Andrew S. Mariki & Abigail R. Wills

Transcription

1 Mpingo Conservation & Development Initiative Mpingo Conservation & Development Initiative Environmental Factors Affecting Timber Quality of African Blackwood (Dalbergia melanoxylon) November 2014 Andrew S. Mariki & Abigail R. Wills

2 Executive Summary African Blackwood (Dalbergia melanoxylon), or mpingo as it is called in Tanzania, is one of the most valuable trees in the world. However, high wastage rates during processing limit the revenue that can be generated by rural communities that trade the timber. Between September 2013 and September 2014, the Mpingo Conservation & Development Initiative (MCDI), with funding from the International Tree Foundation (ITF), investigated the environmental factors specifically site, soil type and vegetation affecting blackwood timber quality and yield in three community forests in south-eastern Tanzania. This report summarises the findings of the research. Vegetation type had the strongest influence on blackwood timber yield and quality in the community forests surveyed; the branching height, and thus bole length and harvestable volume, of timber of blackwood trees in open woodland was significantly higher than those in less-dense wooded grassland. However, these taller trees in open woodland were also subject to more severe fluting. These findings suggest that stratifying community timber inventories and/or harvesting operations by vegetation type could: (a) provide a more accurate representation of timber availability, and (b) allow better identification of suitably long, straight lengths of timber for use in manufacturing processes. However, this will not come without additional costs; stratification will require effort to determine the spatial distribution of vegetation types in community forests, will add complexities to the inventory method, and could result in inventorying and/or felling trees with more defects, such as fluting. Further research is recommended to assess how other aspects of timber quality in blackwood, e.g., colour, heart rot and other defects, are affected by environmental variants. Such research will probably involve measurements of pre-felled logs, which were beyond the scope and budget of this research. Citation: Mariki, A. S. & Wills, A. R. (2014) Environmental Factors Affecting Timber Quality of African Blackwood (Dalbergia Melanoxylon). Mpingo Conservation & Development Initiative, Tanzania. Work funded by the International Tree Foundation Front Page Image: Evaluating harvestable African Blackwood by Anne-Marie Gregory enquiries@mpingoconservation.org Page 1 of 20

3 Background African Blackwood (Dalbergia melanoxylon), or mpingo as it is known in Tanzania, is one of the most valuable timbers in the world. It is renowned as one of the finest woods to make musical instruments with. Between 7,500 and 20,000 trees are felled to make musical instruments each year, mostly from Tanzania and Mozambique. Sustainably harvesting blackwood timber is a significant potential revenue earner for rural communities in areas where stocks are economically viable. The Mpingo Conservation & Development Initiative (MCDI) is a Tanzanian NGO that supports communities to realise these benefits. They do this by building local capacity to sustainably manage timber stocks, and by linking communities to markets through which they generate ethical and long-lasting income by trading responsibly harvested blackwood, and other valuable hardwood species. Unfortunately, the revenue that can be generated by communities is limited by high wastage rates of blackwood during the manufacture of musical instruments. This is because the sections of wood, or billets, used must be completely free of defects, otherwise they will split during machining. While some flaws are easy to see in the forest, a fault as small as a pin-hole will lead to a log being rejected at the sawmill. Unfortunately the shape of mpingo trees mean that it is often hard to find long, straight sections of wood to form billets. Thus, wastage rates at sawmills are over 90% and even more timber in the form of the smaller branches is abandoned in the forest. With support from the International Tree Foundation (ITF), MCDI set out to tackle this issue by performing research into the environmental factors affecting timber quality and yield of blackwood trees. This will enable better sustainable management and harvesting of the resource, thus maximizing the benefit flow to rural communities and increasing local incentives to manage forests sustainably. Aims & Objectives Overall Aim To improve understanding of environmental factors affecting timber quality in African Blackwood (Dalbergia melanoxylon; and, by extension, other miombo tree species) to facilitate better sustainable management and harvesting of the resource. Specific Objectives 1. Determine how the quality and yield of timber in blackwood trees varies with site, soil and vegetation type. 2. Develop simple, participatory methods for measuring potential determinants of timber quality and yield so that rural communities can use them to better assess their own stocks of blackwood. Page 2 of 20

