Early Selection of Tree Species for Regeneration in Degraded Woodland of Southeastern Congo Basin

: Miombo woodland (MW) has several multi-purpose forest species, which are over-exploited for agriculture, charcoal and logging. Industrial plantations are among the promising solutions for sustainable management of MW, but high-yielding MW species are still lacking in the forestry sector. In this context, the present study assessed the growth of MW species, with respect to their early (ES) or late (LS) successional status. Seedling development was assessed for eight tree species, which were ES ( Combretum collinum , Pterocarpus tinctorius ) and LS ( Brachystegia boehmii , B. longifolia , B. spiciformis, B. wangermeana , Julbernardia globiﬂora , J. paniculata ), 1, 2 and 4 years after planting. Germination and survival varied in relation to ontogenetic traits of the species, but not the successional status, as was the case for growth and productivity. Seed germination was 60% to 88% for all species (except C. collinum ). Seedling survivorship was greater than 70% for most species, except for J. paniculata (36%). ES species have greater growth compared to LS, due to the allocation of biomass to diameter and the length of the root, respectively, from the early stage of seedling development. These two strategies of biomass allocation are positively correlated with height. We further observed two distinct phases of seedling development (0–2 and 2–4 years) for all species. Growth differences among LS species were identiﬁed 4 years after planting. Based on seedling biomass, B. spiciformis , C. collinum and P. tinctorius are the most productive potential candidates for reforestation of degraded MW. We concluded that seedling growth and productivity varied according to strategies of biomass allocation that were related to successional status (ES (Chipya group) versus LS (Miombo group)) and to age following planting.


Introduction
Miombo woodland (MW) occupies an area of about 2.7 million km 2 of central and Southern Africa. MW is characterized by the dominance of trees in the genera Brachystegia, Julbernardia and Isoberlinia [1]. The vegetation is dominated by drought-deciduous woody plants (canopy: 10% of the ground surface), which experience a dry season of three months or more [2]. This woodland is spread across Angola, Zambia, Democratic Republic of The study area is dominated by Ferralsols (FAO; Oxisols, USDA 7th Approx.), particularly haplic ferralsols [33]. On the study site, we identified a clayey ferralsol with acidic pH, low nutrient content and high Al and Fe concentrations ( Table 1). The experimental site was covered with Miombo woodland, which was currently degraded and consisted of a dense herbaceous cover and sparse trees. Hyparrhenia diplandra (Hack.) Stapf (thatching grass) was the most abundant grass species (more or less 90%). We identified other herbaceous species in isolated tufts, including: Affromum sanguineum (K. Schum.) K. Schum, Imperata cylindrica (L.) P. Beauv., Smilax anceps Willd. and Panicum maximum Jacq. The tree layer was composed of Albizia adiantifolia (Schumach.) W. Wight, Annona senegalensis Pers, C. collinum, Pericospsis angolensis (Baker) Meeuwen, Strychnos spinosa Lam., Syzygium guineense Wall. and Terminalia sp.

Species Selection and Seed Collection
Species were selected based on the concept of ecological groups in the dynamics of natural regeneration of degraded woodland [31]. Lawton [31] classified the species into four ecological groups according to their successional status: Chipya, Uapaca spp., Miombo and Mateshi. In this study, we used only early and late successional specieswhich were respectively Chipya (ES) and Miombo (LS)-based on functional traits and seed availability. According to Lawton [31], the Chipya group is composed of pioneer species that are heliophilic, resistant to fire and grass competition. This is the case of species in the genus Combretum (Combretaceae) and the genus Pterocarpus (Fabaceae, Papilionoïdeae). In contrast, species in the Miombo group (Brachystegia and Julbernardia) belong to the family Fabaceae (legumes without nodules), subfamily Caesalpinïoideae. They are shade-tolerant and require protection against fire and competition from grasses to progress from the sapling stage to the canopy ( Table 2). Seeds were collected from eight woody species (Table 2), from June to September 2010, in the Mikembo Floral-Faunal Reserve (Lubumbashi, DR Congo). Seed collection was carried out on 10 trees, selected for their desirable characteristics (e.g., healthy tree, straight stem, availability and viability of seeds), in three areas each of 4 to 10 ha. For species of Brachystegia and Julbernardia, seeds were collected under the canopy of the seed trees, disseminated by autochory (dehiscence of the pods). For C. collinum and P. tinctorius, seeds were extracted from the dried fruits and non-dehiscent pods, respectively. The seeds were sorted and mixed to form a composite sample of 1500 viable seeds per species, stored in polyethylene bags at 22 to 25 • C. A preliminary germination test was conducted at the Seed Conservation Laboratory, Faculty of Agricultural Sciences, University of Lubumbashi, DR Congo. The test determined that soaking seeds in lukewarm tap water for 24 h was the best seed pre-treatment for all species.

