Tree species diversity and utilities in a contracting lowland hillside rainforest fragment in Central Vietnam

The study site

The study site was a secondary tropical evergreen broad-leaved monsoon forest fragment located on a west-facing hillside (elevation 120–340 m a.s.l.) within local authority of Hương Phú Commune (HPC) in Nam Đông District, Thừa Thiên-Huế Province, 30 km south of Huế City (Fig. 1). Nam Đông Valley extends south and is bounded by mountains in the east (Bạch Mã National Park), south and west (Central Trường Sơn or Annamite Ranges). The valley bottom is characterized by undulating topography with a patchwork of cultivated land, acacia plantations, and remnant forests. The forests of Nam Đông were traditionally used by Katu people (local minority ethnicity); today ethnic Vietnamese (Kinh) represent the majority (60%) of inhabitants in more urban areas of the district (Wetterwald 2003).

HPC-forest lies within the biodiverse but poorly studied and highly diminished/threatened ‘Northern Vietnam Lowland Rain Forests Ecoregion’ which extends for 22,500 km² along a narrow coastal strip northwards from Thừa Thiên-Huế Province. The forest is, however, also in close vicinity to the better protected and documented ‘Southern Annamites Montane Rain Forests Ecoregion’ which covers adjacent mountainous areas (Wikramanayake et al. 2002). The climate in Nam Đông is wet-tropical (mean annual temperature 24.5 °C ± 0.4 °C; annual precipitation 373 cm ± 103 cm; relative humidity 85% ± 3%) with a fairly dry period January-July (but sufficient intermittent rainfalls to maintain evergreen ‘rainforest’), and a very wet monsoon period August-December (Fig. 2). Owing to climatic, biogeographic and geological factors, forest biodiversity is generally very high (Sterling et al. 2006). In nearby Bạch Mã National Park >2100 plant species have so far been recorded in addition to a richness of wildlife (Dickinson and Thinh 2006; Huynh et al. 2016).

Valuable timber has been extracted throughout the 20^th century, changing the forests’ structure and composition. Acacia plantation forestry started in the 1990s (Thiha et al. 2007), displacing natural forests. Until 2003 HPC-forest was still embedded within a large forest area covering ~750 ha (Additional file 1: Figure S1). In 2010, when this study was conducted, HPC-forest was completely isolated within plantations, covering 76 ha. The cover was further reduced to 46 ha in 2014 and 39 ha in 2015. Twenty-three of forty plots surveyed in 2010 were thereby destroyed (Fig. 1). In 2015 the nearest contiguous forest was located 1.5 km east; another remnant forest (164 ha) was located 500 m north.

The vegetation survey

Within forty 20 m  ×  20 m square main plots (covering 1.6 ha or ~2% of HPC-forest) the vegetation was surveyed June-August 2010. The plots’ locations were determined in a semi-random fashion, using random points (set on Google Earth^TM maps) as ‘starting points’, but finally selecting locations along accessible pathways (ridges or creeks) closest to the points. Plots were at least 30 m apart. Ten plots were finally located on slopes along a southern hillside ridge, nine along a northern ridge; remaining plots were located in lower-lying areas on slopes or local ridges mostly near to creeks (Figs. 1 and 3).

Within the plots we recorded species of all dicotyledonous trees, saplings and shrubs with a diameter at breast height (DBH) >2.5 cm, as well as woody monocotyledonous plants (only palms were found) of all sizes. DBH was measured 1.3 m above ground using a tree girth tape. We refer to ‘saplings’ (P) meaning trees/treelets/shrubs with a DBH ranging between 2.5 and <6 cm, following IFRI (2004; correspondingly, palm seedlings/saplings were defined using plant weight categories, cf. Additional file 2: Table S1-m). Stems broken below 1.3 m were measured at break point, and notes were taken. For palms smaller than 2 m the bole diameter was measured at plant mid-height (monocot stems are approximately cylindrical, Brown 1997). Canopy height (bottom to tree crown top) was measured by two persons using an inclinometer and 50-m tape. The species and numbers of tree ‘seedlings’ (defined as woody dicotyledonous species with a DBH <2.5 cm, respectively smaller than 1.3 m) were recorded within five 2 m  ×  2 m sub-plots, four located in the corners of the main plot and one at its center.

Using a digital camera, images were taken of the leaves, twigs, and bark of any species which were unknown, and leaf samples were collected using a plant press. The field images and collected material were then compared to herbarium specimens at Huế University of Agriculture and Forestry (HUAF) and/or the Consultative and Research Center on Natural Resource Management (CORENARM) in Huế City, and were compared with descriptions in the literature (i.e. Ho 1999; FIPI 1996; Henderson 2009; Flora of China 2015). Species which could still not be identified were numbered A, B, C, etc. Taxonomy followed Flora of China (2015) and Tropicos (2015), i.e. the most recently updated sources.

Vegetation parameters derived from field data

A full list of recorded species, including summary statistics and additional information, is provided in Additional file 2: Tables S1 and S2. This includes (in footnotes) detailed descriptions of calculation methods (including allometric formulae) used to summarize the plant species data. MS Excel was used for calculations.

