't'{, l( - Impact monitoring of Forest Management in Cameroon

Ministère de I'Environnement et des Forêts, UTO Campo-Ma'anEsitimating elephant density and distribution in the OperationalTechnical Unit CamPo-Ma'anA manualCAMPO-MA'AN DOCUMENT NO 16Projet d'Aménagement et de conservation de la Biodiversitéde CamPo-Ma'an, CamerounF{'r} ifl'j g;@WUR#hlmmNIISWÛ'I{\!'17,'zKribi/ I2001't'{, l( )

CAMPO.MA'A}{ DOCUMENT NO 16Estimating elephant density and distributionUnit Campo-Ma'anin the Operational TechnicalA manualPatricia D.B.M. BekhuisU'agen i nge n LIn wersi t.v, trtr'age n m ge nCc rapport est le résultat d'une étude exécutée dans le cadre du Programme GEF-Biodirersité Cameroun. Projetd'Aménagement et de Conservation de la Biodiversité Campo-Ma'an. un projet du MINEF.Financement :Maître d'æurre de cette étude :Exécution de l'étude :GEFÆlanque Mondiale (Don FEM n" TF28673CM)Gouvernement néerlandais (Don DGIS no. TF20957 NETH)Gouvernemenl camerounaisOrganisation Néerlandaise de Développement (SNV)Tropenbos International (TBI)Université de WageningenMINEFMinistère de I'Environnement et des ForêtsiRùTlsc5 |\iRArnN{Tropenbos InternationalWURWageningen Universitl- and Research CentreProjet d'Aménagement et de Consen'ation de la Biodirersité de Campo-Ma'anBekhuis. P.D.B.M. 2002. Estmating elephant densiq'and distribution in the Operatonal Technical Unit Campo-Ma'an: a manual. Campo-Ma'an Document No. 16. Tropenbos International / Projet Campo-Ma'an / WUR.Projet Campo-Ma'anB.P. 219- Kribi. CamerounKribiAilageningen, 2002

Estimqting Elephqnt Density ond Distributionin the UTO Compo-Mq'on,A MclnuqlPaticia D.B.M. BekhuisLatge Mammal Program in the UTO Campo-Ma'anBiodiversity Conservation and Management Pro iect Campo-Ma'anKribl 2002

Proiect Campo-Ma'an Larqe Mammal Proqram Patricia BekhuisAcknowledgnentsThe Campo-Ma'an Project, specifically financed by the World Bank and TropenbosInternational, supported this work. Special thanks to MINEF for letting this happen. To thedepartment of nature conservation and ecology of vertebrates of the University ofWageningen in particular Profl Dr. H.H.T Prins for his support with respect to content andmoral. Much of gratitude also goes out to the two fieldwork assistants Medjo Imocent andMenye Menye Justin Elie for their perseverance in the field and for sharing their fieldknowledge. Big thanks for all others that supported the bush: the cooks, Obam "Dr. Ganago"Elisé, Medjo Medjo Casimir, Essola Blondine and Ondo Daniel and the drivers, AmougouThomas, Aba'a Aba'a Louis, Mbonda Hubert and Asso'o Jean. Appreciation goes out toChristiaan v.d.Hoeven for his useful comments, the GlS-section and last but not least to thePrincipal Technical Advisor of the Project, Mr. Tinus de Kam.8.P.219 Kribi Cameroon TellFax +237 346 21 37

Proiect Campo-Ma'an Laroe Mammal Prooram Patricia BekhuisContentsAcknowledgement2lntroduction4MethodBackground for method choiceFieldworkDung counts on striP transectsh-rput of fieldwork data in cotnputerData analysesCalculating elephant densitYStatistics557991010t2Reporting the resultsInterpretation of resultsFormatting the reportL61616Protocol18References cited in manualI98.P.219 Kribi Cameroon TellFax +237 346 21 37

IntroductionIIIIIIThe purpose of the manual is to help people who lvill continue the work on elephant densityand distribution in the uro- It has been written as comprehensively as possible; however,some basic knowledge of biology and statistics is presumed to fully understand it. The manualwill explain the whole course. In chapter one the method is explained, starting with thefieldwork until the statistical analyses of the data. chapter two will he$ to interpret the resultsand chapter three explains how to convert results and interpretations into a clear report. For aquick overview of the whole procedure read the last chapter called protocol. This chaptersummarized the work about how to get to interpreted biological results useful for managersand other interested parties. This manual is closely linked to the report on elephant densityand distribution in the southern part of National Park Campo-Ma,an (Bekhu is,2002).8.P.219 Kribi Cameroon Tel/Fax +232 346 21 37