Installation and Assessment of Seedling Development
The development of the seedlings was evaluated on a cohort of seeds that were sown in a nursery from December 2010 to March 2014. In November 2010, a plot (15 × 15 m) was established by felling shrubs and mowing the grass cover, followed by plowing to 25 cm depth. The plot was divided into subplots of 1 m 2 , aligned in six linear blocks of eight subplots ( Figure A1A). Within a given block and between blocks, the distance separating subplots was 30 cm. Seeds were planted on 10 December 2010 (100 seeds/plot) equidistant within subplots (10 × 10 cm) for all species listed in Table 2. Each species randomly occupied a subplot of 1 m 2 , in a linear block of eight subplots, which corresponded to a randomized complete block design ( Figure A1A,B). In order to obtain a good survival rate and growth, seedlings were manually watered during the dry season (May to October) and the four years of plantation, on three alternating days per week, with 10 L of well water per plot for all species.
Seedling emergence was estimated during the rainy season from the number of seedlings per 100 seeds planted. The number of seedlings was counted at 15 day intervals from 10 December 2010 to 29 March 2011. Subsequent survival of seedlings was assessed in March, June and September 2011 (dry season), by counting the number of live seedlings present on the plot. In September 2011, thinning was conducted to leave three seedlings per plot for all species and selected according to their diameter and height, which were 30 cm apart and circumscribed by a triangle. Regrowth on thinned seedlings and competing vegetation were cut monthly (September to April: rainy season) and at two months (May to August: dry season) to eliminate competition with seedlings. In March 2012, stem trait (height and diameter) and root traits (diameter and length) were measured for all seedlings (1 and 2 years old). For root traits measurements, destructive method was used by excavating the seedlings from the soil. Indeed, stem diameter and root diameter are diameters measured at root collar level (1 and 2 years; or RCD) and at 15 cm from collar level (RD), respectively. In March 2014, only height and stem diameter at 10 cm from the ground surface were measured for the 4-year-old saplings. All allometric measurements were made on three seedlings per subplot (3 blocks) for each of eight species with a caliper for diameter and a ruler for height.
From the primary data, we determined the stem height/root length ratio (SH/RL ratio), the stem height/stem diameter ratio (SH/SD ratio) and the seedlings volume index (SVI). The SH/RL ratio was calculated by dividing the stem height on root length (adapted from [34]). The SH/SD ratio was estimated by dividing the stem height by stem diameter. Finally, the volume index of the seedlings was calculated using the following formula: SVI = (D 2 * H)/100, where D is stem diameter (cm) and H is stem height (cm) [35].

Data Analysis
Data that were collected in our study were submitted to repeated measures analysis of variance (RMANOVA, such as split-plot) using the MIXED procedure of SAS (version 9.4, SAS Institute Inc., Cary, NC, USA). In order to test the differences between variables, the species (8 levels) and block were used as the principal factor and subject, respectively; while age was considered as the subplot (within-subject, 2, 3 and 5 levels for growth traits, survival and germination, respectively). We used 3 and 6 block levels for growth traits and seedling emergence, respectively. Compound symmetry (cs) was used as the covariance structure. To validate the statistical model in our data (ANOVA assumptions), data were log 10 -transformed for seedling emergence (germination and survival), stem height and root length. Log 10 (x + 1) transformation was applied to data for stem diameter, root traits (root collar diameter and root diameter), SH/SD ratio, SH/RL ratio and volume index of seedlings. Means were compared through Tukey HSD tests (p < 0.05), including for interaction (species and age). Simple comparisons (1 degree-of-freedom contrasts) were used to examine the differences between and within groups of differing successional status (ecological groups) 4 years after planting.
To test the degree of correlation between variables, Pearson product-moment correlation (r) analysis was performed using the rcorr function from the Hmisc library [36] in the R statistical environment (R Core Team, 2019). To explain seedling mortality, we measured the correlation between seedling survival and growth variables (stem and root diameter, stem height and root length), 9 months after planting. The shoot growth strategy of the species was identified, two years after planting, based on Pearson correlations between heights, root length and stem and root diameter. For this last correlation, analysis was performed by pooling all species data for the Chipya (ES) group and, again, for the Miombo group (LS).