For all plots we calculated the tree densities (extrapolated to counts per hectare) and total tree basal areas (TBA in m²∙ha^–1; Bonham 2013), and the mean and maximum tree DBHs and canopy heights. Above-ground dry-matter woody biomass of dicotyledonous trees (eAGB_d in kg) was estimated using an allometric formula provided by Chave et al. (2005) and data on tree species’ wood densities provided by Zanne et al. (2009). The eAGB_p of palms was estimated using a formula by Goodman et al. (2013), setting wood density at 0.37 g∙cm^–3. The summed-up eAGBs of all trees per plot were then extrapolated to total above-ground woody biomass (TeAGB in Mg) per hectare.

The importance value index (IVI) was calculated for each tree species by summing up the species’ relative density, frequency and dominance (Curtis and McIntosh 1951); IVI served for overall ranking (IVIR) of the species’ importance. To describe plant species diversity on plots biodiversity indices (i.e. Shannon Index H′, Shannon Evenness J′, inverse Simpson’s Index 1/D) were calculated using formulae provided by Magurran (2007).

Physical and biological plot data recorded during the survey

Plot UTM coordinates and elevations were recorded using a hand-held GPS receiver (Garmin 12XL, GARMIN International Inc., Kansas City). Slope inclinations of plots were determined using an inclinometer and a 50-m tape. Using a gouge auger five random 40-cm-deep soil samples were taken in each of the five sub-plots; the 25 samples were mixed to be representative for one main plot. The samples were weighted in the field and after being oven-dried at HUAF, whereby relative weight differences provided indications of soil moisture on plots at respective times of sampling (termed ‘soil field capacity index’, FCI). The samples were then brought to the Asian Institute of Technology (AIT, Bangkok) for further analyses of soil pH (by pH meter), color and nutrients. Dried pestled soil samples were photographed (using a flash), and the RGB (red, green, blue) chromaticity values were determined from digital images using Adobe Photoshop. Soil organic matter (OM) was determined through weight loss on ignition, and total nitrogen (N) and phosphorus (P) were determined using the Kjeldahl method (Bremner 1960; Nelson and Sommers 1996).

Data analyses and assessments

Minitab 15 (Minitab Inc., State College, Pennsylvania), MS Excel and R software were used for calculations and statistical analyses. A vegetation classification was first performed on species’ biomass data (respective TeAGB plot data, transformed to normal distribution) by using dendrograms. Manhattan distance measure and Ward linkage method provided a dendrogram with distinct vegetation classes; these classes could also be pertinently explained from plot distributions within the hilly terrain (Fig. 3). Based on this classification a species association table was composed (following Kent 2012), illustrating floristic differences among vegetation types. The complete table is represented in Additional file 1: Table S1, whereas a sub-set of the sixty most important species (in terms of IVI) is shown in Table 2. More detailed information on the thirty most important species (and another four species of interest) is provided in Table 3. Species rank/abundance diagrams and rarefaction curves (including real data and corresponding models from log-logistic regressions) are shown in Fig. 5; these illustrate patterns of tree species richness overall and within the two main vegetation types. Summary information of various parameters (relating to vegetation structure and diversity, terrain and soils) is provided in Table 4. This Table also provides information (obtained from FIPI 1996; Ho 1999; Wiart 2003; Tanaka and Nguyen 2007; Henderson 2009; Flora of China 2015; Fern et al. 2016) on the recorded tree species’ utility (value categories: ‘no’ [0], ‘moderate’ [1], ‘good’ [2], ‘excellent’ [3] or ‘unknown’ [-]) of timber and non-timber forest products (NTFP; utility of fire wood was not assessed). Simple T-tests were used to assess differences of parameters (transformed to normal distribution) between the two main vegetation types (Table 4). In addition, in order to illustrate vegetation distributional patterns in relation to bio-physical gradients (in terrain, soil parameters, biomass and stocking density; Fig. 4) a canonical correspondence analysis was performed on the thirty most important species (accounting for 69% of TeAGB, Table 3).

Tree species diversity and floristic-structural forest patterns

In total 170 species of 88 genera and 45 plant families were recorded in a sample of 2,384 trees (of which 27.6% were ‘sapling’ sized) in the survey (on 1.6 ha). In addition, two rattan species were only recorded at ‘seedling’ size. Most species (128) could be clearly identified, but 32 were identified only to genus, and 12 rare species remained unknown. The most important plant families by tree count were the Arecaceae (representing 15% of all trees/shrubs; 15 species), Euphorbiaceae (7%; 15 spp.), Fagaceae (7%; 5 spp.), Burseraceae (7%; 5 spp.), Myristicaceae (6%; 3 spp.), Moraceae (6%; 9 spp.), Cannabaceae (5%; 2 spp.), Lauraceae (5%; 18 spp.), Myrtaceae (5%; 4 spp.), Sapotaceae (4%; 4 spp.), Sapindaceae (4%; 7 spp.), and Rubiaceae (3%; 3 spp.).

A majority of identified species (44%; including dominant species), had a predominantly southward-tropical distribution, but many species (36%) were distributed within the Sino-Himalayan mountain regions or were regional endemics (within the Trường Sơn Ranges; 18%) (Table 1). A large majority of species was rare; rank/abundance patterns did not differ substantially among different tree size strata (Fig. 4). No species was recorded on all plots, but several common species were found more or less throughout HPC-forest (Tables 2 and 3). Trees with a DBH ≥10 cm were represented by 129 species (i.e. 75%). Calculations using log-logistic models (Fig. 4) indicated that on average ~70.2 ‘large’ tree species (DBH ≥10 cm) and an additional ~23.4 smaller-sized species may be found on one contiguous hectare.

