Aquatic monitoring: report N° 3, section IV, analysis and recommendations

AQUATIC MONITORING , [r{ I t I I I t I I I' I t I I I I, I ,'c) lq I I OttcHocgRc IAs t s Corurnol Pnoe nAMME Report No 3 Secrroru IV Anauvsr s AND RecommrruDAT r oNS Fl '' September 1979SalfordUniversity of ONCHOCERCIASIS CONTROL PROGRAJVII'IE. AOUATIC lvl0NIT0RING : DATA HANDLING. THrRD REPoRT,, 17tu sEPTEMBEn 1979, PRINCIPLE INVESTIGATORS : ADVI SORS H.S.M, GENEVA RESEARCH OFFICERS I I a a TECHNICAL ASSISTANTS DR C P FAIRHURST DRRDBAKER I DR S FROST DR M PUGH THOMAS MRS M LOSSKY MRS M S CURTIS MISS E LEE MISS F M GRIFFITH MR P A HOLDEN RECoT'TIIENDATToNSmSECTION IV ANALYSIS AND M [ | 'fqqf; Acknowledgements: The presenters of these reports would like to acknowledge the invaluable help of l{r J D Marr and Miss F Baker of O.C.P., Geneva, Mr J DuppentJraler, Mr J Dupuy and other staff from H.S.M and E.D.P., Geneva in providing background material and help in gathering information relevant to the project. At Seilford, the Department of Biology and the Computing Laboratory have provided every support possible to aid the work. Mrs J Annis, Mrs B Cheadle, Miss T McKie, Mrs D Monks, Mrs J Sastri and Mrs E M Williams have ensured the successful presentation of the reports. The World Hea1th Organisation has provided financial backing for this work. Its \s \-- .a :l-) 'r \-f c '/'!.5i I i' : , rt . - / "'" L-tJ-\'-'-L' Ia.rst l''.1..',/' x-Je 7ro !<t leht. 'o i<) ) . ta,'Lc tcc l9.Lc lH!'- -:. i,'it - o l*J.'cl -r rD I\ or__.... -), a .P- r' = ,/tE,/ ?t ^ u.l E o ! o 7E\e'no> k,, t o N oo oN o @ 3 H/=/- )o ) :SB3 ; q ) s"/I}to- I te.\ 6 --n'E ,Pg g &oo<05?EB &,Hd a)vu,d;F =oO ,S-/ZA-- an - i' -;" .\.--.--!-'-- \ 3 H ..r.p o al 5tt.. s:Hi r.t$n& /; i --H,, -trP l Flpooee trl LU ul ca ul iE E(, oE o.ga F tJ-l -2 r-o'-oa ./, -? o oz.Erlr -)ono=rZ 5a il T\He ? :tr{ (5 .t,trr-1 ; oHa .* j ,", g @Ie !n !n r^ : n eo 1"' t^ * .", p t * U - P f i : HH ' F\ N . rrr 6.t \o tr . r-r r-{ r-{ F{ .:n.OvNClOF{pH 4 N @ c.t r{ ca (n crr F{ c\tO.H (, Ft 19 ,i<4 t c oc)qrodEtzr'. z&OHIrI(,zuOr.1HH€Z .$XX N6rOO trl F{ ; r{Hl.\Or@.S FlONl{fr co dtl':8 (n},g .d.Otr\OOO fEI N c\f Cq Fq $F.l q, Jdqt € oAd --t{ooEt,o\oo'd.;F{.r{qlfJlCH+{tg,(t)\6roi,tE_qo+.fu)op{o og..{p_.nr.l rl-.iooqoo.go.ilh ol o }l'd _ op !u r{ .r{ E o E d E .Ur: g 6 .: eE t b..d H j ; fl E E.is IEr o o s{ tr tr, tr .o o. n E E E i n Ei ? e -o 6 .F{ .q -o -o q o E .o t,o rE o o o tr r6i 6 6B. Fa Z E a{ A{ M A A t u, Fq p" i n rE u) ii Z, G., Fl \ra g dOtr Eq s Es *P$-FrF{;o;;qrvtoHt{ooonook!d r. EE It*.!?i ;n;E.gt.B?o \o q q q, J b o, gl,rl 5 j4 J 5 6 6 ;S o r{ Iq o g- $ +r +r E - T t o ,6;)q E E 6 ! "{ trEE E ds Ess H * H E 'r c: Fq Fa Z crl F.l E B B o Fa & Fq pq ji fi X ;, @@ooNf\N Cft \O \t r\ F\ l.)(\t HzFltrl B rqt(4ES H H H H H ts H E H E H H H H H E i ts HrtIA H H&H t\r{ \O lln F{ 6l (f} \+ !o \o t\ @ 01 A Ft c\r (n .$ rn \o t\ oo o\F{Flr{r{F{F{F.lFl;; +,qroP doLoo -;Px$Hhq,o r.' e o rl. dC)9g===o.od 'H () 5 H c, tf) \o rr -lN('t.+ Hut F{dod ffsE*A<>Td!rEHC, u)l fEl Iolol :l HI2l EI otAzHHE]caOo(.) 