4 Methodology Study Site The research was conducted in three communal forest reserves, or Village Land Forest Reserves, VLFRs, as they are legally designated in Tanzania: Kikole, Kisangi and Nainokwe (Figure 1; Table 1). All three of the VLFRs are dominated by miombo woodland and are under sustainable timber harvesting programmes by communities, supported by MCDI. Prior to data collection, the purpose and design of the research was introduced to the Village Council and Village Natural Resources Committee (VNRC) 1 in each village, and approval obtained to perform the research. Members of the Village Council, the VNRC, and wider community in each village were involved during field data collection. Figure 1: Map of the study site. 1 The Village Natural Resources Committee is a sub-committee of the Village Government with the responsibility for overseeing management of community-owned forests, known as Village Land Forest Reserves. Page 3 of 20

5 Table 1: Details of the three Village Land Forest Reserves where research was conducted. Village Land Forest Reserve Kikole Kisangi Nainokwe Year established Year of first timber harvest Area (ha) 454 1,966 8,047 Local population 1, ,980 Timber density (trees/ha) African Blackwood density (trees/ha) Participatory Assessment of Factors Affecting Timber Quality Site, soil and vegetation type were selected as environmental variables that: (a) are most likely to impact blackwood timber quality and yield, and (b) can be assessed using simple cost-effective participatory methods involving rural communities. The abundance of blackwood trees as well as the estimated harvestable volume of timber per tree was used to quantify timber yield, and fluting was used as a proxy for timber quality. Fluting was chosen because it is easy to identify in the forest without requiring the assessment of pre-felled logs, which was beyond the scope and budget of this project. Data was collected in two stages. Firstly, the distribution of soil types in each of the three VLFRs was mapped. Secondly, surveys were conducted to assess the abundance, quality and timber yield of blackwood trees in each soil type, and to report on other environmental variables, including vegetation type. Soil Mapping Prior to electronic soil mapping inside the VLFRs, focus group discussions were held in each village with 16 members of each VNRC, 10 Village Council members, and four additional community representatives with basic knowledge on local soils. During these focus groups, participants were asked to list all types of soil found in the village, using vernacular names together with a description, and then to discuss where the soils are commonly found, and develop a hand-drawn map depicting the distribution of each soil type within their VLFR (Appendix 1). After consulting the soil descriptions provided by local people in all three villages, the soil types were grouped into three mutually exclusive categories based on these local descriptions (Table 2), which were later used to inform data analysis. Table 2: Soil types found in the three Village Land Forest Reserves where research was conducted. Soil type Black cotton Loam Sandy loam/sand Definition Black in colour with small grains and very compact; slippery when wet and retains water for a long period of time; forms cracks during the dry season. Dark grey in colour with small grains and semi compacted; contains moisture and very fertile for agriculture. Light grey in colour with large grains and not compacted; dries very early in the dry season; not smooth in texture. Page 4 of 20

6 Ten participants from the focus group were selected to participate in GPS soil mapping inside the VLFR in each village. The team, led by MCDI s Monitoring Officer, performed a series of walkover surveys to identify the different soil types present following the typology which had been developed previously during the focus group discussion along with their distribution. Waypoints were recorded at the interface between each soil type, and were later used to map soil distribution electronically using GIS software (Appendix 2). Ecological Measurements of African Blackwood Trees Blackwood trees were sampled using a random walk design in each VLFR, stratified by soil type. A random start point was located in each soil type; these were identified by first referring to the location of the boundaries of each soil type using the electronic maps produced during soil mapping, and then using a random number generator to identify coordinates that fell within the limits of each soil type. Once the team arrived at a start point, they embarked on a random walk, as follows. 1. A random compass bearing was generated using a random number generator and the team walked a transect in that direction until they found an eligible blackwood tree. 2. Any blackwood tree that was larger than the legal minimum harvestable diameter at breast height (DBH) of 240 mm encountered within five metres either side of the central transect line was considered eligible. 3. Each consecutive tree was determined by generating a new random bearing and walking in that direction until the next eligible tree was encountered. 4. If no eligible trees were encountered within 100 metres, a new random bearing was generated and the sampling continued in that direction. A minimum of 60 blackwood trees were sampled from no less than five start points in each VLFR, and a minimum of 20 trees were sampled in each soil type; the number of start points and trees sampled increased with the size of the forest and soil type. For each tree sampled, the following measurements were recorded: UTM GPS location Elevation Soil type Surrounding vegetation type Distance to nearest tree with DBH > 30cm DBH (mm) Branching height (cm) Number and length of harvestable bole and branches (cm) Depth of fluting (mm) Statistical analysis An estimate of the harvestable volume, HV, (m 3 ) of wood from each tree was generated using the following equation. HL refers to the length (cm) of each harvestable portion of timber, including the bole and branches. DBH is given in mm. For each tree, fluting severity, F, was calculated as a percentage of the DBH, using the following equation. FD refers to the depth (mm) of fluting. Page 5 of 20