Germination and Survival of Seedlings
To determine whether the rate of germination and survival of seedlings differed between species, we tested the effects of species, age (day after planting) and the species by age interaction. Germination was significantly different among all species (F (7. 40) = 44.27, p < 0.0001) and was affected by age (F (4. 160) = 15.34, p < 0.0001) and the species × age interaction (F (28, 16) = 3.25, p < 0.0001). The mean germination averaged between 50% (SD = 10.21) and 83.3% (SD = 3.09) for all species, except for C. collinum (26.3%). Most species germinated 15 days after planting ( Figure 1). Yet, peak germination was reached respectively at 30 and 45 days for all LS species and one ES (C. collinum). Seedling mortality was noted from 60 days forward for all species, except P. tinctorius, the seeds of which germinated up to 75 days after planting.  A significant difference in survival was observed in age (F (2. 80) = 87.29, p < 0.0001) and among species (F (7. 40) = 15.74, p < 0.0001). Given their interaction, temporal responses depended upon species and vice versa (F (14, 80) = 27.2, p < 0.0001). Seedling survival was greater than 73.35% (SD = 21.84), and there was no significant difference between 4, 7 and 10 months after planting for six of the eight species. Yet, seedling survival of J. globiflora and J. paniculata significantly differed between 10 and 7 months and 10 and 4 months ( Figure 2). Ten months after planting (September 2011: hot, dry season), seedling survival of J. paniculata significantly decreased drastically (average = 36.8%, Figure 2).

Influence of Age on Seedling Growth
We conducted tests regarding the influence of age on seedling growth and found that growth of ES and LS species differed significantly between 2 and 4 years (Figure 3, Tables A2 and A3) after planting. Age has a more significant effect (10-20 times) on height (F (1, 16) = 586.96; p < 0.0001), diameter of the stem at collar level (F (1, 16) = 355.95; p < 0.0001) and seedling volume index (F (1, 16) = 637.24; p < 0.0001) than the species factor and interaction (species x age) (Tables A2 and A3). These important differences indicate two phases of seedling development, which are distinguished by strong growth in the root (0-2 years, Figure A2), height, RCD and seedling volume index or biomass (2-4 years, Figure 3). The first phase shows two different root growth strategies between the ES and LS species (Table 3a, (LS). N = 3 replicates (subplot) and 3 seedlings per replicate and per species Box and whisker plots summarize the 25th, 50th (median line) and 75th percentiles defining the box, together with the 10th (lower whisker) and 90th (upper whisker) percentiles, for the responses for each species. Letters above the boxplots indicate mean differences with respect to age and species, according to post-hoc Tukey HSD tests (p < 0.05). . N = 3 replicates (subplot) and 3 seedlings per replicate and per species Box and whisker plots summarize the 25th, 50th (median line) and 75th percentiles defining the box, together with the 10th (lower whisker) and 90th (upper whisker) percentiles, for the responses for each species. Letters above the boxplots indicate mean differences with respect to age and species, according to post-hoc Tukey HSD tests (p < 0.05). Table 3. Pearson correlation (r) between growth variables for early (a) and late (b) successional species, two years after planting. ES species allocate biomass as reflected in the diameter and length of the root (RD and LR; Figure A2b,c). However, the root diameter (RD) is positively correlated with height growth (Table 3a). In contrast, LS species invest their biomass more in root length, which is positively correlated with height growth (Table 3b). In addition, the SH-RL ratio ( Figure A2d) is significantly low for LS compared to ES (C. collinum and P. tinctorius) (F = 65.18, p < 0.0001).
In Figure 3, the increase in variation (CV) in height was positively correlated with the variation in stem diameter for all species (Spearman's rank correlation: r s = 0.738, n = 16, p = 0.02). A great variation was observed between 2 and 4 years for all species (CV = 100.4% and 71.1% for height and stem diameter, respectively). It was average between species for the same year (2 (CV = 66.9% and 40.2%) and 4 years (CV = 69.3% and 40.1%) for height and diameter, respectively.

Variation in Seedling Growth and Productivity
The simple contrast analysis showed that differences in growth were highly significant between early and late successional species (ES vs. LS) for all observed variables (Table 4). Height was significantly greater for ES species compared to LS. The same pattern was observed for the stem diameter, the stem/diameter ratio and the volume index of seedlings. No difference was detected between the ES species in terms of any indices of growth. Yet, we found significant differences within LS species, between the species of the genus Brachystegia and the genus Julbernadia, and within these two genera (Table 4). Brachystegia spiciformis and J. globiflora had better growth (height and volume index) compared to other species within the genus Brachystegia and Julbernardia, respectively. Seedlings of B. spiciformis had a mean height (90.8 ± 2.6 cm) and volume index (3.28 ± 0.97 cm 3 ) that were greater than B. boehmii, B. longifolia, and B. wangermeeana; this response was also the case for J. globiflora (64.9 ± 6.5 cm and 1.23 ± 0.24 cm 3 ) compared to J. paniculata (28.4 ± 2.1 cm and 0.31 ± 0.06 cm 3 ) (Table 4). Based on the volume index of the seedlings, C. collinum, P. tinctorius and B. spiciformis were more productive (Table 4).

