The Future of Walnut–Fruit Forests in Kyrgyzstan and the Status of the Iconic Endangered Apple Malus Niedzwetzkyana

The future of walnut–fruit forests in Kyrgyzstan and the status of the iconic Endangered apple Malus niedzwetzkyana

B RETT W ILSON,MORENA M ILLS,MAKSIM K ULIKOV and C OLIN C LUBBE

Abstract Forest ecosystems are rich in biodiversity and pro- Keywords Apple, firewood collection, Kyrgyzstan, live- vide valuable ecosystem services, but are declining world- stock grazing, Malus niedzwetzkyana, species distribution wide. Malus niedzwetzkyana, an Endangered wild relative modelling, walnut–fruit forest, wild relative of domesticated apples, is an important component of the Supplementary material for this article is available at walnut–fruit forests of Central Asia. Its iconic pink blossom https://doi.org/./S and genetic properties give it special cultural and scientific significance, but livestock grazing and firewood collection threaten its survival. The conservation of the species and its native forest ecosystem is critical and urgent. This Introduction study provides information on the ecology and population orest ecosystems support biodiversity and provide of M. niedzwetzkyana and the threats affecting its habitat, essential services to people (FAO, ). Estimates improving our understanding of its distribution and pro- F suggest that forests contain . % of all terrestrial plant posing measures to reduce threats. We collected ecological and animal species (Secretariat of the Convention on data and assessed population structure and threats at four Biological Diversity, ), but many forest ecosystems are forest sites in southern Kyrgyzstan. We mapped  indivi- declining and becoming more degraded as a result of duals, creating the largest known dataset for this species. We increasing human disturbance (Hosonuma et al., ). developed species distribution models for M. niedzwetzkya- These threats continue to intensify as the human population na to identify climatically suitable regions and potential grows and more people inhabit biodiverse areas (Fisher & areas for restoration. Sary-Chelek Biosphere Reserve con- Christopher, ). tained the largest expanse of pristine forest and the most The walnut–fruit forests of Central Asia have declined stable M. niedzwetzkyana population, followed by Kara- dramatically in the last  years; only –% of their original Alma Forestry Unit. Forests in the Gava Forestry Unit and area remains (Djanibekov et al., ). They occur at alti- Dashman Reserve were most extensively damaged by tudes of –, m (Hemery & Popov, ) in the foot- humans and livestock. The wild apple’s favoured habitat hills of the Tian-Shan mountain range, within the Central was south-west facing slopes with a gradient , ° and a Asian biodiversity hotspot (Critical Ecosystem Partnership relatively open canopy. Overall, the study population was Fund, ). These forests play an important environmental vulnerable to extinction with limited regeneration poten- and cultural role in this region, providing ecosystem ser- tial. We recommend short-term population enhancement vices, including soil protection and water regulation, and re- through planting projects and increased protection of indi- sources for wildlife and local communities (Cantarello et al., viduals in pristine sites. Community-based conservation in- ). The largest remaining area of walnut–fruit forest is in itiatives should be prioritized in extensively damaged sites, Kyrgyzstan (Epple, ), the most forested country in and larger-scale reforestation of these forests needs to be Central Asia (Djanibekov et al., ). However, over the considered. past  years the forest area has been greatly reduced, driven by the economic recession following Kyrgyzstan’s inde- pendence. Rural communities rely increasingly on natural

BRETT WILSON* (Corresponding author) Imperial College London, London, UK resources and agricultural practices, leading to expansion E-mail [email protected] of agricultural areas, increased grazing pressure, and unsus- MORENA MILLS Department of Life Sciences, Imperial College London, Silwood tainable exploitation of forest products (Borchardt & Dorre, Park, UK ). MAKSIM KULIKOV† CEN Center for Earth System Research and Sustainability, Today % of the country’s population live in rural areas Institute of Geography, University of Hamburg, Hamburg, Germany (Djanibekov et al., ) and % live below the poverty line COLIN CLUBBE Royal Botanic Gardens, Kew, Richmond, UK (Fisher & Christopher, ). Firewood collection and live- *Currently at: Department of Plant Sciences, University of Cambridge, stock grazing are the greatest immediate threats to the forest Cambridge, UK. E-mail [email protected] –  †Formerly at: Fauna & Flora International, Bishkek, Republic of Kyrgyzstan ecosystem, with Kyrgyz households collecting m of Received  March . Revision requested  May . firewood per year and livestock generally allowed to graze Accepted  September . First published online  March . without restrictions in the forest (Djanibekov et al., ).

Oryx, 2019, 53(3), 415–423 © 2019 Fauna & Flora International doi:10.1017/S0030605318001230 Downloaded from https://www.cambridge.org/core. IP address: 170.106.35.76, on 27 Sep 2021 at 00:32:04, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0030605318001230 416 B. Wilson et al.