7 Pearson s Chi-squared tests were used to determine differences in the number of trees recorded between vegetation and soil types. Two-way ANOVA tests were used to analyse differences and interactions between the affect that site, soil type, and vegetation type had on blackwood trees in respect to: (1) distance from the nearest large tree; (2) branching height; (3) harvestable timber volume; and, (4) fluting severity. Where differences between groups were significant, but the interaction was not, a post-hoc test the Tukey test was used to determine where the differences were. If the interaction and/or differences between groups were significant, Simple Main Effects analyses with Bonferroni corrections were used to determine where the differences were. A Simple Linear Regression was used to examine whether the branching height was a function of distance to the nearest large tree, and also to determine whether harvestable timber volume and/or fluting severity were functions of branching height. Results Site Characteristics Three different broad soil types were identified in the project area: black cotton, loam, and sandy loam/sand. Loam was the most widespread type of soil, covering almost 40% (39.93%; 5,112 hectares) of the project area, followed by black cotton (35.40%; 4,532 hectares) and sandy loam/sand (24.66%; 3,157 hectares). The local dominance of each soil type varied between VLFRs (Figure 2). Figure 2: Distribution of black cotton (colour), loam (colour) and sandy loam (colour) soil types in: (a) Kikole, (b) Kisangi, and (c) Nainokwe Village Land Forest Reserves. A total of 252 blackwood trees were sampled. The majority of blackwood trees were recorded in loam (41.12%, n = 104) or black cotton (39.53%, n = 100) soil, while approximately one fifth (19.37%, n = 49) were found on sandy loam/sand. The number of trees recorded in each soil type different significantly between VLFRs (Pearson s Chi-squared test, X 2 : , p < 0.001, n = 264; Figure 3). Page 6 of 20

8 Figure 3: Number of African Blackwood (Dalbergia melanoxylon) trees per Village Land Forest Reserve that were encountered in each soil type. Blackwood trees were recorded in five vegetation types (Figure 4), although indicators of timber quality were analysed between open woodland and wooded grassland only. This is because a sufficient number of trees (n > 50) were recorded in each of these vegetation types, and because they were encountered in more than one VLFR. The proportion of trees recorded in open woodland and wooded grassland differed significantly between VLFRs (Pearson s Chi-squared test, X 2 : , p < 0.001, n = 233). Figure 4: Number of African Blackwood (Dalbergia melanoxylon) trees per Village Land Forest Reserve that were encountered in each vegetation type. Page 7 of 20