Discussion
We made the first comprehensive comparison of growth, 4 years after planting, between ES and LS of the woodlands. ES had early growth (shoot and productivity) compared to LS, as reported for four tropical trees (2 ES and 2 LS) from India [37]. In the present work, the difference is due to different root growth strategies between ES and LS, which were identified 2 years after planting. ES species allocate more biomass in terms of increasing root diameter, while LS species allocate more biomass to the root length. These strategies of root growth are positively correlated with shoot height, in agreement with Chidumayo [19] who showed that the slow shoot growth in LS (B. spiciformis and J. paniculata, Miombo trees) is a result of biomass allocation in root growth.
Low seed germination of C. collinum (ES) can be explained by low viability of the seeds and the adaptation to regeneration by resprouting. In this study, seeds lost their viability after six months of storage, given that laboratory germination rates were high (>80%) using seeds that were one-month-old. Loss of germinability has been reported after one year of seed storage for J. globiflora [39] and J. paniculata [19]. For seeds that were planted in the field, as was the case in our trial, viability is affected by fungal infection, as observed by Chidumayo [39] for J. globiflora. Low germination rates were detected in a species within the Combretaceae (Terminalia sericea Buch. ex Dc.), i.e., 12.7% to 51% [23] and 35% to 67% [20]. This suggests strong variability in the germination capacity of seeds for species within the Combretaceae family.
Previous studies have reported that C. collinum regenerates easily through stump resprouting in Kenya [40] and South Africa [41]. This species appears to have adapted to vegetative regeneration, thereby reducing its reliance on viable seed. In fact, resprouting is a resilient strategy for most species in tropical savannas and dry forests [42]. This is the case of C. collinum, which was inventoried in degraded woodlands in Botswana [43] and Kenya [39], in the savannas of Limpopo Province (South Africa) [40], and in wasteland that is subject to fire in Zambia [7,31].
A high number of seedlings was obtained 15 days after planting. Except for P. tinctorius (ES), mortality was observed from 60 days onward for the seedlings of C. collinum (ES), B. boehmii, B. longifolia, B. spiciformis, B. wangermeana, J. globiflora and J. paniculata (LS). This mortality is probably due to insect herbivory (field observations) rather than to drought, which is a major cause of seedling mortality [19]. In our study, 45 days after planting corresponds to the middle of the rainy season, i.e., the month of February [15].
In September 2011 (10 months after planting), seedling survival was low for J. globiflora and J. paniculata (species LS) compared to other species. For seedlings of J. paniculata, 64% mortality is higher than that observed for J. paniculata (54% [19]) and J. globiflora (57-60% [39]). In our study, watering did not reduce seedling mortality of these two species, while Chidumayo [19] reported that seedling mortality (B. spiciformis and J. paniculata) was caused by drought. Adaptation to the unimodal rainfall regime means that irrigation during the dry season would not improve the survival of some species of the woodland (e.g., 20% seedling survival with Uapaca kirkiana Müll. Arg. (mahobohobo or sugar plum), despite irrigation [44]). The low survival of seedlings can be explained by a low growth (height, stem and root diameter, Figure A2) positively correlated with survival (Table A1). Seedlings did not accumulate considerable biomass in terms of height and diameter to resist desiccation in September 2011 (end of dry season and hot temperature), as reported by Chidumayo [19]. In particular, J. paniculata regenerates mainly by stump resprouting in charcoal production sites [45].