Many regions have implemented management strategies Study area to achieve sustainable use of forest resources (Kyrgyz  –  Government et al., ; Fisher et al., ). However, in We conducted fieldwork during May June in many forest tracts these management and protection rules four sites within the Jalal-Abad region: Sary-Chelek Bio- are not adhered to (Orozumbekov et al., ), leading to sphere Reserve (Sary-Chelek), Dashman Reserve (Dashman), extensive environmental damage (Hemery & Popov, ; Kara-Alma Forestry Unit (Kara-Alma), and Gava Forestry Herold, ). Unit (Gava). These sites are also the focus of Fauna & ’ Our study focuses on the Endangered Malus niedzwetz- Flora International s (FFI) regional conservation work for kyana, a close, but rarer, relative of Malus sieversii, which is the Global Trees Campaign, and FFI has established strong thought to be the ancestor of most domesticated apples working relationships with local rangers. We chose smaller (Cornille et al., ). Both species are exploited heavily fieldwork areas within each site, to focus our research efforts  for their fruit and firewood (Morgan & Richards, ). on forested areas (Fig. ). Anthocyanin pigments give the blossom of Malus niedzwetzkyana a distinctive pink colour, making this species Methods easily distinguishable from other trees in these forests and giving it an iconic status with local people. The fruit is dis- Species distribution modelling tinctively red both inside and out because it also contains these pigments in its skin and flesh. Anthocyanins are fla- We selected the maximum entropy model, Maxent v. .. vonoid compounds whose antioxidant, anti-carcinogenic, (Phillips et al., ), as suitable for modelling the distribu- and anti-inflammatory properties confer extensive health tion of M. niedzwetzkyana because it is highly accurate with benefits (Bagchi et al., ; Boyer & Liu, ; Konczak small sample sizes (Elith et al., ; Qin et al., ), can be & Zhang, ). This species could therefore be an impor- run with presence-only data (Phillips et al., ), and is tant genetic resource that could lead to the development widely used in conservation science (Trisurat et al., ; of red-fleshed cultivars with benefits for human health Liang et al., ). We obtained occurrence records from (Sekido et al., ) whilst also retaining genetic diversity the Bioresurs project, which thoroughly sampled the wal- (Velasco et al., ). nut–fruit forests of the Jalal-Abad region (Cantarello et al., Only a few small, isolated populations of M. niedzwetzkya- ; Orozumbekov et al., ), providing  location na are extant, making this species exceptionally vulnerable points for M. niedzwetzkyana. to extinction, with the largest known population recorded Initially, we considered  environmental variables as po- in the Jalal-Abad region of Kyrgyzstan (Cantarello et al., tential indicators of the distribution of M. niedzwetzkyana. ; Orozumbekov et al., ). This area is the primary These variables are widely recognized as factors that affect focus of conservation efforts to maintain walnut–fruit forest plant distributions and have been effective in determining ecosystems in general and M. niedzwetzkyana in particular. the distribution of species in a similar threatened situation However, many conservation actions currently lack a clear to that of M. niedzwetzkyana (Kumar & Stohlgren, ). scientific basis. In this study we develop an evidence-based We obtained  bioclimatic variables and elevation data at  approach to the conservation of this species, supporting the  second resolution (c.  km ) from WorldClim (Hijmans growing movement for data-driven conservation initiatives et al., ). We generated slope and aspect raster files from (Sutherland et al., ). the elevation data, with the QGIS . terrain analysis tool We used species distribution modelling to identify (QGIS Development Team, ). A global land cover file climatically suitable areas for M. niedzwetzkyana across (GlobCover) was also included (Arino et al., ). Kyrgyzstan and to determine fieldwork locations. We de- We tested multicollinearity between all variables by cal- vised a threat assessment to provide insight into which of culating a Pearson correlation coefficient matrix. A correl- these climatically suitable sites should receive targeted fund- ation coefficient for each pair of variables was generated ing, and explored the most appropriate interventions for and using these values we selected the variable with the each location. We analysed population structures to deter- strongest correlation with the majority of predictors from mine the regeneration capacity of M. niedzwetzkyana within each group of highly correlated variables (r . .) to be in- the sites, and recorded ecological data across all survey sites cluded in the model (Kumar & Stohlgren, ; Qin et al., to develop a basic understanding of the species’ ecology. We ). This reduced the model to  parameters (Table ), present key information on the population and ecology of, making it more likely to detect variables that caused the and the threats to, this species and its habitat, to improve greatest effect (Porfirio et al., ). Because of the small local conservation decision making. This will enable better sample size (,  individuals), we ran the model using support for communities and protection of M. niedzwetz- only the linear, quadratic and hinge features of the kyana so that this species remains a key component of the Maxent programme (Elith et al., ). We selected %of iconic walnut–fruit forests of Kyrgyzstan and Central Asia. the Bioresurs tree location points as test data for the model;

FIG. 1 Study area in Kyrgyzstan, showing fieldwork sites and locations of Malus niedzwetzkyana individuals.

