Planning from scratch: A new modelling approach for designing protected areas in remote, data‐poor regions

Protected areas are vital tools to safeguard the remaining strongholds of biodiversity, championed by High Ambition Coalition for Nature and People Post-2020 Global Biodiversity Framework as one main instruments protect nature help save species from extinction (Jetz et al., 2007) (HAC: www.hacfornatureandpeople.org, global biodiversity framework: https://www.cbd.int/conferences/post2020). However, gaps in quality data on distribution remain a major challenge planning protected areas, especially remote regions that often rich but poor (Amano 2016). Factors such political/social insecurity, inaccessibility, language barriers or lack local expertise prevent collection many around globe & Sutherland, 2013). Together with dynamic species, these spatial temporal ecological may jeopardize analyses required plan new their zoning, is delineation land units associated specific purposes (Ladle Hortal, While birds best documented taxon databases, only 0.5% deficient (IUCN, 2021), even they can be very unevenly sampled (Christie 2021). Vast central Asia Sahel Africa have enormous importance, yet among world's most terms birds. The makes them less accessible conduct extensive standardized surveys 2016; Amano 2013) use traditional bird survey methods. effective methods estimate abundance large (Collen 2008)—such tropics (Robinson 2018)—can lead possible bias inferred population parameters (Buckland 2008), hampering efforts identify conservation interest (Girardello 2018). As poorly places threatened changes 2021) larger ones play significant role key ecosystem services overall maintenance (Watson 2016), need target halt loss worldwide (Di Marco 2019). Scientific evidence has shown creation and/or reinforcement particular priority 2007), where human footprint rapidly increasing (Anderson Mammides, 2020). In this context, response request managers southern Chad, data-poor Sahelian country, we developed relatively rapid approach design area zoning targeting mammals, given an almost total quantitative data. combines distance sampling (DS), density surface modelling (DSM) systematic software Marxan: combination our knowledge not previously been used peer-reviewed study delineate (however, see combining DS-DSM-Zonation software: Winiarski 2014). approaches require lot (Ball 2009), challenged (Kukkonen Tammi, One way generate homogeneous observations ground difficult obtain aerial DS. then integrated into Marxan through DSMs. This was carried out framework African-Eurasian Migratory Waterbird Agreement (AEWA) Ramsar Convention RESSOURCE project (‘Strengthening Expertise South Sahara Birds Rational Use Communities Environment’: https://www.swm-programme.info/ressource-project). Bahrs Aouk Salamat floodplains Chad remarkable vertebrate under anthropogenic pressure (Brugière Scholte, 2013; Zwarts so partnered organization African Parks develop preliminary region, aim centre proposed vast pristine wetland (Lake Iro its tributary floodplain) sustaining high concentration birds, possibly mammals. To address geolocated data, conducted surveys, recording both mammals analysed collected using unique suite potential primarily Our produced alternative scenarios were discussed view implementing measures biodiversity. cover 50,000 km2 (Figure 1) characterized tropical dry climate. landscape region dominated dense shrubby forests (Raimond 2019) typical floodplain vegetation Andropogon sp., Hyparrhenia Cymbopogon Echinochloa Oryza sp. Acacia Lake (145 km2) supplied Bahr (river) hosts resident migratory great importance wildlife, including elephants Loxodonta africana, Kordofan giraffes Giraffa camelopardalis hippos Hippopotamus amphibius, two areas: Zakouma National Park Faunal Reserve. latter buffer zone park activities authorized, except hunting construction settlements, which theoretically prohibited. largest sites Africa, six vulture (all either Critically Endangered IUCN Red List), black-crowned crane Balearica pavonina northern hornbill Bucorvus abyssinicus, listed Vulnerable List particularly wetlands caused (Zelelew 2020) classified national heritage Chad. Human low, towns (sub-prefectures) small villages hamlets found across area, well nomadic camps during season. addition sedentary communities, stakeholders include