synodontis punctifer 541 4 s+r o SynoQgntis schall Synodontis bastiani 5420 sqzt .S_U_n 9 d qnli s kctsnsrs _ Synodontis comoensis 6102 Hemichromis fasciatus 6401 Chromidotilapia 6501 Sarotherodon galilaeus 6507 Tilapia zillii 6801 Lates niloticus 6901 Ye Tableau 4 : Code et nom des especes prises en compte dans les analyses des variations annuelles. Brycinus imberi ll r70 y2 5 2 45 ro 4 5 '.\ 6r r t09 t 605 5 3 O,4 I -o,70 -o,lO N 0-() cPr o,6l au sur Figure 4 '. ACp des abondances annuerres moyennes par especePont frontibre sur ra L6raba (station or). pran aes espe-cesres deux premieres composantes principales, repr6sentant respectivement 38 et 17t des variitions tot.ales. station 01 - cPl '-" "' cP2 4,0 3,0 ' 1,0 r 0,0 74 75 .76 -1,0 -2,O 3,O ,02 v,o E .9(, c o, od, co00o E o(, -+ -- .: I -'--- + - -- --+_-_1--- 77 78 79 bp 81 83 84 85 'a6 87 8 89 90 .€1 '92 92 300 250 20,0 150 100 50 o I ffi ao0t . septembre f octobre t, ar) E 2 € ! 80 81 82 83 84 85 86 87 88 89 90 91 ann6e Figure 4 : Evolution temporelle du peuplement de poissons au Pont frontidre sur la Leraba resumde par les 2 premiers axes de I' ACP des moyennes annuelles des abondances specifiques (en log).D€bit moyen des mois de hautes-eaux (les valeurs nulles correspondent aux donn6is manquantes). 38 3 680J t70? o,2 o -q4 t - o.27 cP I o,68 Figure 6 z ACP des abondances annuelles moyennes par espEce A Gans6 sur la Como6 (station O2). PIan des especes sur les deuxprerniEres composantes principales, repr6sentant respectivement 3B et 153 des variations totales. N (L() ilo 30 39 4 54 6r station 02 - cP1 """- cP2 ,o ,0 -2,O 2toq) 6q o '= o oo c6a oc E o() 4,0 , 3,0 - O,O' + -- 74 75 76 77 78 79 'g0.. 1,0 3,0 -4,O -_+ ___,1__ +_ - +- 1-' t" 89 99 91 s2-'82 83 4 85' ffi ao0t ,- septembre I octobre e rt E E E 900 800 700 600 500 400 300 200 100 o I 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 ann6e Figure 6 : Evolution temporelle du peuplement de poissons i Gans6 sur la Como€ resurnEe par les 2 premiers axes de I' ACP des moyennes ennuelles des abondances sp6cifigues (en log) Ddbit moyen des mois de hautes+aux (les valeurs nulles correspondent aux donn€es manquantes). l', 1 -0,28 o,7 5 -gl5 N 0- o cPt o,64 a Ies 300r a 6 6 ?ao2 ? ,ror/ il09 r20l 3 5a a0 90. . Figure 8 : ACP des abondances annuerres moyennes par especeNiaka sur Ie Bandama branc (station o3): plan des espEces surdeux premiEres composantes principares, reprdsLntant respectivement 34 et 2Lz des variations Lotares. ) I ) I I l 4,O 3,O ' 75 76 t.7. -78 79 80 station O3 83 8s -86' I E9'90- 91 92 oo Eo o .E c oo cooo o. E o o 2,O 1 0 o,o - cPl """'cP2 ffi ao0t septembre I octobre -1,0 - -2,O -3,0 350 - 300 250 - 200 r 150 - 100 - 50, 0 -- ---- v, at E 5 E