The presentchapter is structured in two parts. The first one explains different methodsused for collecting data and the second one is related at the data analysis,statistical methods and software utilized for data treatment. 3.1 Data collectingThey were relative to ethnobotanic and economicsurveys as well as entomologic data collecting. Before all this, literaturesurvey were done in different documentation centers of FLASH, ABEE, BIDOC, LEA,IITA, Faculties of Forestry and of Biology I at the University of Freiburg.Furthermore deeply understand of the topic were worked out via Internetwebsites research.
Climatic and demographic data were collected at ASECNA andINSAE respectively. The section will be presented according to objectives. Theresearch questions were investigated within the mixed method framework thatused different tools and approaches for collecting quantitative and qualitativedata from field respondents by medicinal plant users living near to forests.
Inour study data were collected through structured survey questionnaires, focus group discussions key informantinterviews and, personal fieldwork observations (entomofauna collecting) . 3.1.1 Ethnobotanic surveyAccording theirproximity to different forests, five (05) and four (04) surrounded villageswere selected near Lama Protected Forest (Lama Forest) and Lokoli Swampy Forest(Lokoli Forest) respectively for ethnobotanical study. A total of nine villages surrounding LamaForest (Akpè, Koto, Massi, Agadjaligbo, Zalimè) and Lokoli Forest (Koussoukpa ,Lokoli,Dèmè, Samionta) were covered for the survey.
Structured and semi-structuredinterviews were carried out from September 2012 to February 2013 to identifythe main tree species used to harvest tree barks for medicinal purposes. Atleast 30 informants per village have been interviewed. They included differentsocio-professional groups (Table1), such as collectors (harvesting bark to sellit), Traditional healers (harvesting bark for medicinal purposes), farmers andothers (craftsmen, housewives etc) from different socio-cultural groups (Fon,Holli and Aïzo). As numbers of different target groups were unknown, we chose30 informants per village in order to cover all groups since none statisticalanalysis was competed at village scale.
To recruit Traditional healers andcollectors we relied on a snowball technique, to know where recruits areparticipated and answered the getting names of them. Farmers and others wererandomly questioned. The first part of our questionnaire asked for socio-demographicdetails of the respondent. The second part addressed the purpose, harvestingmethod, harvesting frequency (daily, weekly, monthly etc.) and identity oftrees bark-harvested, different diseases healed, the processing method(decoction, maceration etc.) as well as ‘awareness’ questions (e.g.do you remarksome debarking impacts on the tree after harvesting?).
Variables such as: Sap Loss, Crown Decimation, FlowersAbortion, Leaves Loss, Insect Infection and Change Shapewere asked (for details see Annex 1). A total of 261 informants wereinterviewed during this survey, 104 from the Lama Forest and 157 from the LokoliForest. It is worthwhile mentioning that in the Lama Forest people were lesswilling to participate, therefore we could not get higher sample sizes. Usingdata from the ‘Institut National de laStatistique et de l’Analyse Economique’ (INSAE, 2004), thepopulation sizes in the villages surrounding Lokoli Forest and Lama Forest wereestimated to 1452 and 1919 inhabitants,respectively.
Hence, the sample sizes correspond to 10.8 % of Lama Forest’s and5.4 % of Lokoli Forest’s inhabitants. 3.1.
2 Diversity of Coleoptera xylophagous of most important medicinaltrees debarked126.96.36.199 Sampling method As explained and described above, both Lama and Lokoli Forests aredifferently managed. While the first one is protected ecosystem with forbiddenaccess, the second one suffers from human pressure via the use of forestresources such as tree barks for medicinal purposes because of its free access.Then, studying the effect of forest disturbance such as debarking on insectdiversity especially the xylophagous and their natural enemies is very importantin order to evaluate the impact of local population on forest resources. Thiscould be made by several methods and one of the most important is beingassociated with better design of field experiments: BACI analyses which meanBefore-After/Control-Impact methods (Stewart-Oaten et al.
, 1986). In the present study,we adopted the Control-Impact method by setting up insect traps on debarkedtrees and non-debarked tree which served as control. For that, we selected treemostly cited in the previous study in inside both forests and added tree harvestedin the field which is more open habitat, since farmers spared some species intheir fields. It is worthwhile to precise that, due to agriculture system, wecould not consider data from open habitat in Lama Forest since sites selected wereburned by farmers. Thus in Lama Forest we considered only trees selected inforest such as Anogeissus leiocarpa,Dialium guineense and Khayasenegalensis while in Lokoli Forest species such as Naucleadiderrichii, Syzygium owariense andFicus trichopoda were selected in the forest and in the field we worked on Parkia biglobosa, Pterocarpus santalinoides and Bridelia ferruginea.