1. ANALYSIS OF INVERTEBRATE DATA The data set comprises uonthly samples, with replicates at a maximum of 19 sites, using four main methods; of up to 60 taxa and 5 abiotic variables. The sheer bulk of the data is the biggest obstacle to understanding. S:atistical analysis comprises a large number of tools (multivariate analysis, time-serles analysis, etc.) with many transformations, and we must be careful ln our choice of appropriate technique. our work so far has demonstrated interrelatlonships and posed further questions of the data set. The comparison of aiiterent sites, different methods, and some understanding of the sinilaritles between the 60 taxa is inevitably of value. Cluster analysls has resulted ln dendrograms of taxononic E tructure. These "tree" diagrams summarise the relationshJ.p between the taxa in a compact way and flgures 1 and 2 Lllustrate the reEernlclance between species based on tJlelr seasonal distribution ln night drift and substrate. These simllarlties wiII change when we fully consider the value of the "other" categorles and taxa rarely recorded. Also, figure 3 represents the relationship between samples based on the conmunity structure rather than any lndividual group. Ordination is another technigue of cluster analysisl it reveals the relationships between the taxa in the form of a scatter plot, t{here siml-Iar taxa will be clustered together. More is written about these techniques and the use of our cluster analysis package PODS, in the PODS user guide appended to section II of-this report. The months at a given site can also be clustered, giving an ordination and a dendrogram, so that the similarities betwqen the months can be seen. Flgure 4 is drawn from Danangoror on the River Maraoue; (site 8) where attention can be drawn to the position of all post treatraent samples. The work so far done on individual sites reveals that the first principal component or t'factort' only accounts for a small fraction of tJle total variance of the data. This should improve when further work on transformations has been aone (oide infra) and when certain taxonomic clarifications have been made. Cluster analysis, W tre way, links up with factor analysis, in that the ordination axes are factors. Seeking an understanding of the meaning of these, we see that the first factor loadings on the taxa have nearly all the same sign. we can interpret this factor as "general species abundance". The use of standardisation, and the acquisition of further abiotic data will elucidate the biological meaning of the factor loadings. ,When aII the si-tes are included in the analysis, a dendrogram and ordination of sites will reveal clusters of sites. Thus representative sites from each cluster can be chosen, and others isolated for special attention. This is cLearly of great usefulness. Figure 5 is not drawn from the current data set, but from results of a survey carried out in the Volta Basin by Drs Frost and Leveque in 1974. Although a simpler method of analysis has been used in this case, the groupings have been identified by the dominant taxa which are characteristic for each cluster of sites. A more accurate way of assessing the taxonomic individuality of sites and samples is shown in figure 6. An example from this site concerns months 1-4 (May - August 1975). These samples differ from the rest in having more than average Tanypodiinae (46) and less than average Libellulidae (2t') and. Hydracarina (55). The potential of t"rris technique on an appropriate data set cannot be overemphasised. Both the Shannon and Brillouj-n diversity indj-ces are included in the data base. The latter is often preferred through having less bias with Iow nunbers. Many other indices need to be tested for their application to such a data set. The use of the Taxonomic Hierarchical Diversity has a possible role if strong correlations are found between familj-al, generic and species diversities, and an index of comrnunity structure is the main requi-rement. Then, the specialist taxonomist is not necessarily required for the routine