H l* l:* 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 ann6e Figure 8 : Evolution temporelle du peuplement de poissons i Niaka sur le Bandama Blanc resum6e par les 2 premiers axes de I' ACP des moyennes annuelles des abondances spdcifigues (en log) D6bit moyen des mois de hauteseaux (les valeurs nulles correspondent aux donn6es manquantes). fr 380 6r 3 33 o \ 5I 3 50 6640r-i t70 45 680 450 l706---- 650r o,28 -o,45 - O,95 c Pt o,67 Figure 10 : ACP des abondances annuelles moyennes par esp6ce au Pont de Ia route de Dabakala sur Ie N'Zi (station 05). PIan des espEces sur les deux premiEres composantes principales, repr6sentant respectivement 5l- et 15t des variations totales. N 0-() station O5 3,0 2,O : I 1,0 o o, E o,o o I -1,0 6 ?.r,o oo -3,o -4,O -5,O l- t- -+ l- -r - cP1 " "'-' cPz 68 69 70 71 72 73 74 75 76 .7.7 78 79 80 81 I '8ii 84 586 ffi ao0t septembre I octobre o(, E .= € E 600 500 400 300 200 il 1OO r -81- 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 ann6e Figure 10 : Evolution temporelle du peuplement de poissons au Pont de Dabakala sur le N'Zi resum€e par les 2 premiers axes de l' ACP des moyennes annuelles des abondances sp€cifiques (en log) D€bit moyen des mois de hautes-eaux (les valeurs nulles correspondent aux donn6es manquantes). o,o8 - q57 - O,50 cPt Figure L2 : ACP des abondances annuelles moyennes par espEce auPont de s6mien sur ra Sassandra (station 06). pran des espEces sur les deux premiEres composantes principares, reprdsentant respectivement 35 et 31? des variations totales. N (L (J O,93 t70 t? 300 ro l5 680 54r 5 3 390 6507ilo8 oa_=-_.] stltion OG 1,5 1,0 0,5 o,o {,5 -1,0 -1,5 -2,O -2,5 -3,0 -3,5 4,O 3t -9oc Et tG ,oIE o C' 69 70 71 72 73 74 78 79'.gO 1r283848586 - cP1 cPz I aoot n scptcmbrc I octoure a G, E € E 1000 900 800 700 600 500 400 3@ 200 100 0 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 ann(h Figure 12 : Evolution temporelle du peuplemcnt de pdssms au Pmt dc S6nrien sur la Sassandra resumtc par les 2 gemiers axes de I' ACP des moyenncs anntrclles des abondanccs sp6cifiqrrs (en log). D6Ut moyen des mois de hautes€aux (les valeurs nulles correspondent aux donn6es manquantes). J t r_t_ -- i + t + I I + + + 1? 11 t0 I 8 7 I 5 a o 2 I A/estes baremoze t102 dacdrg. (unirs) 123456 aulocon6ldion I 1 an d'inlaruCc p4,@1 auloconlleion I 1 rr d'hlenralr p4.01 Etyunus nurse 1709 dSoalegc (mnlor) 12 3 4 5I 6 Synodontrs Hra/l Sdld eulooonfldion l1 rn d'inhrveh p4.01 fop rara fop rarc 1? 