Variation in the Growth of Tree Species in Relation with Successional Status
We observed significant differences between ES and LS species, in terms of their growth (height, RCD, SD, RD, SH/SD ratio and volume index of the seedlings). The early growth for ES would be linked to adaptations to intense luminosities in degraded woodland [31] and biomass allocation strategies (stem versus root) [37]. ES species have large leaf areas and a high rate of leaf renewal [37], which permits high photosynthetic capacity and productivity. In this study, we did not measure light interception variables, a crucial research question for the future. Yet, Inman-Narahari et al [46] observed reductions in seedling growth (11%-12%) due to light effects on six Hawaiian species. As was the case in our study, Shukla and Ramakrishnan [37] reported that ES and LS species allocate biomass to the stem and root, respectively. In this study, biomass is allocated both in the stem (height) and in the root (RD and RL) for ES species (C. collinum and P. tinctorius) during the first phase of growth (0 to 2 years). Our observations reveal two different growth strategies based on a positive correlation of height with diameter and length of root for ES and LS, respectively. The ES species allocate the biomass to growth in diameter and length of the root, while the LS invest their biomass more in terms of root length which would explain a slow shoot growth during the development of the seedlings, as observed in Miombo seedlings trees (B. spiciformis and J. paniculata [19]). However, for 66 tree species of temperate forest, specific length of the root and root diameter were not clearly related to growth rate [38]. The relationship between stem growth and root traits requires more data and remains an open question for future research of the woodland tree species.
Woodland species regenerate both sexually (seeds) and vegetatively (stump resprouting and root suckers) [30,45]. Regeneration of Miombo species (LS) is slow for planted seeds compared to stump resprouting [30]. This study also reveals the low growth rate of LS species (in the genera Brachistegia and Julbernardia) compared to ES, due to the low stem height/root length ratio, in the first growth phase, for all Miombo species (LS). Chidu-mayo [19] reports a high allocation of biomass to root growth during seedling development of Miombo species, in agreement with our findings.

Substantial Change of Seedling Growth in Relation to Age
In the secondary succession of abandoned woodland in Zambia, Stromgaard [7] reported three distinct phases of growth over a 25-year period (0-1, 2-6 and 6-25 years). In our study, the four growing seasons can be split into two phases (0-2 and 2-4 years), causing a substantial change in growth, especially in terms of seedling biomass (Seedling volume index). Evaluation of the biomass at 4 years also allowed us to identify significant differences between LS species. The differences between LS species, which were characterized by slow initial growth, were not observed by Chidumayo [19] between the biomass of seedlings of B. spiciformis and J. paniculata, two years after planting. Growth as a function of age agrees with Chidumayo [30], who reported that age better predicts annual increment in diameter.
The standing biomass of certain species, which was estimated using allometric models, does not clearly show the differences between species in mature or degraded open forests. This is due to variability in tree diameter and height [26,27] differences between sites [28] and use of the woodland [6,29]. In contrast to these studies, our results clearly showed interspecific differences in the growth of seedlings of similar age being subjected to the same environmental conditions. The volume index of seedlings, which indicates their productivity [35], shows that C. collinum and P. tinctorius (ES) are more productive because the index is 3-to 40-fold higher compared to LS species (Miombo). Based on the volume index of seedlings, the three best-performing species were C. collinum and P. tinctorius (ES), and B. spiciformis (LS).
C. collinum and P. tinctorius (ES) had an average height greater than 1 m (157.6 ± 5.4 and 155.8 ± 4.5 cm), a diameter less than 3 cm (27.3 ± 2.5 and 27.8 ± 2.3 mm) and a seedling volume index of 11.87 ± 2.43 and 12.17 ± 2.09 cm 3 , respectively. In contrast, B. spiciformis has a height of less than one meter (90.8 ± 2.6 cm), a diameter of 18.8 ± 2.7 mm and a volume index of 3.28 ± 0.97 cm 3 . The mean height is in the range of that reported by Chidumayo [30] for both typical Miombo (LS) species and other canopy species, following 4 year old coppice regrowth. Yet, for the same age of the trees (4 years), our dendrometric measures are far lower compared to the growth of an exotic model species (Acacia auriculiformis A. Cunn. Ex Benth. (northern black wattle), diameter = 10.6 to 17.8 cm; height = 10.4 to 15.3 m and biomass = 53.6 to 150.9 kg), which is used for fuelwood in six provinces of the Democratic Republic of Congo [10].

Conclusions
ES species (C. collinum and P. tinctorius) have early growth (height, RCD, SD, RD and volume index of seedlings) compared to LS species (B. boehmii, B. longifolia, B. spiciformis, B. wangermeeana, J. globiflora and J. paniculata). Based on seedling productivity (volume index), B. spiciformis, C. collinum and P. tinctorius are more productive and potential candidates for reforestation of degraded woodland. Variations in growth are due to the allocation of biomass in the diameter and length of the root, from the first phase of seedling development of ES and LS, respectively. A substantial change in seedling productivity was observed for all species between 0-2 and 2-4 years, indicating two distinct phases of seedling development. Yet, growth differences between LS species (Miombo group) were observed 4 years after planting. The transferability of these results to reforestation programs, is very limited, because of site variability (different responses of tree species to rainfall and physicochemical soil properties) and multiple uses of the forest.       Letters indicate differences at the 5% level, based on post-hoc Tukey HSD tests. Lowercase and uppercase letters, respectively, represent comparisons between species (same year) and between years (for one species).