Biogeographic distribution (according to literature)	Species	Dominance (%)
	(%)	Trees	TeAGB
unknown	30.5	8.0	7.4
endemics (central Vietnam)	8.6	17.0	6.1
endemics (wider Vietnam)	7.2	1.6	1.7
Vietnam, Laos/Cambodia	3.9	4.0	10.4
Vietnam and Malaysia only	2.3	3.0	1.8
wider mainland Southeast Asia	17.2	15.7	12.5
wider Southeast Asia (w. insular)	25.0	25.4	22.4
Vietnam and South China only	17.8	10.0	12.6
Vietnam, China, Laos/Cambodia	5.5	5.9	7.2
wider South Asian-Himalayan	10.9	4.2	10.0
wider Asia (including temperate)	3.1	4.3	6.7
introduced species	2.3	0.8	1.1

‘Wider Southeast Asia’ includes islands of Malesia (Indonesia/Philippines). ‘Wider South Asian-Himalayan’ refers to a distribution mainly along the southern Himalayan arc from Vietnam/China to Burma-India-Nepal-Pakistan

Endemics of ‘central Vietnam’ occur only in Thừa Thiên-Huế and adjacent provinces

The species with an ‘importance value index rank’ (IVIR) from 1-60 are shown (abbreviation key of plant families at end of table). The species weight on the plots (400 m²) is indicated by a normalized biomass index, scaled between a minimum of 1 and a maximum of 100 (color scale in upper row). Inserted numbers in yellow indicate seedling counts on plots (sub-plots, area of 20 m²)

Information is provided on the species’ importance (IVIR/IVI), plot frequency, total basal area (TBA), and the total estimated above-ground dry woody biomass (TeAGB). For each main vegetation type (R and B) the average number of seedlings, as well as the number of small (biomass eAGB smaller than 10 kg), medium (10–100 kg), and large (>100 kg) trees is listed. Further information is provided from the literature (cf. Additional file 2)

The vegetation types were determined by species distributions (Tables 1 and 3; cf. Discussion Section); no distributional gradients within the vegetation were discernible in regards to entire plant families (Additional file 2: Table S1). Compared to R-forest, B-forest was characterized by significantly lower densities of trees, saplings and seedlings. Correspondingly, various tree growth parameters (mean DBH and height, maximum height, TBA, dead biomass) differed between R- and B-forest (Table 4, Fig. 4). Soil N and herbal ground cover were slightly higher in B-forest, but other soil parameters did not differ significantly (Table 4).

Tree species richness was significantly higher in R-forest as compared to B-forest. Model calculations indicated that around 99.0 tree species (thereof ~70.2 with DBH ≥ 10 cm) and 84.4 species (~62.9 with DBH ≥ 10 cm) would be found on average on one contiguous hectare of R- and B-forest, respectively (Fig. 5b). Correspondingly, plot-level species richness as well as Shannon and (marginally) Simpson diversity indices were higher in R- forest as compared to B-forest (Table 4).

Utilities and conservation values of forest tree species as inferred from literature

Of the trees (DBH ≥6 cm) present on plots 35% were in harvestable timber range with a DBH of ≥15 cm, and 8% with DBH ≥30 cm (Fig. 6). Among the 172 woody species recorded 12 species (7%) were assessed to provide timber of ‘excellent’ value, 32 (19%) of fairly ‘good’, 44 (25%) of ‘moderate’, 41 (24%) of ‘insignificant’ (or undescribed), and 43 (25%) of unknown value. Only 4.7% of the biomass (TeAGB) stock on the plots was in ‘excellent’ harvestable timbers. The biomass of ‘good’ timber was slightly higher in B-forest compared to R-forest (Table 4). In between 2010 and 2015 larger portions of B-forest (70% of plots) were converted to acacia as compared to R-forest (45%; Tables 1 and 4), but cut and non-cut plots did not differ significantly in terms of TeAGB of ‘excellent’ and/or ‘good’ timber.

Most identified species were known to provide some potentially useful non-timber forest products (NTFPs) ranging from edible fruits and nuts, saps used for medicine, to materials (e.g. rattan rods, dies, leaves) useful for handicrafts (Table 3). Eight species (5%) could provide NTFPs assessed as ‘excellent’ value, 35 species (20%) of ‘good’, 41 (24%) of ‘moderate’, 43 (25%) of ‘insignificant’ (or undescribed), and 45 (26%) of unknown value. The TeAGB of tree species providing ‘excellent’ NTFPs was relatively high (14.7%). Compared to R-forest, B-forest was characterized by higher ratios of trees (in terms of TeAGB) providing ‘excellent’ NTFPs, and lower ratios of trees with ‘moderate’ NTFPs, but due to high variability among plots the differences between vegetation types were only marginally significant (Table 4). The plots which were cut in between 2010 and 2015 were, however, characterized by a significantly (T-Test: t = 2.96, p = 0.005) higher biomass ratio (20.0% ± 16.0%) of ‘excellent’ NTFP species compared to the non-cut plots (7.4% ± 8.4%) (other differences being insignificant).