TABLE 1 List of model inputs, their source, and percentage contri- (Qin et al., ). We imported the results of both models bution to both models produced using Maxent software. into QGIS and presented them using the classification system employed by Kumar & Stohlgren (). % contribution % contribution (land cover (land cover Environmental variable included) excluded) Threat assessment Global land cover1 53.5 Precipitation of coldest 19.7 26.6 Using the Maxent ecological model and a forest cover raster 2 quarter (Hansen et al., ), we selected areas for fieldwork within 2 Isothermality 14.3 28.8 each of the four areas classified as having high ecological Slope3 6.8 18.2 suitability for the species. To ensure all landscape features Mean annual temperature2 2.0 16.7 Annual temperature range2 1.8 4.5 were accounted for we used stratified random sampling to Aspect3 1.7 2.1 distribute random sample points across all strata. We cate- Max. temperature of warmest 0.1 0.0 gorized the landscape of the fieldwork areas into seven strata month2 (clusters) using a K-mean cluster analysis (Rubin, ), 2 Precipitation of driest quarter 0.0 3.0 with each cluster capturing distinct landscape features. We Precipitation of wettest 0.0 0.0 2 produced the clusters using a digital elevation model ob- month tained from the Shuttle Radar Topography Mission (SRTM;  Source: GlobCover  (Arino et al., ).    USGS ; Jarvis et al., ), slope data (radians), Source: WorldClim (Hijmans et al., ).  aspect data (radians), relative slope position data calcu- Source: Generated in GIS from Digital Elevation Model. lated with the SAGA GIS ‘Basic Terrain Analysis’ module from SRTM, and NDVI (normalized difference vegetation all other settings were kept at their defaults (Qin et al., ). index) calculated from Landsat- images (NASA Landsat We produced two models: one using the global land cover Program). We generated five random GPS points within file, and one excluding it, to highlight climatically suitable each cluster at each site, with a minimum distance of  areas both within and outside areas currently forested. m between them, resulting in  points at each site. Where The model’s accuracy was determined by its area under the generated points fell in locations that were inaccessible the receiver operating characteristic curve (AUC) value: because of geographical features, we took measurements as AUC . . = very good; AUC .–. = good; AUC , . close as possible to the original position, to ensure landscape = uninformative (Swets, ; Baldwin, ). We used a sampling remained representative. We sampled additional jackknife analysis to determine the most important factors points where the landscape showed unexpected features affecting habitat suitability. Factors with a contribution to that needed to be captured to ensure model accuracy, result- the model of , % were deemed redundant predictors ing in  sites assessed.

We created a north facing quadrat of  ×  m(. ha) at to evaluate the extent of threats from firewood collection each site, using the GPS point as the centre. In a few areas that and livestock grazing (Supplementary Figure ). were difficult to access we reduced the quadrat size. This was We assessed the variation in mean undergrowth height recorded and accounted for when scoring the area by exam- between and within our sites as a proxy for undergrowth ining the extent of damage indicators and then scaling up damage. The Shapiro–Wilk test revealed variation in mean appropriately, ensuring the scoring remained comparable undergrowth height between sites to be non-normal and between all sites (Orozumbekov et al., ). so we transformed the data using a Box–Cox transformation Inside each quadrat we assessed damage caused by live- with the R package MASS (Venables & Ripley, ). We stock grazing and firewood collection, using six indicators ran a one-way ANOVA to determine whether there was sig- of grazing damage and four firewood collection indicators nificant variation of undergrowth damage between sites and (Supplementary Material ). The grazing indicators were a Post-hoc Tukey HSD (honestly significant difference) test adapted from Reimoser et al. (); the firewood indicators to determine the relationship between each pair of sites. were developed based upon FFI’s Sary-Chelek monitoring We plotted the frequency distribution of the DBH of programme and personal communication with the FFI trees at each site, to examine whether the distributions con- office in Bishkek. Each indicator was visually scored: ,no formed to the inverse J-shape, also known as type  struc- threat; , light threat; , moderate threat; , high threat; ture, of a healthy regenerative population (Newton, ; , very high threat (Reimoser et al., ). The indicators Orozumbekov et al., ). A stable population should were weighted equally to derive a mean overall score in have a large number of individuals in lower DBH classes; each quadrat. The firewood indicators ‘burning signs’ and fewer individuals are expected in the larger DBH classes be- ‘felled wood abundance’ were ignored if slopes were . ° cause of an increasing likelihood of mortality over time. We or if a settlement was within  m. We recorded altitude, also tested the relationship between the aspect, slope and slope angle, and aspect of each site, using a GPS and a clin- percentage canopy cover at M. niedzwetzkyana locations  ometer, and the mean undergrowth height (with a tape and that of the general environment using χ tests, to measure) and per cent canopy cover (estimated to the near- describe the ecology at tree locations. est % by eye, using the same observer for each measure- ment; Supplementary Material ). Results