9 All three soil types supported open woodland vegetation. By contrast, blackwood trees in wooded grassland were only found on black cotton or loam soils, and never on sandy loam/sand (Figure 5). These differences were significant (Pearson s Chi-squared test, X 2 : , p < 0.001, n = 233). Figure 5: Number of African Blackwood (Dalbergia melanoxylon) trees in each vegetation type that were recorded in black cotton soil, loam soil, or sandy loam/sand. A two-way ANOVA was conducted to examine how the distance between blackwood trees and the nearest large (DBH > 30cm) tree was influenced by VLFR and vegetation type. There was a statistically significant interaction between the effects of VLFR and vegetation type on the distance between trees (F 1, 212 = 7.881, p = 0.005). Mpingo trees in wooded grassland in Nainokwe were located significantly further away from neighbouring large trees than those in Kikole (Simple Main Effects: p = 0.006; trees were not recorded in wooded grassland in Kisangi), while the distance between trees in open woodland did not vary between villages (Figure 6). Figure 6: Site-specific mean distance of African Blackwood (Dalbergia melanoxylon) trees from the nearest large (DBH > 30cm) tree in open woodland (dark green) and wooded grassland (light green). 2 2 In all box and whisker plots, the ends of the whiskers are set at 1.5 times the inter-quartile range above the third quartile and 1.5 times the inter-quartile range below the first quartile. If the minimum or maximum values are outside this range, then they are considered outliers. The minimum and maximum outliers are displayed. Page 8 of 20

10 Factors Affecting Timber Quality and Yield Soil A two-way ANOVA was conducted to examine how tree branching height was affected by soil type between VLFRs, and a statistically significant interaction was found (F 2, 263 = 6.757, p = 0.001). Blackwood trees in black cotton soils in Kisangi branched significantly higher than those in Nainokwe (Simple Main Effects: p < 0.001). Trees in sandy loam/sand tended to branch lower than those in Kisangi and Kikole, although not significantly so (p = 0.089). The branching height of blackwood in loam soils did not vary between VLFRs (Figure 7). Figure 7: The site-specific mean branching height of African Blackwood (Dalbergia melanoxylon) trees in black cotton soil (dark green), loam soil (green), and sandy loam/sand (light green). A two-way ANOVA was conducted to examine how the harvestable volume of blackwood trees was affected by soil type between VLFRs, and a statistically significant interaction was found (F 2, 244 = 3.276, p = 0.039). Trees in black cotton soils in Kisangi were likely to have significantly larger harvestable volumes than those in Nainokwe (Simple Main Effects: p < 0.023). The harvestable volume of trees in loam and sandy loam-sand soils did not vary between villages (Figure 8). Figure 8: The site-specific mean harvestable volume (m 3 ) of African Blackwood (Dalbergia melanoxylon) trees in black cotton soil (dark green), loam soil (green), and sandy loam/sand (light green). Page 9 of 20

11 A two-way ANOVA was conducted to examine how the degree of fluting in blackwood trees was influenced by soil type between VLFRs, and a statistically significant interaction was found (F 2, 263 = 4.924, p = 0.008). The severity of fluting in trees in loam soils in Nainokwe was greater than those in Kikole (Simple Main Effects: p = 0.058), and significantly more severe than those in Kisangi (p = 0.008). Conversely, fluting of trees in black cotton soil and sandy loam/sand did not vary between VLFRs (Figure 9). Figure 9: The site-specific mean fluting severity (% of DBH) of African Blackwood (Dalbergia melanoxylon) trees in black cotton soil (dark green), loam soil (green), and sandy loam/sand (light green). Vegetation A two-way ANOVA was conducted to examine how the branching height of blackwood trees was influenced by vegetation between VLFRs (Figure 10). Trees branched significantly higher in open woodland than in wooded grassland (F 1, 232 = 7.132, p = 0.008). Branching height also differed significantly between villages (F 2, 232 = 5.436, p = 0.005); blackwood trees in Nainokwe branched significantly lower than those in both Kikole (Tukey test: p = 0.001) and Kisangi (p = 0.001). There was no statistically significant interaction between the influence of vegetation type and VLFR on branching height (F 1,232 < 0.001, p = 0.986). Figure 10: The site-specific mean branching height (cm) of African Blackwood (Dalbergia melanoxylon) trees in open woodland (dark green) and wooded grassland (light green). Page 10 of 20