IIIMethodBackground for method choiceKnowledge concerning the size and density of a population is a vital prerequisite to managingit effectively. If answers concerning abundance could simply be obtained by counting theanimals, the method would be simple. Unfortunately, counting animals in a large forested andmountainous environment, Iike the Campo-Ma'an area is impossible. If direct methods i.e.counting animals don't work, indirect observations can be the solution. One can ask a numberof local hunters about the number of animals they think are present. This method is called thepooled-expert-opinion-method and described by Van der Hoeven (2002). Or, as described inthis report, one can do fieldwork with indirect observations; in generally this implies countingdung. Counting all dung in the UTO will be inaccurate, expensive and will consume too muchtime. The alternative is sampled counts and extrapolation.Extranolation: adding new values outside the extremes using knorvn valuesSurface of sample (on2)Surface of Park (km2)Five common used methods for sample counts are:1) mark-recapture 2) point sample 3) plot sample 4) strip transect and 5) line transect.l) With the mark-recapture method, a sample of the animal population is trapped, marked andreleased. After a while, a subsequent sample is taken to estimate the ratio of marked tounmarked animals in the population. With this data the size of the population can be estimatedbut it doesn't give reliable information on the distribution and/or dispersion of the population.B.P.219 Kribi Cameroon Tel/Fax +237 346 21 3Z

IIIIIRecapturing large ntammals, like elephants, is difficult and sometimes dangerous and mostprobably would give a sample that is too small for accurate results. This makes the methodtotally unsuitable to identifr elephant populations in the campo-Ma,an area.2) Point samples have been used near saltlicks or waterholes. The assumption is that allanimals of a target species will visit that place. The observer stays in one place and recordsevery animal visiting the point. Estimates can be obtained with this method but double countsmight occur. Further more, like in the mark-recapture method, the method does not giveinformation on the distribution and dispersion of the animals.In Central African Republic, research on elephants by direct observation at an open spot in theforest was executed. The researcher observed the elephants for 8 years from a tree hut andknew all individuals differently. That way double counts were avoided. In the UTo CampoMa'an the two main assumptions of this method can not be met, l) no particular place wasdiscovered where all elephants nreet (no saltlicks nor savannas rvith good quality food) 2)nothing is klown about the area surface in which these elephants go round.3) Sampling a plot, counting all dung within a quadrate, has the advantage of simple estimatesdue to a constant size of the sampling unit. However here is the assumption that the animalsare evenly dispersed over the whole area. If this is not the case, a quadrant is likely to land ina patch of high density or a patch of low dersity. Due to different environments within theUTO and the animals' behaviour, their distribution and thus dung distribution is probablyclumped. Therefore, plot sampling is likely not to be accurate at all.q e' , A sufficient long transect is more likely to cut through areas of high and low densityof dung, making estimates much more accurate than do quadrants.Two types of transect can be distinguished;- 'bounded' transects also called strip transects (4)- 'unbounded' transects also called line transects (5)Bounded transects are transects where data is gathered within a fixed range at both sides. Allobservations made beyond that distance are not taken into account. Bounded transects havethe same advantage as quadrants i.e. a constant sampling unit size. The strip transect doesassume that all dung within the given strip is counted; therefore the transect width should beadapted' Where as the line transect does assume that transects are walked with constant speed8.P.219 Kribi Cameroon Tel/Fax +237 346 21 37

IIIIIlProiect Campo-Ma'an Larqe Mammal Prooram Patricia Bekhuisand attention. Non-trained locale personnel should therefore not execute the latter as theygenerally fail to be attentive for several hours.If observations are rare and every track seen needs to be included in the data set in order tohave a sufficiently large sample one may use line transects. With line transects only the dungseen from the census line is counted and the boundary is estimated after fieldwork. Eventhough no method is perfect the line transect method will give most accurate estimates forcounting rare mammals in a varyng forested area. However, if abundant throughout, likeelephant dung in our case, the strip transect method is easier and a good choice.LliStrip and line transect visualized,Bounded (strip) transectseen from above.Un-bounded (line) transectCensus lineStrip widthFieldworkTo result into accurate estimates, transects should represent the census area and thereforepreferably cross different habitats and different levels of hunting pressure. Systematicsampling instead of strict randomly is therefore a better option. To make sure transects cutthrough different habitats, they should preferably be perpendicular to rivers or other largeremarkable features like hills, roads etc. However, if possible the already existing transectsshould be used instead of cutting new transects. These are and were used by other researchers,making it possible to compare data over time and between different subjects (vegetation, soil,animals etc.). The location of the existing transects has been visualised on a map (see ReportBekhuis 2002).8.P.219 Kribi Cameroon Tel/Fax +237 346 21 37