monitorlng. Such formulae may be modified to indicate the niche breadth of a taxonomic group before and after treatment. Then, the object is not to estimate the dj-versity of taxa in a particular site,/month, but the diversity of site/nonths for a given:dIr Deseasonalisation has been usefully applied, to Show the fluctuations of diversity about its seasonal norm. This would consist merely of sr:btracting the monthly mean, if the measurements were made at tlte sane date each year. In practice, a more complex method has to be used. Good examples are given in figures 7 (Neoperla) and I (Simulium dannoswn) for the control site 8. In futurer aIlY factors having a biological meaning, can be graphed and deseasonalised, and used as indicators of changes in the ecosystem. 3. Deseasonalisation has been mentioned as an aid to seeing tocal changes in structure. The deseasonalised (spearman) correlations have been calculated for site 8, between taxa and taxa, and taxa and abiotic variables. The interpretation of these is a biological task; their statistical meaning is that a positive correlation exists if one variable is greater than usual for a given month, and the other variable tends to be greater also. These correlations differ from the naive correlations used implicitly in the cluster analysis, in that two taxa are now only correlated if tJley behave similarly in response to changing conditions, not if they have similar seasonal crycles. We can see, for example that pairs showing strong correlations are tlte Baetida.e and Chironomini and, fite Philopotanidae and, S. dannnoswn. As this relationship exists before treatnent and these taxa have differing life cycles, excellent indicator ratios to gauge treatment effects may emerge. Lagged correlations have also been calculated; i.e. these are correlations between two variables where one has a lagged value, ( its value one month or more prevlous to other variable). A positive value means that if one variable was greater than usual last month, another will be greater than usual this month. The idea of looking at Iagged quantities is t}at biological systems nay take time to respond to changes. Hence, abiotic factors including Abate uay be tested both for their direct and su.bsequent effects on the fauna. A particularly interesting correlation is that between a variable and its own J.agged value. This is an autocorrelation. Some taxa (e.g. Ecnomido,e) are hlghly correlated from month to month, t.e. they wander in big swings about the seasonal norm. Others are uncorrelated, in that they perhaps adapt. quickly to currently present conditions. Presunably one use of this difference could be to pick out slowly-adapting taxa as indicators of say, Abate effects. An extension of these procedures is in forecasti* an" nurnbers of a taxon, particularly useful if changes in treatment or management are envisaged, or to test experimental methods carried out in the laboratory or outside the Volta Basin. Another important point concerns interpolation when there are uneven sarmpling intervals. 4. t"lany of the ideas given here are speculative. one immediate task wiII be to find ttre best transformation of the data that stabilises the variances of the taxa. At present loglO (1+x) is used, where x is the taxon count. This does a lot of good, but is not ideal. The advantaqe of logarittrus is also that any multiplicative model of taxon nuulbers becomes additive on taking togarithms, whereas some other power of x may not be simply related. ebiotic variables may also need to be transformed, e.g. Secchi disEreading is a measure of the clearness of the water - it would be interesting to look at a measure of turbidity r e.g. the reciprocal of the present reading. Much remains to be done, and the usefulness of the various statistical- tools available wiII only be discovered through experience. 