12 A two-way ANOVA was conducted to examine how the harvestable volume of blackwood trees was influenced by vegetation type between VLFRs (Figure 11). Trees had significantly higher harvestable volumes in open woodland than in wooded grassland (F 1, 219 = 9.258, p = 0.003), and this was consistent in all VLFRs (F 2, 219 = 0.029, p = 0.971). There was no statistically significant interaction between the influence of vegetation type and VLFR on harvestable volume (F 1,219 = 0.119, p = 0.730). Figure 11: The site-specific mean harvestable volume (m 3 ) of African Blackwood (Dalbergia melanoxylon) trees in open woodland (dark green) and wooded grassland (light green). A two-way ANOVA was conducted to examine how the degree of fluting in blackwood trees was influenced by vegetation type between VLFRs (Figure 12). The severity of fluting in trees was significantly higher in open woodland than in wooded grassland (F 1, 232 = , p < 0.001). Further, fluting in Nainokwe tended to be more severe than in Kikole (Tukey test: p = 0.099), and was significantly more severe than in Kisangi (p < 0.001). There was no statistically significant interaction between the effects of vegetation type and village on fluting (F 1,232 = 0.027, p = 0.870). Figure 12: The site-specific mean fluting severity (% of total DBH) of African Blackwood (Dalbergia melanoxylon) trees in open woodland (dark green) and wooded grassland (light green). Page 11 of 20

13 The distance of blackwood trees from the nearest neighbouring tree had a significant effect on the branching height (Simple Linear Regression, F 1,242 = 4.452, p = 0.036, R 2 = 0.018, n = 243; Figure 13a) which, in turn, was related to the trees harvestable volume (Simple Linear Regression, F 1,249 = , p < 0.001, R 2 = 0.142, n = 250; Figure 13b), as well as fluting severity (Simple Linear Regression, F 1,262 = , p = 0.001, R 2 = 0.041, n = 263; Figure 13c). Figure 13: The relationship between African Blackwood (Dalbergia melanoxylon) tree branching height (cm) and (a) distance (m) to the nearest large (DBH > 300mm) tree; (b) harvestable volume (m 3 ) of timber; and, (c) fluting severity (% of total DBH). Page 12 of 20

14 Discussion & Conclusions In the community forests surveyed, blackwood trees tended to be more abundant in loam and black cotton soils when compared with sandy loam/sand. However, this varied between sites. For example, in Kisangi, more trees were recorded in loam, despite sandy loam/sand covering a larger area. By contrast, more blackwood trees were encountered in sandy loam/sand in Kikole which covered a larger area and where dense open woodland occurred than in loam, which supported less dense wooded grassland at the site. Therefore, it is likely that the occurrence of blackwood trees in these forests is affected by a combination of interacting factors, including soil and vegetation type which influence substrate fertility, drainage, stand density and inter-specific competition (Koch, et al., 2004) among other site characteristics. The surrounding vegetation in which blackwood trees were located had a significant effect on the tree branching height, and thus timber yield (a partial function of bole length and thus branching height). In all three VLFRs, trees in open woodland tended to branch higher up, and thus had longer boles, than those in wooded grassland. This is probably a result of differences in stand density, as trees in open woodland were typically less distant from neighbouring large timber trees, which had a significant positive effect on branching height and timber yield. This is a typical physiological response trees in dense forests are under more competition for light, which places a premium on height growth meaning that trees grow tall (Koch, et al., 2004) and thus may also explain why branching height, bole length and timber yield, of blackwood trees also differed between VLFRs and soil types. Blackwood trees in Nainokwe, the VLFR in which wooded grassland was most abundant, had significantly lower branching heights than those in both Kikole and Kisangi. Furthermore, trees in black cotton soils in Kisangi, which only supported dense open woodland at the site, had larger harvestable volumes of timber than those in Nainokwe, which, although supporting open woodland, also supported the more sparse wooded grassland vegetation. Vegetation also affected fluting in blackwood trees. Fluting was more severe in open woodland than in wooded grassland. Given that fluting severity has been positively correlated with tree growth rate and stand age in Western Hemlock trees (Singleton., et al, 2003), this could be because open woodland is a denser habitat in which trees are under more competition for light, and therefore grow taller and faster than those in wooded grassland. Fluting was also positively related to branching height, which would be expected as flutes generally start where branches grow out from the stem (Julin and Farr, 1989). Trees, in general, fluted less in loam soil than black cotton soil and sandy loam/sand, although it was also more variable. This is probably due to differences between VLFRs, as trees in loam soils in Nainokwe were likely to have more fluting than those in Kikole, and significantly more than those in Kisangi, while the fluting of trees in black cotton and sandy loam/sand soils did not vary between VLFRs. The reason for this is unclear, although it could be due to differences in soil fertility; trees situated on fertile, well-drained soils such as loam grow rapidly, and this may promote fluting (Julin and Farr, 1989). It could also be a result of stand density as fluting tended to be more severe in Nainokwe, which had the highest density of timber trees, than in the other two VLFRs. Previous disturbance, such as clear cutting or wind throws, and mechanical stress can also induce fluting (Julin and Farr, 1989). Conclusion & Recommendations Although tree growth can be influenced by factors such as local topography, resource availability, and previous disturbance (Julin and Farr, 1989; Koch, et al., 2004), vegetation density and therefore competition for light was the most important factor influencing blackwood timber yield in the three forests surveyed. These findings provide a basis for the communities to stratify timber inventories and/or harvesting operations by vegetation type, to provide a more accurate representation of timber availability. They will also facilitate better identification of suitably long, straight lengths of timber for manufacture, which could reduce the number of branches which are abandoned in the forest. However, this will not come without additional costs; stratification will require effort to determine the spatial distribution of vegetation types in community forests, and will add complexities to the inventory Page 13 of 20