If new transects need to be cut make sure that transects do not become paths, since this willIgive biased results. Elephants might be attracted to the transects and transects can no longerrepresent their surrounding. When in the field, transects should only be visible ôr the trainedhuman eye and certainly not for the elephant.IThe number of transects you have to monitor depends on how heterogeneous the environmentIis from the point of view (in our case) the elephant. If the environment would be absolutelyhomogeneous theoretically one transect will do to extrapolate the results. In practice theIenvironment is never homogeneous but at the same time the capital you can spend for yourfieldwork has a limit. The rule for the number of transects you monitor is the more the betterwithin the means you have- If you want your data to be accurate make sure you cover alldifferent areas and if you want your data to be precise make sure you monitor as many aspossible.Transect lengths can vary a lot. Which transect length to take, again depends mainly onpractical reasons. If money and time are limited just a few long transects can be monitoredonce to get results, par example I time 5 transects of 20 km as done by Blom et at (npress).However much more precision and accuracy will be get by repeatedly monitoring of severaltransects. This way the decay time of dung can for the most part be avoided (see further;calculating elephant densilies). A feasible transect lengtll used in former years, is 5 km. Thislength can normally be surveyed in one day.The width of the strip transects depends on the density of the forest. If undergrowth is sparsethe width might be 5 meter both side of the census line. But if vision is taken away,the stripwidth should become nanower. Always decide upon before starting monitoring, do not adapttransect width while walking strip transects.As said before, repetition of monitoring results in more accurate and precise results. Repeatedmonitoring, upgrades your sample size, which in turn will lead to small confidence limits ofyour density. Every two weeks is sufficient for a good precision of elephant densitiesestimates. If year round monitoring is not possible then monitor from September till April,thus including the core rainy season (September, October) and the major dry season (January,February). This way population sizes can be retrieved during the different seasons and datacan be compared with data from former vears.A.P.219 Kribi Cameroon Tet/Fax+237 34621 3T

IIItIProiect Campo-Ma'an Larqe Mamrrl prooram Patricia BekhuitThe distribution of a population or species gives information on the area occupied by thepopulation or species. In order to obtain information on the distribution of the animals overtime par example through different seasons, data should preferably be gathered year round.For spatial distribution differences, data gathered should cover a large area. In both cases,spatial and time specific, sample size should be high to be significant!Dung counts on strip transectsWhile walking transects all dung seen will be marked and place and time noted. To assuredouble counts can not occur, one can mark dung by attaching a red band on a nearby branchbut one can also decide to simply remove the dung from the strip transect. The red band hasthe advantage of follou'ing the dung pile over longer time, par example when parallelcollecting data on deca1, rate. If follorving dung is not necessary, removing dung is probablythe easiest way. Before removing, distance on the transect (to compare possibly with otherexisting data) and perpendicular distance from the census line to the middle of the dung pileshould be measured for every dung pile (set of boli). The latter to assure one only gather datawithin the beforehand fixed strip width (5 meter both side of census line). Distances are notrecorded of dung piles that are impossible to detect at 2 meûes range in the undergrowth asthey are assumed to have disappeared. Make sure you write your observations clearly downon waterproof paper always mentioning date and place, so other people (or yourself) do notcome across problem deciphering your notes. Write down what you see and not what youthink!! If you see several dung piles, count them and write down how many you saw. It iswrong to count one and leave the other dung piles as you assume they are from the sameelephant (even if you are sure due to footprints)!! Note that a dispersed dung pile is still onedung pile. Write also down if you did not see anything, since no observation is valuableinformation too.Input of fieldwork data in computerThe programme Excel is one of the best to file your data, since easy transfers can be madetowards other computer programs and also some statistical calculation can be completed. Firstput all information in per dung pile thus also including extra data gathered like distance on the8.P.219 Kribi Cameroon Tel/Fax +237 346 21 37