Then closedhabitat here concerned traps established on tree species debarked or not insideforests and open habitat concerned those set up on tree species in fields. Toallow comparisons, a standardized collecting method was used. It consisted ofestablishing interception traps made with funnels surmounted of plastic with anarea of 10cm x 15cm (Picture 3).
This funnel leads to a recuperation tube whichcontained formol (0.5 %). Per tree species, two treatments with six replicateseach were installed. The first treatment consisted of a debarked tree where anarea of 10 cm x 20 cm was experimentally debarked at breast height.
The secondwas a control tree where no bark was removed. A total of 108 interception trapswere installed. Traps were sampled every two weeks for a period of four monthsduring the dry season. Data collected were: insects caught by trap and presence/absenceof insect holes and number of insect holes3.1.
2.2 Sorting scope and identificationWe focused on xylophagousColeoptera because they were most abundant and are most relevant for treehealth when it concerned bark harvesting. All taxa with a potential relevancefor wood were considered, such as bark beetles (Scolytidae), ambrosia beetles(Platypodidae), longhicorn (Cerambycidae), jewel beetles (Buprestidae) andtheir natural enemies (Cleridae, Staphylinidae, Histeridae etc.). All Coleoptera were first sorted to ‘morphospecies'(sensu New 1998) and then taxonomically identified at the InternationalInstitute of Tropical Agriculture (IITA) in Benin. 3.1.3 Economic surveyAfterstudying the effect of debarking on tree species, we tried to understand thereason leading local people to continue harvesting bark on medicinal treespecies in Lokoli Forest.
Is the activity of selling bark profitable to collectorsor not. For doing that, a survey was carriedout from January to March 2013 in the villages next to Lokoli Forest: Lokoli,Dèmè, Samionta and Koussoukpa. It was based on interviews using questionnaires(Annex 2 Table 1) with 30 collectors asrespondents who harvested bark directly in the forest and sold it at differentmarkets. In a pre-survey, three groups of collectors were identified: Group 1(G1) collectors sell their bark locally, in the villages next to the forest;group 2 (G2) and group 3 (G3) collectors sell it at medium- (< 50 km) anddistant markets (> 50 km), respectively. We collectedfollowing data: (i) Bark species and quantities as well as their prices(selling price) (ii) All costs induced by the activity, such as purchase, travellingcosts (round trips for bark harvesting and travelling to the market), costs ofmaterial used and their depreciation, number of persons involved in theactivity as well as the number of working days before going to the market, (iii)All taxes paid by the collectors and the number of times they went to themarket per month (Annex2, Table 2).
Annual economic parameters were obtained byextrapolating monthly values. Since collectors were active only during the dryseason, we considered a period of six months for estimating annual values. 3.1.
4 Factors influencing medicinal tree attacks by xylophagous Determiningmain factors which influence medicinal tree attacks by xylophagous is an interestingbut in the same time very astringent because of the cryptic life mode ofxylophagous. Biotic (forest compostion, naturel enemies, and abiotic factors are capable to influenceforest insects. As we could not consider all factors, we emphasized on some whichwere relatively easy to follow such as tree species, habitat, season andharvesting rate.
Data were collected in one year allowing following allseasons. Interceptiontraps were set up on tree species for four months and then substituted byemergence traps for four months during both dry and rainy seasons. Data werecollected from October 2014 to November 2015 in both degraded and non-degradedforests and in the fields. Here, we followed the description and classificationof Lachat et al, 2006 in selecting of different habitats in Lama Forest.
Forbetter comparing results between habitat types, as both forests areecologically different, we selected some common debarked species to bothforests and those belonged to the most tree bark sale by collectors (see table2). Three different habitats were selected; non-degraded forest, degradedforest and field. In Lama Forest, species of non degraded forests wereestablished in Humid dense forest of Synometra megallophylla and species ofdegraded forests were selected in typical dense forest. In Lokoli Forest,species of degraded forest were established both within mash forest and at theedge of forest while trap of non-degraded forest’ species were settled withinswampy forest were very few human impacts were noted.Thefollowing table 2 shows all selected species debarked or not in differentforests. As tree species selected were more important than in previous study,we limited the replication at three in order to be able to collect entomofaunaand to make sorting of different xylophagous and their potential naturalenemies which have been captured. Per season, 72 and 90 traps were establishedin Lama and Lokoli respectively.