Ecology and population structure Species distribution modelling

We used previously recorded M. niedzwetzkyana locations The Maxent models provided insight into historical forest from the Bioresus project and identified additional trees cover and guided our fieldwork. The model containing using ranger knowledge, where possible. We only measured global land cover data (Fig. )was‘very good’, with an trees found within our designated fieldwork areas as this is AUC of . ± SD ., showing the model performs where conservation interventions are currently planned. much better than a random model. The Maxent internal Trees were studied within three sites: Gava, Kara-Alma jackknife test highlighted global land cover as the most im- and Sary-Chelek. The fourth site, Dashman, contains no portant predictor of suitable habitat for M. niedzwetzkyana, M. niedzwetzkyana individuals. For each tree we measured explaining .% of the variation associated with species the diameter at breast height (DBH) and recorded its loca- location (Table ). Precipitation of coldest quarter (.%), tion. At each tree location we recorded altitude, slope angle, isothermality (.%), and slope (.%) were also significant aspect and ecological factors, as described for the threat explanatory variables. assessment quadrats. The Maxent model with global land cover excluded (Fig. ) was also ‘very good’, with an AUC of . ± SD .. Visually the range of M. niedzwetzkyana was much broader Data analysis in the model with land cover excluded. There was no clear single important predictor; isothermality (.%), precipita- We compared the distribution of threat scores for firewood tion of the coldest quarter (.%), slope (.%) and annual collection and livestock at each site, to identify which sites mean temperature (.%) were all important variables in were most damaged. We also compared the distribution of predicting suitable habitat (Table ). threat scores across all sites, for both firewood collection and  livestock grazing, using χ tests on contingency tables cre-     ated in R . . (R Core Team, ). When significant vari- Threat assessment ation was detected between sites, a further comparison  between each pair of sites was undertaken, again using χ The threat assessment provided information on which sites tests. The distribution of scores for each site was assessed were least threatened and the distribution of threats across

FIG. 2 Predicted suitable habitat for M. niedzwetzkyana in Kyrgyzstan, produced using Maxent software, (a) including land cover data and (b) excluding land cover data.

each site. There were significant differences between sites in Gava, and Dashman and Kara-Alma (Table ). Firewood col-  threat from livestock (χ = .,n=,P, .)and lection was a more localized threat than livestock grazing   firewood collection (χ = .,n=,P, .). We acrossallsites(χ = .,n=,P, .). scored Sary-Chelek Biosphere Reserve as the least threatened There were significant differences in mean undergrowth    ,   site overall, Kara-Alma as the second least threatened site, height between sites (Fig. ; F, = . ,P . ). A post and both Dashman and Gava were deemed equally degraded hoc Tukey HSD test showed there were significant differences and the two most disturbed sites visited. The ranking of loca- between Sary-Chelek and Dashman, Sary-Chelek and Gava, tions based on threat level was the same whether livestock Kara-Alma and Dashman, and Kara-Alma and Gava. Only grazing and firewood collection were considered separately Kara-Alma and Sary-Chelek, and Dashman and Gava pair-  or combined in an overall score. Pairwise χ tests revealed ings were not significantly different (P . .). This informa- that livestock threat was significantly different between all tion supports the results of our site ranking recorded in the site pairings, except between Dashman and Gava (Table ), threat assessment, with Sary-Chelek and Kara-Alma having and that variation in firewood collection was significantly dif- the highest undergrowth, and Dashman and Gava having ferent between all site pairings, except between Dashman and equally degraded vegetation.