transect (your basis file). Later on you can group dung counts per transect and time; makecolumns of date, place, number of dung (your work base file). During analysis add columnsfor density calculation.Example data input in Excel; work base file15t10t0116t10t0131t10t01Data analysisWhen field data is collected and filed away in data sheets (excel) it is ready to be analysedand interpreted. Thrs part is an imported part since endless pages of raw field data do not helpany manager with decision takings.Analysing the data should start with calculating elephant densities ûom dung counts.Cal cul ating e I ep hant de ns it i e sTo calculate the species density from dung density, information is needed on the productionof dung per animal per day (defecation rate) and the number of days in which most of thedung disappeared (the decay time).A few researchers retrieved a defecation rate for forest elephants (Table Mt). Wing and Buss(1970) discovered defecation rates to be different for seasons in the Uganda forest. During thewet season they calculated a defecation rate of 17 whereas the dry season demonstrated adefecation rate of I I- Powell (199s) did research in the southwestern Cameroon and came upwith a defecation rate of 14,4 by following a collared animal. Tchamba (Igg2) revealed a rateof 19,77 with a standard error of 0,23 while working in the Santchou reserve in Cameroon.For research in the Campo-Ma'an region I selected the rounded defecation rate of 14 (afterPowell) to be trustworthy.8.P.219 Kribi Cameroon Tet/Fax +237 346 21 3Tt0

IilIIIIIITable M1. Defecation rates of the African forest elephantComments19,77(SE0,23) Sanffi17Wet season, Kibale, Uganda11Dry season, Kibale, Uganda14,4South-western CameroonTchamba. 1992Wing & Buss, 1970Wing & Buss, 1970Powell, 1998The decay rate of the elephant dung is more complicated. According to White (1995)differences between decay rates in the distinct seasons surely occur and are enforced byvarious amounts of fruit in dung (fruit availability being season dependant) but he alsoconcluded that most elephant dung piles disappeared after 60-70 days. Evidentlydistinguishing a fixed decay rate is very difficult and one has to except certain variability.However in this research a fixed decay time is taken of 65 days. During the first time atransect is walked, the decay time stands for the maximum dung accumulation time(estimation of 65 da1's). If the same transect is walked for a second time after 14 days, thedung accumulation period is 14 days and not one day rnore or less. Obviously this will giveresults that are more accurate then the frst time using decay estimation. It is for this reasonthat repeated monitoring of the same transects about every two weeks is advised.To estimate elephant density (number ofelephant individuals per km2) from dung counts, thefollowing equation is used:filephant density:fxlxwxdn:1-t-a-d:number of dung encountereddefecation rate (days- I )transect length (/crn)transect width (km)maximum dung accumulation period (days)8.P.219 Kribi Cameroon Tel/Fax +237 346 21 37il

IIIIIIIIICalculating elephant density frstthe formula, a correction needscould have been accumulated.Exam ple density calculation16t10t0131t10t01should be done per field trip per transect because, as seen into be made for the different numbers of days within dungStatisticsWhen all different elephant densities are calculated separately, statistical methods are neededto calculate the mean and the confidence limits of the elephant density and to see if significantdif[erences occtll between dif,Ferent areas or time periods. Comparing or grouping data shouldnot be done at randomly. One can compare all the data from a specific area with all the datafrom another specific area if you would expect a difference (Dipikar island versus Southerncorridor, inside versus outside park) or compare data between different seasons (rainy seasonversus dry season).In our case density data is normally distributed and at interval level.To calculate means per transect or per season you have to sort out your data from your workbase file. For means per transect, data should be grouped per transect, for means per seasonsort out your data on date. V/ithin Excel this can easily be done (!data, !sort out).Simple statistics as calculating confidence limits should be done step by step; beware not touse the standard function in Excel for confidence intervals since it assumes the z value insteadof the t value, making your interval smaller then they should be, handy but not correct. Takethe effort of calculating it yourself following the order given below:8.P.219 Kribi Cameroon Tet/Fax +237 346 21 3Zt2

IproiecttIIIIIIIIIIIIIICampo-Ma,an Larqe Mammal proqram patricia BekhuisExample of calculating confidence limitsDuring the month October 2 transects has been walked 2 times and four densities havebeen calculated: 0,514; 0,176; 0,095; O,1gO.Value - Mean (Value- MeanDensities October ValueValue) 2o,022-0,059Variance = sum of squares / (sample size - 1)Variance of October = 0,0053 / 3 = 0,0018Standard Deviation = square root of varianceStandard Deviation of October = 0,0419Standard Error = Standard Deviation / root of sample sizeStandard Error of October = 0,0421 | 2 = 0,0209Confidence interval = standard error x t (d.f. = 3 fl to,ozs= 3,192)Confidence interval of October = 0,0666Upoer confidence limit = mean + confidence intervalUpper confidence limit of October = 0,2205Lower confidence limit = mean - confidence intervalLower confidence limit of October = 0,0872When confidence limits are calculated for densities in two different areas or for two differenttime periods, significant differences with a probability of 95o can be identified.8.P.219 Kribi Cameroon Tel/Fax +237 346 21 37l3