Examining the ‘natural history’ and ecology of tree species diversity within HPC-forest

Species richness in Indochinese rainforests, and especially from Yunnan southward along the Trường Sơn Range, is very high, with up to ~200 woody species per ha recorded in southern Yunnan (Lü and Tang 2010). Often >100 tree species (with DBH ≥10 cm) per ha are found, but at this scale of assessment the region’s diversity cannot compare to the most diverse forests in South America and insular Southeast Asia where often >250 species∙ha^–1 are recorded (Wikramanayake et al. 2002; Corlett and Primack 2011; Corlett 2014). HPC-forest has been significantly disturbed and modified by human activities. Nonetheless, with an estimated average of ~70 tree species (DBH ≥10 cm) per contiguous hectare, the forest contained about 1.0–3.0 times as many species as were found on comparably-sized and often better-protected sites of evergreen broadleaf forests in locations in Northern Vietnam, 0.7–1.8 times the species recorded on plots in the Central Coast, 0.8–2.0 times species richness in the Central Highlands and in Southern Vietnam, and 0.7–2.7 times the richness in southern tropical Yunnan, China (Cao and Zhang 1997; Blanc et al. 2000; Wode 2000; Tran et al. 2013a; 2010a; 2015; Corlett 2014; Thinh et al. 2015). At slightly higher elevations (280–380 m a.s.l.; Fig. 1) maximally 48 species∙ha^–1 were recorded by Ha (2015) in Bạch Mã National Park. Even though species such as Gironniera subaequalis, Palaquium annamense and Scaphium macropodum were abundant in both studies, only 26 of 63 species recorded by Ha (2015) were also found in HPC-forest. This points at appreciable changes in floristic composition along relatively short spatial gradients - as was itself observable within HPC-forest (Table 2).

The floristic composition and associated structure of HPC-forest is an outcome of inherent longer-term (evolutionary/biogeographic) and shorter-term (ecological/anthropogenic) processes. The composition largely reflects the region’s intersection of southern wet-tropical (Southern Indochinese, Malesian) and more northern sub-tropical mountain-associated (Sino-Himalayan, Indo-Burmese) floras (Wikramanayake et al. 2002; Averyanov et al. 2003). Like in other parts of Southeast Asia (Ashton 2003; Culmsee et al. 2011; Culmsee and Leuschner 2013) the vegetation in the montane and sub-montane parts of the Trường Sơn Ranges is characterized by species from plant families of temperate northern origin, and partly of southern or Ancient Asian origin (Rundel 1999; Kuznetsov and Guigue 2000; Averyanov et al. 2003; Sterling et al. 2006; Huynh et al. 2016). In HPC-forest, tree species with a mostly submontane-colline distribution (Castanopsis sp., Syzygium spp., Cinnamomum spp., Canarium album, Cratoxylum pruniflorum, Illicium griffithii, Engelhardtia spicata, Melicope pteleifolia, Breynia fruticosa, Artocarpus styracifolius) were common along the higher-lying ridges of the hill (R-forest), and especially on the hill top (R1t). In contrast, species with a wide distribution in Southeast Asian (including Malesia) wet-tropical lowland forests (e.g. Gironniera subaequalis, Artocarpus rigidus, Morinda citrifolia, Garuga pinnata, Knema globularia, Litsea glutinosa, Schefflera octophylla, Pometia pinnata, Elaeocarpus griffithii, Plectocomia elongata, Mimusops elengi) were mostly either distributed throughout the forest or on more moist plots on lower slopes at the hill bottom (B-forest) (Tables 2 and 3).

In addition to many widely-distributed species, a strong endemic Vietnamese-Laotian (Annamite) floristic component was recognizable (Table 1). Several important tree species (e.g. Palaquium annamense, Schefflera violea, Wrightia annamensis, Hopea pierrei, Madhuca pasquieri) were Vietnamese endemic representatives of widespread tropical genera. Thừa Thiên-Huế and adjacent provinces are a hotspot of palm diversity and endemism (Henderson 2009), and 40% of the palm species in HPC-forest were indeed endemics. Some tree and palm species could not be clearly identified due to the absence of distinguishing plant parts. Some of these unidentified species could possibly represent more than one species: for example, in nearby Bạch Mã National Park several species of Castanopsis (16), Syzygium (19) and Ficus (35) have so far been recorded (Huynh et al. 2016). Furthermore, it is not unlikely that some unidentified species represented local endemics not yet described by botanists at the time of the study. For example, the abundant (IVIR 3) Licuala sp. was possibly the only recently described Licuala dakrongensis (Henderson et al. 2010), and Hopea sp. was perhaps Hopea vietnamensis (Hoang et al. 2013). The taxonomy of several species is still contested. For example, Knema globularia was listed in other sources (FIPI 1996; Ha 2015; Flora of China 2015; Huynh et al. 2016) as either K. corticosa, K. conferta or K. tonkinensis. Following closer genetic analyses new ‘cryptic’ species may be discovered among the long list of currently described taxa (Bickford et al. 2006).

Several hypotheses have been proposed to explain the high diversity and endemism (respectively ‘pseudo-endemism’, Turvey et al. 2016) in forests along the Trường Sơn Ranges (Wikramanayake et al. 2002; Sterling et al. 2006). The floristic composition of Indochina is partly a result of ancient plate tectonic collisions, and resulting terrains and associated local climates (Hua 2013; Corlett 2014). More recent processes during the climatically variable quaternary period, however, appear most important to explain today’s biogeographic patterns; such processes include sea level changes, cool/dry periods when forests retreated and were isolated, and climatic shifts along elevational gradients (Woodruff 2010). Evidence suggests that the Trường Sơn Ranges served as a refugium for wet-tropical forest species during cool/dry periods in the quaternary, but the region was itself also a center of species diversification (Sterling and Hurley 2005; Sterling et al. 2006; Corlett 2014). The ‘refugium hypothesis’ is supported by the fact that some species found in lowland Central Vietnam (e.g. Knema pierrei, Elaeocarpus apiculatus, Ficus glandulifera in HPC-forest) also occur in Peninsular Malaysia but are not found in lowland forests in between.