authorities (e.g. territorial village chiefs, nomad sultans), whose influence resource tenure management conflict resolution still significant, administrative heads districts sub-districts, sectoral chiefs inspectors fisheries, livestock, wildlife), who responsible laws provincial regulations, natural management. Administrative sufficient means fulfil missions, example, regulate manage agriculture, ensure sustainable legal fisheries act against illegal activities. Thus, involving communities representatives participatory processes managing resources. Modelling essential asset start build concertation. Other involved biologists experts participating wildlife censuses, Chad's Ministry Environment external non-governmental organizations, reserve site located. case Zakouma's Greater Ecosystem (GZE), located, Parks, (NGO) charge senior staff Environment, NGO. Together, form Management Unit. An unsuccessful attempt field 2020 confirmed extremely limited track network, making barely foot only. economy based mainly livestock rearing, agriculture (including single-crop farming Sorghum durra), all constitute direct indirect threats mammal, fish due habitat harvesting recent decades, droughts conflicts displacements within become increasingly attractive migrants 2019), putting additional ecosystems further protection. poor, over entire thereafter named ‘survey area’ 1a). These yearly 2017 four strata (i.e. sub-regions depending wetlands) along transects following adjusted-angle zigzag (see Figure 1c transect locations). guarantees equal coverage probability complex while being cost-effective compared other more conventional designs 2004). A 192 (mean length = 44.70 km) automatically generated ‘Distance’ v.7.0 (Thomas 2010) surveyed 4 years study. licensed Accord de partenariat entre le Gouvernement la République du Tchad Networks (APN) pour gestion financement Parc son Grand Ecosystème Fonctionnel—2017–2027. aims contribute International Census (IWC) simultaneously wetlands, mid-January early February, when visible. During flown at 90 m above level low speed (~140 km/h) observer each side aircraft. Blind 25 immediately below aircraft right-truncated represent minimum observation 0). divided bands (0–80, 80–180, 180–380 380–780 m, respectively), materialized marks windows aligned streamers fixed wing struts Using portable recorders, observers placed recorded band (one four), mammal size observed cluster, name position aircraft, number, time day date. Additionally, opportunistic any transect) same January rarer list counted Appendix S1). records too scarce basic species-specific models clustered single patch (hippos Iro), process, provided helped assess efficacy (PADs). We aimed collect analysis much information maximize comprehensiveness scientific future area. total, 71 (61 43 waterbirds, 10 mammals) detected All analyses, met threshold. DS performed least 40 2001). Because similar-looking share detection function, pooled waterbirds (<40 records) common waterbird 2008). dataset check similarity waterbird. all, 24 taxa did meet sample threshold modelled individually included layers put extra emphasis analysis: richness below). addition, also biological entities: (1) some genera unidentified few (genera Ardea, Anas Vanellus), (2) (3) richness. categories calculated point. covariates generally known affect detectability (Table S2). Analyses R (version 1.1.383, Core Team, ‘ds’ function package, version 1.0.1.9005 (Miller For entity, first selected functions (from half-normal, hazard-rate uniform functions) adjustment cosine, hermite polynomial simple polynomial) AIC (or AICc <80 observations), providing χ2 goodness-of-fit test satisfactory (Burnham Anderson, 2002; Miller 2013, Moreover, shape checked step model selection procedure verify assumptions underlying added covariate most-supported null non-colinear stepwise manner retained until no improvement could obtained. Density associate count obtained (segments) dividing identified, managers, reduced referred ‘planning 1c), DSMs extrapolated distributions. Raw indicated concentrate southwest Iro, increased pressures occur. part already fishing restrictions interested identifying would benefit general, focused surrounding side. QGIS 3.16—QGIS Development thus restricted 50 km northeast 75 1c). 