2. ANALYSIS OF FISH DATA The length-weight relationship of fish is usually represented as , = .Ib where a is a coefficient characteristic for each species. b is an exPonent between 2 and 4 typically 3 for isometric growth. If a value of 3 is assumed and growth is allometric then the coefficient changes with length. This can be seen in the data set under question and the correct power should be obtained for accurate analysis. So far, the fish have been studied purely to obtain graphs showing values for p.U.E. and coefficient of condition. Some attempt has been made to relate variations to river conditions but the data on the invertebrate forms is not necessarily applicable. Hydrological information is needed for the rivers and times in question. The multivariate analysis techniques mentioned above may also be applicable here, to e><plain the variations in greater detaitr- 5. RECOMI4ENDATIONS An index of drift is essential to the full interpretation of the data set. We are not satisfied that the formula presently used is an accurate reflection of reality. In some sites, to divide the taxon counts by current speed appears to over correct. It is reasonable to assume that the rate of flow through a net and hence the rate of capture of organisms is not the same at the start and end of, say, a 30 minute period. Hence to conbine short and long samples by simply dividing by the tine period wiII lead to inaccuracies. Therefore, standardisation experiments need to be carried out to establish the right correction factor. We are aware that some of this information may already be available. Sinilarly the relative abundance of fauna collected by the substrate methods is weII known. Eor^reverr the rate of recovery from disturbance is not so clear. Again, further standardisation may be needed to aid interpretation. In the Ivory Coast, the common species of fish are sufficiently well represented in each sample to enable an accurate coefficient of condition to be established. In Ghana, although the more regular sampling leads to greater accuracy, the nurobers involved make a reasonable calculation difficult. If less frequent sampllng glves tlme for more measurements this would be an advantage. See table 1 for a monthly comparison of fish nnmbers at some Ivory Coast and Ghana sites. (L702 = Alestes batemozel L7o9 = A. r.tuxsa; 3902 = Eutropius mentalis; L2o3 = PetrocepVnlus Bouei; 5415 = Sgnodontis ganbiensis). To complete the analysis of the fish data, it original 5623 raw fish data be retrieved for 6 and 7. is i-mportant that the 1975 and t976 for sites 1, 2, In any survey of this type, a compromj-se has to be made between the conflicting objectives of a wide range of sites and the detail at each place. If manpower is restricting, one site for each treatment phase (I, Ia, II1W, IIIE and IV) may be sufficient. At other sites considered important, a brief but intensive saupling period in the middle of the wet and,/or dry seasons may adequately reveal year to year trends. If new sampling sites are chosen, the ideal is for a pretreatment period as we have in Site 8. 