High species richness reflects the many rare species in HPC-forest. Numerous hypotheses have been proposed to explain co-existence of many tree species within a small forest area; these mostly consider the range of ecological niches available along small environmental gradients and the spatial impacts (often density-dependent) of a vast diversity of pests and diseases (often host-specific) at different plant life stages (Wright 2002; Freckleton and Lewis 2006; Ashton 2014). Within HPC-forest, species’ habitat descriptions from the literature largely fitted their distributions, with many species described as typically occurring on hill slopes and/or on humid soils (Table 3; Additional file 2: Table S2). The boundaries of vegetation types were however blurry (Table 2). In accordance with findings of other regional studies (Potts et al. 2002; Cannon and Leighton 2004; Noguchi et al. 2007; Du et al. 2013), this suggests that tree species distributions were strongly influenced by a mixture of relatively stochastic (e.g. propagule dispersal, forest disturbance) and more deterministic (e.g. factors relating to habitat) processes across environmental gradients (in HPC-forest largely governed by elevation and terrain, and corresponding soils; Fig. 4). In a study by Nguyen et al. (2016) in nearby A Lươi District 16 out of the 18 most abundant species on a two-hectare study plot (elevation 625–660 m a.s.l.) showed an aggregated distribution in the range of up to 15 m. Furthermore, many significant spatial associations were observed between different tree species. The authors mainly interpreted the patterns as evidence for the species’ dispersal limitations and inter-specific processes of herd protection and/or facilitation. Patterns within the ‘late successional’ forest were, however, barely discussed in regards to micro-topography, soils and potential earlier disturbances. In HPC-forest anthropogenic disturbances in the vegetation were fairly evident.

Anthropogenic influences on forest composition and natural processes

One can only guess what HPC-forest may have looked like before the influence of humans. The forest probably contained many more upper-canopy (20–30 m) and emergent (up to >40 m) trees of ‘excellent’ (hardwood) timber quality, e.g. dipterocarps (Hopea and Parashorea spp.), legumes (Erythrophloeum fordii, Sindora and Peltophorum spp.), and representatives of other plant families (Madhuca pasquieri, Mimusops elengi, Tarrietia javanica, Elaeocarpus spp.). These species (except E. fordii) were still found in HPC-forest, however at low numbers (0.6–6.3 trees∙ha^–1) and small sizes (Additional file 2: Table S1). The only common species with ‘excellent’ timber quality was a slow-growing undergrowth tree (Melanorrhoea laccifera, IVIR 17). Due to its small timber volume it was probably not a target during SFE logging operations; in addition, it was characterised by good regeneration (Tables 2 and 3). In a study in three communes in Nam Đông District (Ngo and Webb 2008; 2016), several excellent timber species (in particular Hopea pierrei, Parashorea stellata, E. fordii, M. pasquieri and Sindora tonkinensis) were mentioned by villagers as species indicating ‘relatively intact forest’. Only H. pierrei and M. pasquieri were however found on field plots, which supports the presumption that the forests have significantly changed due to selective logging.

Logging-out of high-value timber was probably most severe during the 1970s to late 1980s, and some selective extraction by villagers or ‘wood poachers’ may have continued thereafter. Nonetheless, in 2010 the overall per-hectare standing biomass (TeAGB) of harvestable timber (DBH ≥15 cm) was sizeable, with 96 Mg (equivalent to 169 m³) in B-forest and 109 Mg (~212 m³) in R-forest. Per-hectare timber stock of larger trees (DBH ≥30 cm) was 56 Mg (~101 m³) in B-forest and 61 Mg (~125 m³) in R-forest. HPC-forest would thus be classified in Vietnam mostly within the range of ‘average’ forest (100–200 m³∙ha^–1). This is however still markedly lower than primary rainforests which can reach harvestable timber stocks of considerably more than 300 m³∙ha^–1 (‘very rich’ forest; MARD 2009), with above-ground biomass exceeding 350 Mg∙ha^–1 (Hai et al. 2005, and literature cited therein).

Relatively large trees of species with lesser timber quality were found on many plots in the forest (Table 3). Gironniera subaequalis and Artocarpus rigidus (IVIR 1 and 2; with ‘good’ timber quality) were reportedly not a main target for logging by SFEs, but are occasionally used by villagers for house building and furniture (D.T. Ngo, pers. comm.). These species’ high abundance appeared to be partly explained by their dispersal and recruitment capacity under canopy shade (300–2050 seedlings per ha), and associated persistence during and after selective logging operations (including coppicing in the case of G. subaequalis, and tenacity in steep terrain in the case of A. rigidus; Tables 2 and 3; Raich and Khoon 1990; Ding et al. 2012). A. rigidus was possibly also promoted by local villagers because of the tree’s valued sweet fruits and timber (Wetterwald 2003; Lim 2012). In contrast, the success of G. subaequalis may be explained by ecological competitiveness under post-logging conditions. G. subaequalis has the capacity to enrich available N in its tissues (roots and leaves) and use it to produce acid phosphatase to efficiently exploit the more limited levels of available soil P – possibly at the expense of other trees which cannot compete for P in its presence (Huang et al. 2013).