15 method, which is already the most technical step involved in developing a VLFR forest management plan in Tanzania (MCDI, 2010). Unfortunately, stratification by vegetation type could also result in inventorying and/or felling trees with more defects such as fluting, which was more severe in denser habitats, potentially as a result of fast growth and thus it will not solve the issue of wastage following rejection at the sawmill. Thus, considering the associated risks, costs and benefits, it is unclear whether stratifying timber inventories and harvesting operations by vegetation type would be a worthwhile investment. The payoffs of this approach are also likely to vary between VLFRs, depending on their size and the relative dominance of the different vegetation types present. The next step therefore may be to implement stratification in one community forest as a pilot case, to determine whether the returns are sufficient to invest in implementing the method more broadly. This will need to be a voluntary decision made on behalf of the community managing the forest, having been fully informed of the results of this research, including the potential costs and benefits of the transition to stratification. Further research is recommended to assess how other aspects of timber quality in blackwood, e.g., colour, heart rot and other defects, are affected by environmental variants, including vegetation type. Such research will probably involve measurements of pre-felled logs, which were beyond the scope and budget of this study. Page 14 of 20

16 Literature Cited Julin, K. J. and Farr, W. A. (1989) Stem Fluting of Western Hemlock in Southeast Alaska. U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, Oregon. Koch, G. W., Sillett, S. C., Jennings, G. M and Davis, S. D. (2004) The limits to tree height. Nature 428, Mpingo Conservation & Development Initiative (2010) ED02 Group Guidelines on Forest Assessment and Sustainable Harvesting. Version 1.4. Available Online: Singleton, R., Dean S. DeBell, D. S., Marshall, D. D. and Gartner, B. L. (2003) Eccentricity and Fluting in Young Growth Western Hemlock in Oregon. WJAF, 18(4): Page 15 of 20

17 Appendices Appendix 1: Hand-drawn map depicting the distribution of each soil type within VLFRs. Hand drawn map showing distribution of soil type within Kikole VLFR Page 16 of 20

18 Hand drawn map showing distribution of soil type within Kisangi VLFR Page 17 of 20

19 Hand drawn map showing distribution of soil type within Nainokwe VLFR Page 18 of 20

20 Appendix 2: Maps of the distribution and size of soil types in each VLFR. Map showing distribution and size of the soil types of Kikole VLFR (454 hectares). Map showing distribution and size of the soil types of Kisangi VLFR (1,966 hectares). Page 19 of 20

21 Map showing distribution and size of the soil types of Nainokwe VLFR (8,047 hectares). Page 20 of 20