11 10 I I 7 6 5 t 3 2 't station 01 station 02 station 03 station 05 station 06 12 1'l 10 I I 7 6 5 1 3 2 't 12 't1 10 I 8 7 6 5 { 3 2 1 12 tl l0 9 I 7 6 5 I 2 2 t 12 11 10 9 I 7 6 6 1 3 2 t 12 3 4 5 6 123 123a58 123456 2 t2 1l 10 I I 7 6 5 4 3 2 I 12 il l0 s I 7 I 5 I 3 2 t 12 1t l0 I I 7 I 5 I 3 2 t 12 ll 10 I 8 7 6 5 a 3 2 1 Figurc t4 : Conildiont ontr lcr ebondrnoo crphrltr denr lor rneillor do 15 ol 20 run .l ch.cul drr d5:bilr monruclt rtllricun. Sotdor ronl prirar .n oompl. lor orplca prlronbt drrr an molru 7OA dor p0o6or rl donl l'lutoconlldbn d.3 rbondenorr molrann.r rnnrrllor n'.!l pr. dgnilic.hl1r.(+:p0rrp4,6:++:r>0rtp4.01; a+a:rrooro4,@l:-:r4oroo,06:,-:r€dpd,01 ;_-_:ro.t9{,@1). ++ *t+ +?+++ _l_ t_-- -.t-.- - I I + --t--- t + ++ + + +++ ++ + + t^ _t + I , ,-tr. _-+-_l--ir-ii--i l- + i +- I -l _.1 _l I I --f'-'i- -r--- --- -- -li-it I - -{ I --t *'f I ---a-- I -i--- --_-i l-t- I j ---f - I I --.-- i.- -; +- + + + I rt ++ + + + + ++ +- + I t, _ts_-_l-- lt ----t ----1-- ti r,l 1l ++ + + " f--- + I i --L-.-.1. ll lil + --1 1; I i++ a Htrlbc mandbulais drrysrchtl4a maulus hJdrocynus bskah| Brya'nus maao/epidotus ?902 station 01 station 02 station 03 slation 05 etlooonrllJion I t mdlnlcrveh p€.01 et.tooonlldion I 1 rn dhblvrlo p{),01 .503 dSoelego (mn5cr) 12 3 4 5 E rlooonlh[on l1 m d'hlorudo p4.01 12 3 t 5 6 l60d top rarc autoconllelion I I rn d'inbn.l. p4.01 12 3 t 5 6 1? 11 '10 I I 7 6 5 I 3 2 1 l70il top r.r. 6 6 12 tl 10 I 8 7 6 5 1 3 2 1 12 tl t0 I I 7 I 5 I 3 2 1 12 tl l0 s 8 7 6 6 1 3 2 I 12 tl 10 s 8 7 6 5 4 3 ? t 't2 11 10 I I 7 6 5 I 3 2 t 12 It 10 s 8 7 6 6 I 0 2 1 12 t1 l0 I 8 7 I 5 4 0 2 1 12 ll 10 s 8 7 6 5 I 3 2 1 12 1l 10 I 8 7 6 5 1 3 2 I 12 3 4 5 6 l++ I + + I + tili +l I 123456 12 3 { 5 6 6 12 1l t0 I I 7 I I a 3 2 t 12 11 10 g 8 7 6 6 1 3 e 1 12 1t 10 I 8 7 6 5 I 3 a 1 12 11 l0 I 8 7 6 5 1 3 I 1 Flgm 15 : C.onlldionr cntr lor aboodu n3 ceptur6cs dnr llr meilbr dr 16 rl 20 rrn ol ctracun drr d5bitr mrnruols lrll6tLun. Soulos ronl prirca .n compL lor rsplcrs pra3.nt 3 dmr an moklr 70X dcr plctror .l donl l'rutocoralelon dor ebondencl moy.nn.l rnnurllor n'a3l pu dgnfc.f,va.(+:r>0olp4,S;++:rr0ctp<),01:f1:rrortp€,@1 ; -:r{otp<),6;--:14otpe,Ol;---.14rtp<),fr)l) I I t + lil I i---+-- ---i- -- ,.. -t1.1-!^-,-r - -.1 --- ilj -i- -; ;1- ': - i:.-i'--i + -1 _-i 1 -.-'i-.- -_ _.L -J I ___l_.__l____t + I l+l --+--- tJ j_J ++ + !+t -i---1 ;++-i - ;+i-- I + I i++ I I --l -l ll ""1--l-- +t ++ I I - -r._- -l --t I -t- + )r ++i ++ ----..1----- I + J*_ -1 i I + I I I -- -i I I I -T + --t--' +' I I I I i ++ + + + + + ++ -i i.-r-- L- T--i- ; - i 1 ,t-tl I I I i -+ I f--- i' ,+ -----l- | -1-.-f'--i I 1- --+---- + I --i--f ---i--r- --+-- i -t- -t--t- I I --1--- - l----+- - I I I I t + ' --f--'t I -i I .t i + r----!..--i + +--*+ t+ + l-l l----i --i'l+r '--*T----- -+-- .i__r_ti{-, --f-i--'--- --" .--f--i---*---- -. ++ ''f --l I + *r:-1- :---r- *- i'--i' r --i:-i --'' 1 --- l-- -;slalion 06 I i +*+ I i+ , 1_ I rll it+