6. t, .t{ rn the rvory coast and Ghana, sampres are taken per month whirst in uPPer Volta records are every 4 weeks. If treatment schedules are on a weekly basis, the latter sanples will have a fairly constant interval with treatment. This has the advantage of reducing a possible cause of variation and gives good long-term trends. However, a variabre but known period uay in fact tell us more about the medir.un term effects of Abate- Here it is essenti.al that standardisation e>rperiments, using current protocol, sample the change per day over say a week from treatment. Information over 24 hour periods is already available. Thus the incorporation of treatment data and whatever hydrologicar information j-s available, is a high priority. Should any further monitoring exercises, such as chemical analyses and aguatic productivity, be instigated, this information should be compatible with the exj-sting data set. 4- SUMMARY The work incorporated in these reports is summarized on the next two pages. Ehe existing data set has been clarified, reformatted respect to correspondence and notes held at Geneva. has been made for errors and the correction of these in the flow charts and recoded with A complete check is nearing completion. For invertebrates, a semi-automatj-c data handling system has been established for inconing data and produces a variety of reguested outputs. For fish, a simpler system is being developed. The analysis carried out to date underlines the need for standardisation experiments,to apply the appropriate correction factors. OnIy then can the tremendouspotential of this data be realised. The size and complexity of such long term monitoring informati-on, means that an overview can only be obtained from appropriate multivariate analysis. It is our opinion that no firm conclusions, regarding the effect of Abate treatment in the Vofta Basin, can be made until this work has been carried out. FLOh' CHART FOR DATA HANDLING INVERTEBRATES GHAl{A EXISTING DATA SET IIIa IIIb INCOMING DATA SET REFORMAT I RECODE Treatment Hydrology Cl imate RECORDING ER-ROR I S,rU"traEe colonisation BIOTIC I Oritt standardisation I fre and post treatment VALIDATION CORRECTIONS MULTIVARTATE ANALYSIS ABIOTIC SITE FILES FLOVI CHART FOR DATA HAIIDLTNG IVORY COAST GHANA (" VISUAL CHECI(ING PU}CHING 10. FIGURE I CLUSTER AMLYSIS . NIGHT DRTFT TAXi Danangoro - Maraoue (site8). .D t1 R\ sa N $3 a @ a ct) F I I 6 F ,o, r GD B 8 CO ua R .D a1 68 13. F C' n &l t€({ ait afi +a a s a (r, att et ct a\a a R a CED te#&€ S llrJ-Lt, &*C\AVF2 sdur Eils umilirs t I I I I CL:c=DcrCD -i.i-ft FIGURE 2 11. CLUSTER ANALYSIS R & rc r a (! tD r a tsN ? oUA oa t8 = 6 g F ur({ (D G g 88 ?o(\t !c(iUtF e5 i- B iF @ ut1 t^a a?) o) CD t8 . SURBER SAMPLES TAXA Danangoro - Maraoue (site 8) cB-e'e -GxEEe'i-T? * Hjtrt tH t{ EJ t{l+ui eca Fq $3aqO 5 6MDFL_-,O <-trr(\. cYa\i F-Fa@(r) cD (ED t$l+JSiB aD (D O O,Ct HTIH#\mB Eilx ffiuruuslo t.I) aJqs L2. FIGURE 3 CLUSTER ANALYSIS . DAY DRIFT MONTHS Dartangoro -'Maraoue (site8) lq g r (r2 K a.) $B 83 @ t8 c{ 6 c..t lr2 a-6 E8 a{ N .Y @ tn K!9 -| C @0 -E. :e TI N N ts (\ = €o Ct o6r = = (s .E (D <tt Cr cD CD CD .ErsrcD TT .i-(]{ tH $,, H !:l !'-'+{= tlil'{ br*l*li<:g y,it3 $msffiilxu6ni,,il.;tu- q, qHq qq ==qq Ehls 13. I EIGURE 4 CLUSTER ANALYSIS - NIGHT DRIFT MONTHS Sabari - Oti (sire It) ts:(il l)E o +G t* 8= !3 €\{ I x@6tr^ :: a -N-' = .rt e< -' H2of;ta ,< - U' HzoE HzEl E trl& F{ F{ U)oA{ z LB0dtu]ydt}{tu FIGURE 5 % SIMILARITY DENDROGRAM DRIFT DATA FROM FROST AND LEVEQVE 1974 l I x6 v€ az .$ Hd td l{N lz('l dl al EI dl5 OIF.l .f oo o ou o a !U E o0 - € o o aoE o o ok! o a ,.0!A!oO! .Eo od oH (Jd a o o o q kO o @ k !q EN G lc4, I -! ai ' 3 EN-..