Ngo and Webb (2008, 2016) listed G. subaequalis, Palaquium annamense, Syzygium spp. and Canarium album as indicators of ‘relatively intact forest’, whereas A. rigidus and other species (including Horsfieldia amygdalina, Garcinia cochinchinensis, Knema pierrei, Castanopsis spp., Pometia spp., Gonocaryum maclurei) were rather associated with ‘selectively logged forest’. Yet, prime indicators of impacted forest are probably species that do not regenerate under canopy shade. Such species include Garuga pinnata, Lithocarpus amygdalifolius, H. amygdalina and possibly Syzygium jambos – species which in HPC-forest were common in the upper tree size strata but lacked representation by seedlings (Table 3). G. pinnata and H. amygdalina were possibly dispersed by birds into forest gaps and may have profited from the aftermath of logging, but L. amygdalifolius and S. jambos (both with ‘good’ timber qualities and presumably more locally dispersed propagules) were probably mainly promoted after logging via active seeding/planting by villagers. The fact that these species were primarily found in B-forest (and sub-types R1t and R1b; Table 2) suggests that this was indeed the most heavily impacted part of the forest. The elevated herbal ground cover, lower seedling and sapling densities, higher tree mortality and associated lower tree density, and lower species richness and diversity were further indications that B-forest had been more disturbed than R-forest, including effects such as compaction and erosion of soils (Table 4, Fig. 4; Sidle et al. 2006; Clarke and Walsh 2006; Chazdon 2014). Wood collection by villagers and other on-going impacts in better-accessible B-forest may also partly explain the low sapling densities (Hoang et al. 2011; Popradit et al. 2015).

The ‘restoration history’ of HPC-forest is not known in detail. In Vietnam, Thailand and other tropical countries various species of Lithocarpus, Castanopsis, Syzygium, Artocarpus, Ficus, Cinnamomum, Horsfieldia, Garuga, Canarium and Garcinia have, however, been used for more or less successful direct seeding of degraded forests (Woods and Elliott 2004; Cole et al. 2011; Tunjai 2012) and/or as framework trees for forest restoration (Blakesley et al. 2002; Elliott et al. 2003; Khopai and Elliot 2003; FSIV 2003; Acharya and Kafle 2009). Various species of these and other genera (some with capacity to coppice; Table 3) were most probably also promoted in HPC-forest by villagers during initiatives to upgrade and ‘enrich’ local logged-out forests, whereby species selection/promotion was a mixture of practical ecological considerations (what can grow readily under specific conditions?) and the potential prospective utility of planted trees (van Kuijk 2008; next Section). Human influences thereby interacted with natural processes. In a study by Khopai and Elliot (2003) in Northern Thailand G. pinnata, Glochidion eriocarpum and species of the genera Artocarpus, Ficus, Litsea, Antidesma, Dillenia, Castanopsis, Scaphium and Cratoxylum were readily dispersed by animals to forest restoration sites where framework species had previously been planted.

NTFPs and associated values of HPC-forest

The potential utility of many species for NTFPs (in addition to timber) largely explained their presence and/or relative abundance within HPC-forest. Native species which were valued for food (fruits, nuts or seeds) were cultivated and/or promoted by local villagers since time immemorial, and were additionally favoured during government-sponsored forest restoration activities in the 1980s and 1990s. Amongst the thirty most important species were the highly valued fruit tree species Morinda citrifolia (Indian mulberry or noni), Garcinia cochinchinensis (false mangosteen), Scaphium macropodum (Vietnamese malva nut), Syzygium jambos (Malabar plum), Artocarpus rigidus (monkey jackfruit), Canarium album (Chinese olive), Litsea glutinosa (Indian laurel), and Pometia pinnata (island lychee) (Table 3). The three non-native species found in HPC-forest were economically valuable either for fruits (Morus alba - white mulberry, native to China; fairly common, IVIR 35; Nephelium lappaceum - rambutan, native to Malesia; one tree recorded) or camphor wax (Cinnamomum camphora - camphor laurel, native to Taiwan; two trees). The presence of valuable trees does obviously not entail that nearby communities actually collected and used all potentially available NTFPs, nor that they still tended the tree resources in sustainable ways. For example, malva nut trees (S. macropodum; IVIR 18, with highest woody biomass; Table 3), often appeared to be damaged, presumably because the nuts were harvested by cutting entire branches (YT Van, pers. obs.; Huy 2012).