NIIItIIITo accept a significant difference with a probability of 95Yo, the mean density of group Ishould be outside the confidence limits of group 2 and the mean density of group 2 should beoutside the confidence limits of group 1 (two-sided test, t-value 0,02s).Example of two mean densities (A & B) with their confidence limits when significant differenceand when no difference can be demonstratedABSignificant differentNo significant differenceAside from calculating confidence limits step by step there are statistical tests to demonstrateif significant di{lerences occur. Statistical tests can be used if distribution is not normal or ifone wants to demonstrate differences between more then two groups or if one simply wants toknow the chance that there is no difference (p-value).Which statistical method to use for verifiing if differences do or do not occur, depends on theamount of data gathered and how many groups you want to compare (if data is normallydistributed). To choose the right statistical test for differences comparison a key-table fromWhite and Edwards (2000) is included below:p8.P.219 Kribi Cameroon Tel/Fax +237 346 21 37t4

lIproi""t Campo_Ma'"n L"ro" M"**"1 prooram patricia BekhuisIIIIIIIDo you have more than two samples between u'hich youwish to analyse differences ?execule an ANOVA testDo you have more than about 30pairs ofmeasurements in yourtwosamples ?Are your measures atinterval level ?ilExecute the Marur-Whitney U testIf two areas are compared, the measures are not matched pairs. Matched pairs exist if onemeasure from one of the samples can be paired with one (and orùy one) measurement fromthe other sample.If differences are found, note at which confidence level, i.e. 95yo, 99o or 99,9yo thisdifference is found. Less than 95Yo is considered not significant.If no difference can be confirmed with at least 5% significance level using the right statisticaltest, the data should be grouped and a mean density calculated. While reporting a meanelephant dersity, present the standard deviation or interval levels as well.B.P. 219 Kribi Cameroon Tel/Fax +237 346 21 37l5

IIIIItIIilllProiect Camoo-Ma'an Larqe Mammal Proqram Patricia BekhuisReporting the resultsInterpretation of resultsWhen data is analysed, the outcome are numbers (confidence limits, probabilities) which nowhas to be interpreted by the researcher. What do these [umbers mean, how can they be used inreal life.The information like; for the mean density of April and October the outcome of the T-test wasp

IItIIIIProiect Campo-Ma'an Larqe Mammal Prooram Patricia BekhuisMake sure you stand behind the report you have r.r,ritten. Your name is on it and yourreputation depends on it. If you are not confident about your work, ask colleagues or others togive comments on your draft version before finalizing it.Good luck8.P.219 Kribi Cameroon Tel/Fax +237 346 21 37t7

I),-TIIIIIIIIIProiect CamPo-Ma'anProtocolStep IPlace transects if not yet done-> See fieldwork page IStep 2Walk transects and make observations every two weeks-> See dung counts on strip transects page l0Step 3Enter all data in the cot'ttputer-> See input of fieldu,ork data in con'puler page 10Step 4:IlIAnalyse the data-> See data analysis page I IStep 5Interpret and format the results-> See reporting the results page l7IIIIA.P.219 Kribi Cameroon Tel/Fax+237 34621 37l8

IIIttttttIIIllIIIIIIReferences cited in manualBekhuis P.D.B.M. (2002) Elephant density and Distribution in the southern part of NationalP ar k C a mp o - Ma' an. Internal repo rt Campo -Ma' an P roj ect Camero o n'Blom A., A Almasi, I.M.A. Heitkonig, J-B Kpanou, H.H.T. Prins (2002) A survey of the apesin the Dzanga-Ndoki National Park, Central African Republic' Journal of AfricanEcology, in Press)powell J (199S) Biologt offorest elephants in southwestern Cameroon.unpublished Ph. D. Thesis, university of cambridge, cambridge, uKTchamba M.N (1992) Defecation by the Africanforest elephont Loxodonta africana cyclotisin the Santchou Reserve, Cameroon. Mammalia 56: 155-158Van der Hoeven c.A. (2002) Assessment of mammal densities in the campo-Ma'an rainForest by way of pooled expert Opinion.Internal report Campo-Ma'an Project Cameroon'white L.J.T. (1995) Factors affecting the duration of elephant dung piles in rain forest in theLopé Reserve, Gabon African Journal of Ecology 33: 142-150\\4rite L., Edr.vards A eds.. (2000) Consen,ation reseat'ch in tlrc A"û'icatt raittforeslsI atecltnical handbook. Wildlife Conservation Society, New York.wing, L.D. &.I.O. Buss (1970) Elephants andforest. wildlife Monogr. l9:l-92B.P. 219 Kribi Cameroon Tel/Fax +237 346 21 37l9I