Data collected were: individual number ofinsects caught by the traps, presence or absence of insect holes, number ofinsect hole recorded on the tree.3.2 Data analysis 3.2.1Ethnobotanical data analysisAll statistical analyses wereperformed in R (R Core Team, 2014) and the significance level was set to ? = 5%.
Barkuses and tree species: The different diseases cited by respondents wereordered for each forest using the Fidelity Level (FL) of Friedman et al.(1986): Where nj isthe number of respondents who reported the cure of a given disease using thebark of a given tree species j and Nf is the totalnumber of respondents for each forest.This was made in order toselect tree barks which were cited for diseases with FL ?5 % and presented withprocessing method in healing disease. Fisher’s exact test was used to testsignificant differences in bark uses in term of disease healed between Lama Forestand Lokoli Forest. A pattern in bark uses (times a tree was mentioned) in thetwo considered forest types were assessed with Correspondence Factor Analysis (CFA) performed onprofessional groups and tree species involved in the cure of diseases.Identificationof most debarked tree species : We proceeded by weighting the response (as use scoreR) given by respondents as following: all plants harvested daily or weekly arescored as 5; plants used monthly or quarterly are weighted as 3 and thoseharvested biannual or annual are scored as 1. The use intensity (UI) for eachtree species was determined as the average use score (R): Where Rij isthe use score of the tree species j by the respondent i from agiven group of sample size N.
Ineach forest, using Wilcoxon rank sum test (UIj?0),we tested in which forest tree species were more debarked. A cluster analysiswas then performed to group the most important debarked species according totheir UIj values computed per professional group for eachforest. Principal components analyses (PCA) was furthermore applied to the UIjdata to characterize relevant clusters in terms of professional groupand forest.Multiple comparison testsunder Kruskal-Wallis rank sum test were then performed in the R library agricolae(de Mendiburu, 2014) to assess variations in harvesting score with respect to forest, ethnic group, age, sex andprofessional group.Assessmentof debarking method: A Correspondence FactorAnalysis (CFA) was performed to highlight the use of bark according to professionalgroup and harvesting method debarking method recorded during the survey. Assessmentof users’ relative impact on debarked tree species and sustainable debarkingmethods used: Theperceptions of plant bark users on the impact of debarking on trees wereassessed using Impact01 (yes/no) and six impact type variables namely SapLoss, Crown Decimation, Flowers Abortion, Leaves Loss, Insect Infection andChange Shape. Generalized linear models (probit) were fitted to assessthe variability of these seven binary perception variables against forest type,ethny, age, sex and profession. The R function step was used to reducemodel through backward selection of model terms.
Furthermore, a CorrespondenceFactor Analysis (CFA) was perfomed to relate the use of sustainable debarkingmethod according to professional group. 3.2.2 Study of diversity of Coleoptera xylophagous and their naturalenemies The species diversity of insects was evaluated using three common?-diversity indices, the cumulative number (S) of species recorded on a tree,the Shannon-Weaver (1949) diversity index (H’) and the Evenness (J) of Pielou(1966).
The index of Bray-Curtis was computed to assess dissimilarity betweenhabitat types (Lama-forest,Lokoli-swamps and Lokoli-crops).To assessthe importance of each group for each habitat type, the percentages of allcollected insects (both abundance and species richness) by ecological group(Predator and xylophagous) and in different insect families were used. Fisher’sexact tests were performed to test for differences in the frequency(considering both abundance and species richness) of ecological groups betweenhabitat types.Theindicator value index (IVI) of Dufrêne and Legendre (1997) was used todetermine indicator species for habitats types, tree species and treatments(debarking). For a given species iin a group (defined by habitats types, tree species or treatments) indexed k, IVI is given by: whereAki and Bki arerespectively the specificity and the fidelity (define!) of a species? with ?ki the mean number of insects of the species i fromall sampled trees in the group k and p the total number ofgroups; with?kithe number of trees onwhich the species i has been collected in the group k and ;?k total number of treessampled in the group k.To assessthe significance of computed IVI values, randomization tests each based on10000 permutations were performed using the R library indicspecies (DeCaceres and Legendre, 2009). Species with IVI ? 0.