Earlier data from Wetterwald (2003, 2004) gathered in Katu communities in Nam Đông indicated that for 14% of households collection of NTFPs was a mainstay for their livelihoods, 48% collected NTFPs part-time, whereas 38% occasionally or never collected NTFPs. More than half of the households interviewed (mostly those engaged in collecting) traded with NTFPs, whereby the commercially most valuable plant products (besides honey and mushrooms) were cane and baskets from rattans, bamboo shoots, malva nuts and other fruits (e.g. rambutan and rainforest figs), and bark from Litsea spp. (used for incense and medicine). On the plots in HPC-forest no bamboo species and none of the traded rattan species listed by Wetterwald (2003), McElwee (2010) or Polesny et al. (2014) were recorded. The most valued rattan species found in HPC-forest (i.e. Calamus nambariensis and C. rhabdocladus; Henderson 2009) were also the rarest (one seedling each; IVIR 171 and 172). Analogous to valuable timber species, important NTFP which were harvested as whole plants (or substantial plant parts) were therefore largely depleted in HPC-forest. Correspondingly, respondents in a study by Tran et al. (2010b) reported that availability of tradeable NTFPs has markedly declined in Nam Đông during the preceding decade (associated incomes have about halved), except for ‘broom grass’ (Thysanolaema maxima) which prolifically grows in gaps created in forests. Over-exploitation of rattan has become a major problem in the buffer zones of Bạch Mã National Park where new schemes for participative forest protection and resource monitoring have recently been initiated (Ha et al. 2016).

The vanishing of many commercially valuable wild species coincides with changes in local people’s livelihoods and socio-economic interactions during the past two decades. Soon after forest land allocation (FLA) during 1999-2005, home gardens and orchards were expanded and planted with fruit trees (from Artocarpus to Nephelium spp.) and acacia woodlots (for fuel and timber); these covered many household needs and diminished dependencies on fluctuating or declining forest resources (Wetterwald et al. 2004; Tran et al. 2013b). Cash income from expanding larger-scale commercial acacia plantations has become ever more important, however with an income gap opening between rich landholders (making good profits from acacias) and poor households (becoming more marginalized and dependent on casual/subsidiary labor) (Tran et al. 2013b; Bayrak et al. 2015). In accordance with expanding cash crops, the collection of wild products has also become more commercialized, with rich households often maintaining controls over collection/trade of key products (McElwee 2008; Sikor and To 2011; Tran et al. 2013b). Programs of community-based forest management (CFM) were launched in 2006 and piloted in Nam Đông (Tran et al. 2010b), but strong socio-economic dynamics linked to market forces and emerging cash crops outweigh many efforts directed at the conservation and sustainable use of remaining communally managed natural forests. Households in Nam Đông now increasingly use alternative materials to forest products, such as concrete and plastic for house construction, synthetic medicines instead of medicinal plants, and gas for fuel (Tran et al. 2013b). Fewer and fewer people know about traditional forest products, especially the vast diversity of medicinal plants, i.e. 432+ species that were formerly collected in forests around Bạch Mã National Park (Tran and Ziegler 2001; Tran et al. 2013b).

Losing economic value, losing ground: considerations on forest biodiversity conservation

The decreasing appreciation and use of traditional forest products relative to highly-priced merchandises (acacia wood, rubber, meat products) largely explains the demise of natural forests in the lowlands. Today virtually no rainforests remain in the plains and low-lying hills (below ~200 m elevation) in between Nam Đông and the coast, and acacia plantations continue to encroach into rainforest areas in colline and sub-montane rim-zones (Cochard et al. 2017). It is not unlikely that some specialized endemic lowland species have already become extinct. Within the ‘Northern Vietnam Lowland Rain Forests Ecoregion’ few forests remain (almost all affected by logging) in nine small protected areas and numerous fragments. Most of these forests have not been scientifically surveyed. The protected areas (covering ~3.9% of the ecoregion) barely embrace the entire ecoregion’s high biodiversity and endemism; hence, any remaining forest fragments – whatever their small size – are likely of considerable conservation value (Wikramanayake et al. 2002; McElwee 2016).

While extensive intact forestlands are needed to safeguard key biodiversity of rainforest ecosystems, considerable numbers of endangered plant and animal species (especially smaller species) can persist in forest fragments for decades (Turner and Corlett 1996; Hernández-Ruedas et al. 2014). Forest fragments can be important refugia for economically valuable species, and provide a biological store and nucleus from where forest regeneration of surrounding areas can be re-seeded under favorable conditions (Chazdon 2014; Sloan et al. 2016). Besides the presence of little known or even undescribed species, several species found in HPC-forest were listed in the IUCN Red List (IUCN 2016) or the Vietnam Red Data Book (2007) as ‘endangered’ (Hopea pierrei, Sindora tonkinensis, Tarrietia javanica) or ‘vulnerable’ (Knema pierrei, Mangifera minutiofolia, Elaeocarpus apiculatus, Madhuca pasquieri). This underlines the potential of the remaining forest fragments to maintain threatened species, especially species restricted to lowland forests.

To what degree (and in what ways) forest fragments can maintain assemblages of historically occurring species depends on many factors. Fragments only retain a ‘sample’ of the species present within the original non-fragmented forest, and the maximum number of species is related to fragment size (Laurance et al. 2011). For HPC-forest the extrapolative application of the log-logistic model equation (Fig. 5b) indicated that in total perhaps 255 tree species would have been found in the entire HPC-forest fragment (76 ha) in 2010. From simple loss of area this number would have decreased to ~237 species in 2014 (46 ha) and ~231 species in 2015 (39 ha). In contrast, ~319 species were possibly present in the larger connected forest in 2003 (742 ha) and ~287 species may still be found in the total forest area remaining in 2015 (203 ha, including another larger fragment; Additional file 1: Figure S1). The calculations suggest that species losses may be moderate (–28%, 2003–2016) relative to the substantial (–95%) decrease in area (considering HPC-forest alone). Calculations focusing on area alone, however, disregard potential longer-term impacts of fragmentation. Deforestation and reduction of HPC-forest has occurred over a period of just a few years. Hence, in 2010 many slow fragmentation-associated degradation processes were not yet distinguishable from effects associated with other impacts (logging, forest uses). Fragmentation effects will however become more obvious – provided the remaining fragment actually persists.

