2.1 Study area
The study was conducted in homestead forests of Maheshkhali Island under Cox's Bazar District, a coastal area of Bangladesh (Fig. 1), emerged as most vulnerable to climate impacts as documented in the recently published report of the World Bank (WB) [11]. Maheshkhali island is the only hilly island with complex geological system on the eastern cliff coast of Bangladesh, bounded by coastal plain characterizing inimitable geologic, tectonics and geomorphologic particularities [58], located between 21°28' and 21°46' N latitude and 91°51' and 91°59' E longitude [59]. It is bordered on the north by Chakaria Upazila (sub-district), on the south Cox’s Bazar Sadar Upazila and the Bay of Bengal, on the east Chakaria and Cox's Bazar Sadar Upazilas, and on the west Kutubdia Upazila and the Bay of Bengal (Fig. 1). It occupies an area of 362.18 km2, with a total of 33287 households [60]. The island has a moist tropical climate with a long wet season (April–October) and a relatively short dry season (November–March). The mean annual precipitation, temperature, and relative humidity are 3,627 mm, 25.7 oC, and 70 to 90 %, respectively [61]. This region is prone to cyclonic storms, tidal surges, and flood due to proximity to the Bay of Bengal, a source of cyclones, usually occurring during April–May and October–November. The island was separated from the mainland by severe cyclone and tidal bore of 1559 [62].
The island has four subdivisions including active, young, and old coastal plain, and hilly areas with piedmont plain. Geological deposition of sedimentation forms landmasses [58]. Maheshkhali with an accretion rate of 1.2 sq. km. per year since 1972 formed huge landmasses in the southwest coastal plain (e.g., Bara Maheshkhali) and western coastal plain (e.g., Gorakghata) [63], contributing to the land use and land cover changes. The major land uses include salt cultivation, agricultural land, hill forests and coastal forests, which have been changed markedly since 1972. Expansion of salt fields caused a decline in the agricultural land at an average rate of 14.5 hectares per year, and extensive and illegal hill cutting for settlement, betel leaf cultivation, and unpanned development caused reduced hill forests at an average rate of 90 hectares per year [59]. The coastal forests are in an increasing trend due to the planting mangrove trees by the BFD and local people along the eastern and western coast to protect lives and properties from the adverse impacts of cyclonic events and floods. However, shrimp cultivation, another type of land use threatened the coastal forests, to some extent, especially old coastal zone [59].
BFD manages its hill and coastal forests consisting of mangrove plantations of 534.54 ha and 8129.9 ha and non-mangrove plantations of 2667.3 sq. km and 232.66 sq. km, under two range[1], Maheshkhali and Gorokghata, respectively [64]. Forest covers several hills of up to 23 m and the low-laying valleys. Soils of the forest vary from clay to sandy loam and to some extent yellowish red sandy clay, with low pH. Hills are mainly composed of sands and various stones, including limestone, siltstone, and mudstone [65]. Peoples` dependence on hill forests were significant for collecting fuelwood, house and boat making materials, traditional medicine, and non-timber forest products such as bamboo honey, fodder etc. However, from the time immemorial, this overexploitation of the resource declined in biodiversity in the area [66]. The detrimental activities included deforestation, hill cutting for converting forest land to settlement and agriculture, and coastal forests to shrimp culture and salt fields [66]. This situation made the individuals for taking care and managing homestead forests very carefully for their protection from coastal storms, surges, floods. BFD has no management and administrative right over homestead forests as they are managed by owners themselves.
Fig. 1 Map of a) Bangladesh and b) Maheshkhali Island showing three categorized study sites with c) sampling points of homestead forests.
2.2 Reconnaissance survey
Before starting the data collection, three initial field visits for reconnaissance survey were made to get an overview of the study area in February 2019. This included observation of general conditions including geographical location, physiography, hill and coastal forests, and homestead forests of the area. We divided three Unions[2] Chhato Maheshkhali (recently converted as municipality), Gorokghata, and Bara Maheshkhali as the hillside, beachside, and inland, respectively, under Maheshkhali sadar Upazila according to the geographic location (Fig. 1). Then the researchers visited to the respective administrative (Union Parishad) offices of these three Unions to collect relevant data of villages, number of households and homestead forests of each village. Additionally, we interviewed the Chairman of these three Unions as key informants to gather knowledge about the study sites and as well as inform them about the purpose of the study.