TABLE 2 Statistical comparison between pairs of sites for the threats Ecology The canopy cover of areas with M. niedzwetzkyana  of firewood collection and livestock grazing, with the χ value, was significantly more open than the mean canopy cover in  degrees of freedom and probability. the surrounding landscape (χ = .,n=,P, .). Site comparison χ2 df P Malus niedzwetzkyana grows best where canopy cover is – Livestock grazing %. Furthermore, we found a significant difference Dashman – Gava 5.068 4 0.28 between the range of aspects found in the landscape Kara-Alma – Sary-Chelek 27.468 4 , 0.001* and the aspect where M. niedzwetzkyana is growing  Dashman – Kara-Alma 14.567 3 , 0.01* (χ = .,n=,P, .). Most commonly, this tree Kara-Alma – Gava 29.189 4 , 0.001* grows on south-west facing slopes. Some individuals were – , Dashman Sary-Chelek 18.959 4 0.001* found on slopes with other aspects, but none on slopes – , Sary-Chelek Gava 21.301 4 0.001* facing north. There was a significant difference between the Firewood collection – range of slopes found across the landscape and slope angles Dashman Gava 0.822 4 0.93 χ    Kara-Alma – Sary-Chelek 19.590 4 , 0.001* where M. niedzwetzkyana is growing ( = . ,n= , Dashman – Kara-Alma 8.186 4 0.08 P , .). Most individuals were found on slopes of –°, Kara-Alma – Gava 10.200 4 , 0.05* evenly distributed across this range. Dashman – Sary-Chelek 30.846 4 , 0.001* Sary-Chelek – Gava 32.896 4 , 0.001* *Significant difference. Discussion

Conservation action for the long-term protection of M. niedzwetzkyana and its habitat is currently limited by a poor evidence base. This study provides evidence that the walnut–fruit forest cover of Kyrgyzstan was once potentially more extensive and that current M. niedzwetzkyana habitat is restricted to a few fragmented forest areas. Most of these are seriously affected by regular firewood collection and livestock grazing; only a few areas in Sary-Chelek and Kara-Alma remain relatively undisturbed. The populations within these fragments are vulnerable to threats and regeneration is limited. We determined that M. niedzwetzkyana is usually located on south-west facing slopes (, °) with relatively open canopy cover, which we assume to be its nat- urally favoured conditions. This information will support the design of a more comprehensive and practical restor-

FIG. 3 Undergrowth height measurements for plots across ation plan for this Endangered tree. Dashman, Gava, Kara-Alma, and Sary-Chelek sites. The thick Walnut–fruit forest area is estimated to have declined by line on the boxplot is the median, the box represents the % in the past  years, with agricultural expansion and interquartile range, and the whiskers show the extreme data extensive clear-felling thought to be the driving factors,   points within . × the interquartile range. White dots are data although little evidence is available to quantify their full points outside of this range, which are assumed to be outliers. impact (Djanibekov et al., ). Today Kyrgyzstan’s forest area is relatively stable because these particular threats have Malus niedzwetzkyana database declined (Orozumbekov et al., ), but localized threats such as grazing still degrade the remaining forest. Our Population structure The locations of  M. niedz- species distribution modelling highlighted that M. niedzwetzkyana individuals were mapped across Kyrgyzstan wetzkyana occurs in currently forested areas. However, (Fig. ) and three subpopulations assessed for regeneration if forest area is ignored and only climatic conditions are capacity. The DBH distribution of M. niedzwetzkyana at considered, the potential species range appears to be consid- Gava and Kara-Alma showed that many DBH size classes erably greater. Much of this expanded area presently com- were underrepresented, indicating reduced regeneration prises mixed forest or cropland and can, therefore, guide capacity. The DBH of the Sary-Chelek population showed a future forest restoration projects. Previously, reforestation basic inverse J shape and Type I structure (Peters, ), has been a key strategy for maintaining forest areas in indicative of a population with some regenerative potential. Kyrgyzstan, with over , ha reforested during – However, this may be limited because of the low number of  (Orozumbekov et al., ). However, this was done individuals in some size classes (Fig. ). with the aim to increase resource availability and efforts