Tree communities in forest fragments that are initially diverse may biologically simplify as a result of isolation and reduced populations of many species. Given that pollinators may become functionally extinct within the fragment (Brosi et al. 2008) and can often not traverse the surrounding matrix (Kormann et al. 2016), rare plant species may eventually disappear from the fragment due to failing pollination and/or inbreeding. Ha (2015), for example, found that allelic diversity of seedlings of the dipterocarp Parashorea stellata was significantly reduced in forest fragments as compared to seedlings in contiguous forest in Nam Đông - despite observed inter-fragment genetic exchange. In several studies (e.g. Lopes et al. 2009; Breed et al. 2012; Zambrano and Salguero-Gómez 2014; Lowe et al. 2015 and studies cited therein) lowered tree fecundity was important to explain differences in tree seedling establishment between fragments and contiguous rainforest. In HPC-forest a majority of species was rare (Fig. 5a). In the fragment remaining in 2016 (39 ha) species recorded by three (9% of all species), two (14%) or merely one tree (25%) may in the fragment on average be represented by 73, 49, or 24 trees, respectively. Seedlings were recorded for only 14% of these rare tree species, and seedling frequencies of species with 1–5 trees∙ha^–1 were on average (3.5 seedlings/tree) almost three times lower than for species with more trees (10.1 seedlings/tree). This indicates that there was an on-going fragmentation-related trend towards losses of certain rare species.

Introductions of seeds from outside the fragment may keep small tree populations viable and/or change the species composition in the fragment. The most important seed dispersal agents in Indochinese rainforests are birds; accordingly, fruit sizes of many rainforest trees are relatively small (Corlett 2014; Table 3). In HPC-forest some common passerines such as babblers, thrushes and white-eyes were observed, and these may have brought in seeds from species such as Melicope pteleifolia, Glochidion eriocarpum, Eurycoma longifolia, Knema pierrei and Schefflera octophylla which were represented by widely-scattered seedlings (Tables 2 and 3; Datta and Rawatt 2008; Snow 1981; Kitamura et al. 2002; Wang et al. 2010). Tree species with larger fruits (>25 mm diameter), such as Pometia spp., Garcinia cochinchinensis, Artocarpus rigidus, Dillenia scabrella, Gonocaryum maclurei and palms were more likely dispersed by civets and/or rodents (especially squirrels), and probably just within the fragment (Kitamura et al. 2002; Corlett 2014). Species with winged seeds (Cratoxylum pruniflorum and Tarrieta javanica, but probably not dipterocarps; Au et al. 2005; Smith et al. 2015) may also be partly sourced from far-away areas by wind (Table 2). No connected lower-lying forests persisted near HPC-forest; the nearest contiguous forest (1.5 km east of HPC-forest) ranged from 300–1,100 m elevation. Hence, while many remaining lowland species within HPC-forest were presumably about to decline, there was possibly an increasingly higher colonization chance by more generalist-montane species. Several species recorded in HPC-forest (120–330 m elevation) had an altitude distribution range up to well over 1,000 m, and according to the literature some bird-dispersed species (e.g. Disepalum plagioneurum, several Litsea spp., Archidendron clypearia, Turpinia cochinchinensis, Eurya nitida; Kitamura et al. 2002) would not be expected below ~500 m.

Data described by Ha (2015) from older forest fragments (estimated 30 years since fragmentation) in Nam Đông indicated that tree species richness in fragments larger than 50 ha was not different to the richness within comparable contiguous forest (~40-48 species∙ha^–1), but smaller fragments were relatively impoverished (i.e. 28–32 species∙ha^–1 in 10-23 ha-sized fragments, and 14–21 species∙ha^–1 in 1–5 ha fragments). Ha (2015) attributed this mostly to fragment edge-effects such as lowered humidity, higher temperatures, and increased light and wind – i.e. effects which may increase mortality of sensitive trees and seedlings, and simplify the vegetation by promoting generalist species (Laurance 1997; Benitez-Malvido and Martínez-Ramos 2003; Bennet and Saunders 2010; Cochard 2011). Edge-effects in small fragments are disproportionally higher than in large fragments; simultaneously, anthropogenic impacts may increase in smaller fragments (Arroyo-Rodríguez et al. 2015).

In 2010 the plots were still located well inside HPC-forest, and edge-effects per se were thus barely obvious. The fragment boundary was, however, pushed back 2010–2015, and more than half of the plots were cut. Clearance was highest in the most accessible lower-lying forest parts; these parts were also relatively rich in harvestable ‘good timber’ (e.g. Lithocarpus amygdalifolius, Scaphium macropodum, Pometia spp. and Syzygium jambos; Tables 2 and 3). The destruction mostly affected the already previously highly impacted and modified B-forest, but also species-rich parts in R2- and R1b-forest (Table 2, Fig. 4). Judging from plot data a large fraction of trees with ‘excellent’ NTFPs was destroyed (77% in terms of TeAGB), including most (91%) of the commercially valuable malva nut trees (S. macropodum). This lends support to the note that the boom of acacia wood production currently outcompetes the economic values of many other commercially tradeable forest products.