3409 segments 1200 × each. Species assigned segment geographical location. ‘dsm’ package detailed Table S3, potentially (Brito 2014, 2009). prior analysis. fit various distributions (quasi-Poisson, Tweedie negative binomial) 2002). Fit qqplots residual plots. Covariates AICc. Retained autocorrelation visual inspection residuals, assessed explained deviance coefficient variation (CV). discarded CV > 50%. grid squares (such unit approximately twice truncation i.e. 580 m), DSM corresponding limit extrapolation errors (Fifield 2017), removed cells predicted values outside range locations. Although false absences, preferred option falsely estimating directly v.2.43 2009) visualization tool Zonae Cogito v.1.74 (Segan 2011) set targets lowest socio-economic cost (Watts 2017). uses simulated annealing algorithm allows flexible time-efficient computing protect, units’ (Game Grantham, Details configuration S4a. features entities optimize of) resulted 20 2 (warthog baboon) perform S5 species). numeric objectives reach entity) consultation Section 2.3.3; S5). primary focus work below) three (total vulnerable crane, needing urgent protection) suitability constructing comprehensive, sound, science-based currently includes restriction Local suggested systematically solutions Marxan, functionality it generates (‘fixing’ in). On account up-to-date validated layer considered. Opportunistic although output maps created inclusion PADs, implementation PAD. minimize area's overlap villages, activity (sorghum cultivation), equally cell proportional portion containing dependent type (either core buffer). location managers. mobility road access crops located walking 3 (Akponikpè 2011). belonging 3-km radius contained radius, 1 if entirely inside 0.01 outside. cells, half value. possibility excluding generates, set, demand exclude proximity Boum Kebir Alako. designed (with moderate targets), levels protection entities, amounts individuals options. Targets criteria five (Ramsar Wetlands, arbitrarily gradient (10%, 25% 50% entity level, respectively). aforementioned crane), penalty factor (SPF) better doubled (compared value should regard criterion six). presents, different feature, rationale behind method them. At had composed restrictive implemented, consist unfragmented population. core, reinforce without requiring strong measures, allowing Note here enlarged area). Hence, simply because larger, higher than alone. tend bigger capture areas. Potential PADs steps: first, (core) scenario, built respectively. Each preceded calibration Penalty Factor (SPF), weight improving Marxan's performance meeting target, boundary modifier (BLM), scaling compactness Ball asked select those solutions, number offers in), order contain medium fitting encompassing targets, Concertation between scientists led final solutions. sensitivity Ardron al. (2010) input removing reliable (species richness, hadada ibis, purple heron) running again evaluate whether lower might detrimentally affected simulation outputs details results S6). requested working nearby development comprehensive regional network general elaborate implement Land-Use Plan GZE, covers Reserve started lacking suitable GZE region. previous years, received showed concentrations diversity avifauna floodplains, present support strategy (step 1, 2). involvement phase, regular discussions team recommended Vogler (2017). From initial conceiving interpreting results, partners interacted frequently (steps 8). context team, turn, made proposals about process 2) 4) refined feedback 7 regularly informed broader Land Plan, sessions Governance Meeting (February governance bodies 5). concertation teams, prioritized views 9). 5506 635 443 recorded, 140 303 derived estimates 28 22 example list). Most coordinates, presence/absence water altitude covariates. estimated 454,404 (Confidence Intervals: min 343,984, max 600,270) robust (Sylvicapra grimmia, Redunca redunca, European ducks, Spatula querquedula, Ephippiorhynchus senegalensis, lapwings) insufficient size. reasonable 30.5%, Concurvity usually (mostly <0.5); 0.7, concurvity. Mean 39.6%. residuals several DSMs, tried alleviate adding autoregressive (Dormann 2007; reduce autocorrelation, remained superior 0.1 ibis Overall, accounted S6a). expected observations, lake. Both predominantly distributed baboons). Park. S3c. (both targeted interest. covered highlighting 4). mean 925 4a). When 1535 1488 4b). averaged 2286 avoid