known trees can be protected and saplings planted at low cost. In addition, the current management of both sites is highly effective, implying strong community support for conservation in these areas. In contrast, Dashman and Gava contain extensively damaged tracts of forest, indicating a failing or absent management system. Dashman’s status as a Reserve does not reflect the park’s current use, and unlike Kara-Alma, the village of Gava’s Forestry Unit system is faltering and needs to be revitalized (BW, pers. obs.). There is insufficient capacity within the socio-ecology of these rural villages to reduce resource use and control livestock (Jalilova & Vacik, ). External help is needed to provide alternative livelihoods, funding and equipment to promote sustainable practices and conserve these forest ecosystems and the goods and services they provide. Consequently, the potential for large-scale ecological restoration in these extensively damaged areas is limited and community-based methods must be the initial focus. Across all sites, our threat assessment highlights livestock grazing as a greater threat than firewood collection because of the more extensive and less controlled nature of the former. We therefore suggest funding should be targeted initially at measures reducing livestock impact (e.g. temporary fencing), to allow selective grazing and facilitate natural regeneration of trees. This could improve the overall outlook for M. niedzwetzkyana. We updated the location map for M. niedzwetzkyana (Cantarello et al., ; Orozumbekov et al., ) within the sites, adding new individuals and confirming previously recorded locations. The low population density supports findings of previous research, highlighting the continued FIG.4 Malus niedzwetzkyana population structure for (a) Gava, (b) Sary-Chelek and (c) Kara-Alma sites. Trees with DBH .  cm vulnerability of the species and the importance of existing were grouped in size classes of  cm increments. Trees with DBH fragments as population strongholds (Orozumbekov et al., ,  cm were not included in the analysis because they have not ). Genetic isolation may also become a significant factor yet survived a single grazing season. in determining the persistence of this species (Young et al., ). Our study provides more detailed information on popu- have therefore focused on common, profitable species lations at specific sites than previous studies (Orozumbekov such as spruce Picea schrenkiana, ignoring rare species et al., ). It identifies both Kara-Alma and Gava popula- such as M. niedzwetzkyana. Our data suggest that reforest- tions as ageing, with limited regeneration, and shows that ation strategies could be broadened to include walnut–fruit only the Sary-Chelek population has some regeneration forest ecosystems and their constituent species, ensuring the capacity (but even this is limited by the low number of conservation of this iconic landscape whilst also continuing surviving trees). We conclude that the currently known to meet resource needs. population is not sustainable without external intervention, Because M. niedzwetzkyana populations are vulnerable providing justification for FFI’s planned sapling planting and resources for conservation are limited, it is important project (Fauna & Flora International, ). Moreover, to implement cost-effective conservation strategies within most trees are in areas susceptible to high threat, increasing the existing forest fragments (Mittermeier et al., ; the vulnerability of these populations. Bottrill et al., ). Our findings will help to focus re- In addition to the population assessment, the ecological sources to maximize the survival chances of M. niedzwetz- data we collected provides new information about the fun- kyana. Tall undergrowth and large areas with few or no damental niche of this species, suggesting that M. niedz- threats in Sary-Chelek and Kara-Alma indicate that these wetzkyana requires open areas of forest to flourish and sites should be the initial focus for conservation work in favours south-west facing slopes of , °. This information the region. Both sites provide an environment where can guide more effective conservation decision-making,