__-_-*- o o o o O(n (:, N <o <N <d Qo- lko fai N (rN-- q a ! q o 3o 'J o o a o o ! o akU 7 o ..1 q ktN a.l - od QN ! a jcool NN@{c o o "o c,, s 9A o a ! E q ^l@jc.oN@-aq9d@so@N€i @ i6 - r r\ e o .E 6 4 -, -, -, o o N 15. DIVISIVE DENDROGRAM. NIGHT DRIFT Sabari - Oci (site l1) FIGURE 6 N 6 t8 ;; 6 a 8l ;n IJ .Ja- N @ e <Y) l-r) LD a v) r.< sa- U o --r ;H = € o oaa @ a{ = = = :? rN <D (., c=t o cD + TH YJgJco L1 sm8nrs,trx uiillliS'i dr o tn FIq]RE 7 G]t @FIL Hu cl F :tr(D Hz o(r)cl UJ C" a-aJ z(f(n H<]U cl o2 J z c)a fi>a @ IUF .H a a -FzOP$lb z IUE t-- C9 o I I lr1, a c9 I o a o to a o s a n a c9 a GI { .-<- -<- B( -+ ( +e t9 B{ D X D> ++ + D{ D{ *-+ +- .<- --+- -<. + ** -+ +>- C9 6t NOXVI UOJ IJIUo JO At90'] akr TIGURE 8 ot (tFILH u cl F :tr(tt t{z a :EFzOP(\l- z H I,UE HF o(Y'c]lrJU' l-lJ zC'a H< u,o clz J zo @ fiDa (I) TUFH @ o c9 I I l/:l a o I o a o ln a 6) o a a a o a$l _-_<_ <t0 --<- { -<-. { + --+ --{- .+ -+->- --{>"-- --{- *fi>.- € t, {- -+tr -<- t>- * -+- -+ D -+ --+- --+-- + --+>*- -* o ( HnsoNHVo HnItnHIS )Zr N0XVJ- J0 A t90't 18. Table Species overall percentage 5623 Monthly catch at SITE 3 L702 L709 3902 L702 different Sites. SITE 4 SITE 9 SITE 11 L709 3902 3902 3902 1203 5415 48% s ot74 N D 340 185 L27 730 28 18 33 13 365 505 33 84 L4 280 46 11 J r M A M Jt75 J A S o N D 4L5 0 L22 29 1s5 23 230 99 42 39 67 58 109 200 35 29 59 18 605 130 0 267 249 L4 J F M A Mt76 J J A S o N D 330 113 563 135 24 76 73 42 338 0 24L67 15 s6 145 308 28 28 286 301 18 225 2L4 22 620 37 4s o o 1 4 2 o o 3 0 o o o 0 o 'I J F M A Mt77 J J A S 0 N D 181 96 29 44 L28 188 2L 115 30 42 30 89 353 73 15 100 100 100 224 4L4 46 46 46 46 0 o o o 66 75 33 8 4L 66 o 10 o o 28 t28 L7 37 9 4 o o o o 0 o 7l 5 o o 2 0 4 0 o 4 2 1 3 1 2 J Ft78 M A M J J 24 L2 36 36 38 13 4 o 0 L6 oo7 OL22 oo5002 31 L7 56% L77" 8% 2L% 4% 20.5% 2t% 6% 5.5"/, Ja I I I t 19. Appendix Taxa and month codes used'in figs. L-4, 6. Site / Sample Type O1 Coelenterata 02 Nematoda O3 0ligochaeta 04 llirudineaO5 Cl-adoceraO6 Copepoda 07 Ostracoda O8 Potamon O9 ltacrobrachi ,,rl 10 Caridina 11 Baetldae L2 Caenidae 13 LepEophlebiidae L4 Heptageniidae 15 Trlcorythidae 16 Ephemerldae L7 Ollgoneuriidae 18 Other Ephemeroptera 19 NeoperLa 20 Gomphidae 2L Libe1l.ulldae 22 Zygoptera 23 Other Odonata 24 Cal.amoceraEidae 25 Eenomidae 26 Hydropsychidae 27 Hydroptilidae 28 Leptoceridae 29 Philoporamidae 30 PoLycentropodidae 31 Psychonyidae 32 Rhyacophilidae 33 Serlcostonatidae 34 St,enopsychldae 35 Other Trichoptera 36 Croixidae 37 NoEonectidae 38 OEher llemlptera 39 CeraEopogonidaetfi Chaoboridae 4L Simulidae (Others) 42 Slmulluur damnosum 43 Chlronomini 44 ?anytarsini 45 Orthocladiinae 46 Tanypodiinae 47 Other Diptera 48 Dytlscydae 49 Elmidae 50 HydrophiLidae 5L Gyrinidae 52 Other Coleoptera 53 Pyralidae 54 Sisyrldae 55 Hydracarina 56 Gastropoda 57 Bivalves 58 Amphibia 59 Pisces 60 Orhers 8/L 812 8/4 LL/Z Dare 75 l'{ayrr June July ,r Aarg rr Septtr OcEIt Nov It Dec 76 Jan " Febtr Marchit ApriLtt Maytt Junet' JuLytt AugIt Sept rl Octtr Novir Dec 77 JanIt Febtt March 't Aprl1rr lvl,ay rr June rr JuLyt, Aarg rr Sep t " octlr NovIt Dec 78 Jan rr Febtr MarchI April rr I,Iayil June rr Julytt Aarg rr Sept " Octrr NovIt Dec 1122 3 43 546s 786 97 108 11 9 L2 1013 11 L4 L2 15 13 t6 14 L7 Is 18 16 L9 L7 20 18 2L 19 22 20 23 2L 24 22 25 23 26 24 27 25 28 26 29 27 30 28 31 2932 30 33 31 34 33 3s 3436 3s37* 36r' I 2 3 4 5 6 7 8 9 10 11 L2 13 L4 15 L6 L7 18 2 3 4 5 6 7 8 9 10 11 L2 r3 L4 15 L6 L7 18 19 20 2L)2 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 37 38 38 39 I 19 20 2L:s 22 23 24 25r, indicates miscoded site 16. { t; t