25 and significant at 5 % level were selected as indicator species(Lachat et al., 2006). The density of insects (N,insect abundance) and the diversity indices S andH’ were used to assess the effect of debarking on the attraction ofbeetles and the variability of this effect across habitat types and treespecies. For each of the three dependent variables (N, S and H’), a repeatedmeasure ANOVA was considered with habitat type, tree species (nested to habitattype) and debarking as fixed factors and sampling time as random factor. ANOVAswere performed using the R library nlme (Bates et al., 2014) withPoisson error distributions for N and S; and a Gaussian error distribution forH’ after Box and Cox power transformation (Box and Cox, 1964). The mean andcoefficient of variation (cv) of the three dependent variables were calculatedfor each habitat type per treespecies considering debarked and non-debarked trees separately.
The abondanceof xylophagous beetles (Nx), the number of predators (Np), thenumber of predators per xylophagous (Npx) and the number of holes (NH)per tree were used to assess the impact of insects on trees, according to treespecies (per habitat type) and debarking. A negative binomial error model wasfitted to each of these variables (Nx, Np, Npx and NH) using the R library MASS (Venablesand Ripley, 2002). Goodness of fit was assessed using R2 and ?2test on residual deviance from each model. Means of Nx, Np, Npx and NHwere computed per tree species for debarked and controls trees separately andcompared using negative binomial error models. 3.2.
3 Economical data analysis188.8.131.52 Profit and Loss accounts AnalysisThe profitand loss accounts analysis estimates the Net Income (NI) over one year. It isthe Operation Income (OI; OI = Qt*SP; Qt: bark quantity, SP: selling price) aftersubtracting different costs.All costs(total costs: Ttcost) faced by economic activities can be broken into two maincategories: fixed costs and variable costs.
Fixed costs (FC) are those that donot change, while variable costs (VC) change depending on the company or aprivate individual’s activity. Fixed costs are linked to staff expenses, totaxes and to equipment while variables costs represent costs linked totransportation, packaging and the purchasing of bark. Net Income(NI) is the difference between OI and FC+VC= TtCost. TheEconomic Profitability (EP) was then calculated as the ratio between NI and thesum of Ttcost induced by the economic activity: 184.108.40.206 Sensitivity AnalysisTheeconomic analysis was based on computation of the Break-even Point (BP) and thesafety margin (SM).
BP is the ratio from the product of OI and FCdivided by OI-VC. SM is OI minus BP.BP= ; SM= OI- BP 220.127.116.11 Analyses of the profitability of bark saleStatistical analyses were performed in R (R CoreTeam, 2014). The significance level was set to ?= 0.05.
A mixed model ANOVA (random intercept linear mixed model) wasperformed, independently for each variable of the Profit and Loss AccountAnalysis and the Sensitivity Analysis using the R library lme4 (Bates etal. 2014). Each time ‘collector group’ and ‘tree species’ were fixedfactors and ‘respondent’ a random factor.
Mean values and standard errors werecomputed per collector group and per tree species to highlight significantdifferences in returns from bark sale.3.2.
4 Factor influencing medicinal plant attacks byxylophagous insectsTo assessthe distribution and the diversity of insects, abundance (N) and diversityindices (species richness (S), Shannon-Weaver (1949) diversity index (H), andequi-repartition index (E) of Pielou (1966)) were computed per habitat withinstudied site and per debarking level and insects ecology. Bray-Curtisand Jaccard similarity indices were also computed between habitat types withinstudies sites to assess the resemblance of these habitat types. The proportionsof entomofauna from each recorded family and order were computed to identifythe most prominent insect families and order in the studied habitats. The indicator value index (IVI) of Dufrêne andLegendre (1997) was used to determine indicator species in different studiedsites and across debarking levels. This analysis was performed with the Rpackage ‘indicspecies’ (Caceres and Legendre, 2009).AGeneralized Linear Mixed Model (GLMM) with binomial error distribution wasadjusted to insect attack event with studied site, presence of water, habitat,tree species, debarking level and season as fixed factors and collection timeas random factor.
Two GLMMs with Poisson error distribution were also adjustedto the number of beetle individuals per beetle species (offset) and to thenumber of beetle individuals on attacked trees using the same set of factors. GLMMswere fitted using the R package ‘glmmML’ (Broström, 2013) and thepseudo-R² (coefficient of determination) of Nagelkerke N (1991) was computed toassess their explanative power. The significance an estimated coefficient wasinterpreted as the significance of the effect the corresponding factor on theresponse. Mean and standard error of the responses were computed to illustratethe results of the GLMMs.