particularly regarding the selection of the most suitable BOTTRILL, M.C., JOSEPH, L.N., CARWARDINE, J., BODE, M., COOK, C., habitat for sapling planting projects. GAME, E.T. et al. () Is conservation triage just smart decision  – Our study comprises the first comprehensive assessment making? Trends in Ecology and Evolution, , . BOYER,J.&LIU, R.H. () Apple phytochemicals and their health of these sites and their threats since Orozumbekov et al. benefits. Nutrition Journal, , –.  ( ) and provides information on historical forest size CANTARELLO, E., LOVEGROVE, A., OROZUMBEKOV, A., BIRCH, J., and the ecology of M. niedzwetzkyana. The population of BROUWERS,N.&NEWTON, A.C. () Human impacts on forest M. niedzwetzkyana is in a critical state, requiring focused biodiversity in protected walnut–fruit forests in Kyrgyzstan. Journal  – research to target conservation strategies, which need to be of Sustainable Forestry, , . CORNILLE, A., GIRAUD, T., SMULDERS, M.J.M., GLADIEUX,P.& site-specific for maximum effect. The walnut–fruit forest ROLDA,I.() The domestication and evolutionary ecology of ecosystem continues to be utilized heavily, with local threats apples. Trends in Genetics, , –. remaining widespread (Orozumbekov et al., , ). CRITICAL ECOSYSTEM PARTNERSHIP FUND () Mountains of Increasing landscape degradation could result in more land- Central Asia. Http://www.cepf.net/resources/hotspots/ slides and soil erosion, which is already a problem in this Europe-and-Central-Asia/Pages/Mountains-of-Central-Asia.aspx [accessed  July ]. region, and reduced resource availability would have nega- DJANIBEKOV, U., VILLAMOR, G.B., DZHAKYPBEKOVA, K., tive effects on the rural population. Additional research is CHAMBERLAIN,J.&XU,J.() Adoption of sustainable land uses needed to assess the accuracy of our models, but consider- in post-Soviet Central Asia: The case for agroforestry. Sustainability ation should be given to expanding the limited walnut–fruit (Switzerland), , –.  forest area in line with the increasing recognition of the need ELITH, J., PHILLIPS, S.J., HASTIE,T.&DUDI,M.( ) A statistical explanation of MaxEnt for ecologists. Diversity and Distributions, for connectivity and reforestation worldwide (Newmark  –  , . et al., ). EPPLE,C.() A vegetation study in the walnut and fruit-tree forests We have provided evidence to guide conservation deci- of southern Kyrgyzstan. Phytocoenologia, , –. sions, a methodology that other projects could replicate, FAO () State of the World’s Forests. th edition. FAO (Food and and established a baseline for long-term monitoring. Our Agriculture Organization of the United Nations), Rome, Italy.  synthesis of broad and local-scale research and the FAUNA &FLORA INTERNATIONAL ( ) Conserving Fruit-and-Nut Forests in Kyrgyzstan. Https://www.fauna-flora.org/projects/ evidence-based conservation approach are critical for conserving-fruit-nut-forests-kyrgyzstan [accessed  January ]. effective conservation planning. FISHER,B.&CHRISTOPHER,T.() Poverty and biodiversity: measuring the overlap of human poverty and the biodiversity Acknowledgements This work was supported by the Global Trees hotspots. Ecological Economics, , –. Campaign. We thank Igor Lysenko, Mick Crawley and Bolot Tagaev FISHER, R.J., SCHMIDT, K., STEENHOF,B.&AKENSHAEV,N.() for sharing their knowledge on trees and GIS, Bakai Ernazarov for Poverty and Forestry—A Case Study of Kyrgyzstan with Reference to translating during fieldwork, Vicky Wilkins, Victoria Price, Jarkyn Other Countries in West and Central Asia. Livelihood Support Samanchina, and Akylai Kabaeva at Fauna & Flora International for Programme (LSP) Working Paper . FAO (Food and Agriculture their support and advice throughout the project, and the U.S. Organization of the United Nations), Rome, Italy. Geological Survey for providing the LANDSAT data. HANSEN, M.C.C., POTAPOV,P.V,MOORE, R., HANCHER, M., TURUBANOVA, S.A.A, TYUKAVINA, A. et al. () High-resolution Author contributions Project design: all authors; fieldwork and global maps of st-century forest cover change. Science, , – data analysis: BW and CC; writing: BW and CC; revisions: all authors. . HEMERY,G.&POPOV,S.() The walnut (Juglans regia L.) forests of Conflict of interests None. Kyrgyzstan and their importance as a genetic resource. Commonwealth Forestry Review, , –. Ethical standards This research was carried out in accordance with HEROLD,B.() The avian communities and their habitat associations the Oryx Code of Conduct. of the walnut fruit forests of southern Kyrgyzstan. Diploma thesis, Ernst Moritz Arndt University, Greifswald, Germany. References HIJMANS, R.J., CAMERON, S.E., PARRA, J.L. & JARVID,A.() Very high resolution interpolated climate surfaces for global land areas.  – ARINO, O., RAMOS PEREZ, J.J., KALOGIROU, V., BONTEMPS, S., International Journal of Climatology, , . DEFOURNY,P.&VAN BOGAERT,E.() Global Land Cover Map HOSONUMA, N., HEROLD, M., DE SY, V., DE FRIES, R.S., BROCKHAUS, for  (GlobCover ). © European Space Agency, Paris, France M., VERCHOT, L. et al. () An assessment of deforestation and & Université catholique de Louvain, Louvain-la-Neuve, Belgium. forest degradation drivers in developing countries. Environmental PANGAEA, Bremerhaven, Germany. Https://doi.org/./ Research Letters, , . PANGAEA. [accessed  April ]. JALILOVA,G.&VACIK,H.() Local people’s perceptions of forest BAGCHI, D., SEN, C.K., BAGCHI,M.&ATALAY,M.() Anti- biodiversity in the walnut fruit forests of Kyrgyzstan. International angiogenic, antioxidant, and anti-carcinogenic properties of a novel Journal of Biodiversity Science, Ecosystem Services and Management, anthocyanin rich berry extract formula. Biochemistry, , –. , –.  BALDWIN, R.A. () Use of maximum entropy modeling in wildlife JARVIS, A., REUTER, H.I., NELSON,A.&GUEVARA,E.( ) research. Entropy, , –. Hole-filled SRTM for the Globe. Version . Http://srtm.csi.cgiar.org   BORCHARDT,P.&DORRE,A.() Changing systems, changing [accessed May ]. effects—pasture utilization in the post-Soviet transition. Mountain KONCZAK,I.&ZHANG,W.() Anthocyanins—more than nature’s Research and Devevlopment, , –. colours. Journal of Biomedicine and Biotechnology, , –.

KUMAR,S.&STOHLGREN, T.J. () Maxent modeling for predicting QIN, A., LIU, B., GUO, Q., BUSSMANN, R.W., MA, F., JIAN, Z. et al. suitable habitat for threatened and endangered tree Canacomyrica () Maxent modeling for predicting impacts of climate change on monticola in New Caledonia. Journal of Ecology and Natural the potential distribution of Thuja sutchuenensis Franch., an Environment, , –. extremely endangered conifer from southwestern China. Global KYRGYZ GOVERNMENT,SWISS DEVELOPMENT COOPERATION,THE WORLD Ecology and Conservation, , –. CONSERVATION &ACADEMY OF SCIENCES OF K.R. INSTITUTE FOR FORESTS RCORE TEAM () R: a Language and Environment for Statistical AND NUT FARMING INSTITUTE FOR BIOSPHERE () Walnut–Fruit Forest Computing. R Foundation for Statistical Computing, Vienna, Action Plan. Bishkek, Kyrgyzstan. Austria. Http://www.r-project.org [accessed  July ]. LIANG, X., KANG,A.&PETTORELLI,N.() Understanding habitat REIMOSER, F., ARMSTRONG,H.&SUCHANT,R.() Measuring selection of the Vulnerable wild yak Bos mutus on the Tibetan forest damage of ungulates: what should be considered. Forest Plateau. Oryx, , –. Ecology and Management, , –. MITTERMEIER, R.A., MYERS, N., THOMSEN, J.B., DA FONSECA, G.A.B. RUBIN,J.() Optimal classification into groups: an approach for &OLIVIERI,S.() Biodiversity hotspots and major tropical solving the taxonomy problem. Theoretical Biology, , –. wilderness areas: approaches to setting conservation priorities. SECRETARIAT OF THE CONVENTION ON BIOLOGICAL DIVERSITY Conservation Biology, , –. () Global Biodiversity Outlook . Convention on Biological MORGAN,J.&RICHARDS,A.() The New Book of Apples. Ebury Diversity, Montreal, Canada. Press, London, UK. SEKIDO, K., HAYASHI, Y., YAMADA, K., SHIRATAKE, K., MATSUMOTO, NEWMARK, W.D., JENKINS, C.N., PIMM, S.L., MCNEALLY, P.B. & S., MAEJIMA,T.&KOMATSU,H.() Efficient breeding system HALLEY, J.M. () Targeted habitat restoration can reduce for red-fleshed apple based on linkage with s-rnase allele in ‘Pink extinction rates in fragmented forests. Proceedings of the National pearl’. HortScience, , –. Academy of Sciences of the United States of America, , –. SUTHERLAND, W.J., PULLIN, A.S., DOLMAN, P.M. & KNIGHT, T.M. NEWTON, A.C. () Forest Ecology and Conservation: a Handbook of () The need for evidence-based conservation. Trends in Ecology Techniques. Oxford University Press, Oxford, UK. and Evolution, , –. OROZUMBEKOV, A., MUSURALIEV, T., TOKTORALIEV, B., KYSANOV, SWETS, J.A. () Measuring the accuracy of diagnostic systems. A., SHAMSHIEV,B.&SULTANGAZIEV,O.() Forest Science, , –. rehabilitation in Kyrgyzstan. IUFRO World Series, , –. TRISURAT, Y., CHIMCHOME, V., PATTANAVIBOOL, A., JINAMOY, S., OROZUMBEKOV, A., CANTARELLO,E.&NEWTON, A.C. () Status, THONGAREE, S., KANCHANASAKHA, B. et al. () An assessment distribution and use of threatened tree species in the walnut–fruit of the distribution and conservation status of hornbill species in forests of Kyrgyzstan. Forests, Trees and Livelihoods, , –. Thailand. Oryx, , –. PETERS, C.M. () Sustainable Harvest of Non-Timber Forest Plant USGS (UNITED STATES GEOLOGICAL SURVEY)() Resources in Tropical Moist Forest: an Ecological Primer. Shuttle Radar Topography Mission. Global Land Cover Biodiversity Support Program, Washington, DC, USA. Facility, College Park, USA. Http://srtm.csi.cgiar.org [accessed  PHILLIPS, S.J., ANDERSON, R.P. & SCHAPIRE, R.E. () Maximum May ]. entropy modeling of species geographic distributions. Ecological VELASCO, R., ZHARKIKH, A., AFFOURTIT, J., DHINGRA, A., CESTARO, Modelling, , –. A., KALYANARAMAN, A. et al. () The genome of the PORFIRIO, L.L., HARRIS, R.M.B., LEFROY, E.C. & HUGH,S.() domesticated apple (Malus × domestica Borkh.). Nature Genetics, Improving the use of species distribution models in conservation , –. planning and management under climate change. PLOS ONE, , VENABLES, W.N. & RIPLEY, B.D. () Modern Applied Statistics with e. S. Springer, New York, USA. QGIS DEVELOPMENT TEAM () QGIS Geographic Information YOUNG, A., BOYLE,T.&BROWN,T.() The population genetic System. Open Source Geospatial Foundation. Https://www.qgis.org/ consequences of habitat fragmentation for plants. Trends in Ecology en/site/index.html [accessed  April ]. and Evolution, , –.