2.3 Sample selection and woody vegetation measurement in homestead forests
Settlement is an important type of land use, and the population density was relatively higher in the hillside (Chhato Maheshkhali), beachside (Gorokghata), and inland (Bara Maheshkhali) compared to the northwestern part of the island [67]. From the key informant (KI) interview with the Chairman of the Union in February 2019 it was known that some of the settlements started in the foot of the hills under the hillside by human intervention in modifying slope of the hills. Since after the loss of lives in 1991 cyclone, people started migrating from Gorakghata to other places [63]. Settlement started earlier in the inland site and population was relatively higher [67]. Settlement was assumed to be associated with the homestead forests in the three sites. The homestead forests of these three sites were also assumed to represent the same kind of ecotype. We hypothesized that C stocks of the homestead forests differ from each other among the three sites.
The sampling procedure followed from Upazila to Union, Union to village, village to homestead forests of the households. From the lists of the number of households provided by the office, with a sampling intensity of 5% as accepted by the United nations [68], a total of 239 homestead forests were determined. Then, based on the total number of households in each of the three sites, 67, 69, and 103 homestead forests from hillside, beachside, and inland, were randomly allocated for the study in Maheshkhali sadar Upazila in 2019 (from February to April). The mean area of the studied homestead forests in these three sites were 0.02, 0.01, and 0.02 ha per household.
Each of the homestead forests was divided into quadrats (5m × 5m) based on the area and the direction from the dwelling. The surveyed data were recorded which included all woody plants identification, with measurement of height (m), diameter at breast height (DBH, cm) and the area of the homestead forests. The owners of the homestead forests helped in identification with local name, and in few cases, herbarium was prepared to ensure the identification with scientific names. The height measurement was made by rangefinder and DBH by diameter tape. The coordinates of each point of sample collections was recorded by using GPS. Herbs and shrubs were not considered as 98% of total forest biomass consists of tree biomass; they may be ignored in estimation of carbon [25]. Homestead forests are well managed and therefore, are usually free from herbs and lianas.
2.3 Soil and litterfall sampling for estimating C stock
A sampling of the litterfall was made in 4 points wherever available for each of the three different sites in 2019 (from February to April), thus making a total of 12 (3 × 4 = 12) samples. All litterfalls at each point of an area of 1 m2 (1 m × 1 m) was collected using a metallic frame. A pit of 30 cm depth, under the litterfall layer sampling point, was dug by using a soil auger and mineral soil samples were collected at 10, 20, and 30 cm depths. This procedure was followed for four samples consisting of 12 (4 × 3 depths = 12) subsamples for each of the three different sites, thus making a total of (12 × 3 = 36) subsamples. Accordingly, following the same procedure, 36 unaltered soil subsamples were collected using a core (10 cm high and 6.5 cm wide) to measure bulk density (BD) at the same three depths in each point, following Blake (1965) [70].
2.4 Data analyses
2.4.1. Estimation of tree (above- and below-ground) biomass, and density and basal area of stands
Above-ground biomass (AGB) was estimated by converting tree data into biomass using allometric Equation (1), (2), (3), and (4) for tropical trees, Cocos nucifera, Areca catechu, and Phoenix dactylifera, respectively (Table 1). Below-ground tree biomass (BGB) was estimated as 15% of AGB [74]. Tree total biomass (TB) was the summing up of AGB and BGB. Finally, total C stock (Mg ha-1) was estimated as C content is assumed to be 50% of dry TB [75]. To estimate AGB wood density (g cm-3), a required variable, which was collected from Bangladesh Forest Research Institute (BFRI) [76]. For those not found in BFRI publications we used global wood density database [77, 78]. Additionally, species level C was also estimated for most frequent tree species and expressed in kg C per individual across three sites. Stand density (D) (individual ha-1) and basal area BA (m2 ha-1) were estimated (Equations 5 and 6). Mean values of tree biomass, density, and BA were compared among three different homestead forest sites.
2.4.2. Laboratory analysis and estimation of carbon of litterfall and mineral soil and bulk density (BD)
To estimate soil organic carbon (SOC), washed silica crucibles were dried in an oven at 105 °C for half an hour and cooled in desiccators, and then weight was taken. Dried soils were ground by pestle and then exactly 5 g of grind soils were kept in silica crucibles and weighted by an electric balance. The crucibles with soil were then transferred to an electric muffle furnace for igniting at 850 °C for one and half an hour. Then crucibles with soils were cooled in the desiccator and reweighted to determine the percent loss of ignition LOI (%), from which, SOC (%) was calculated (Equations 7 and 8). C stocks in mineral soil at three depths were calculated using BD (g cm-3) (Equation 9) and expressed in Mg ha-1 for three different sites (Table 1). We calculated soil BD as the quotient between the dry mass of the fine fraction in the core segment and volume of the cylinder [82].
To estimate the biomass of litterfall, after taking the fresh mass of the original samples collected from each point of litter collection, adequate subsamples from the weighted original sample were made and labelled. In each plot, the number of original samples was four and subsamples three to five, depending on the wet masses of the original samples. The wet masses of all the subsamples were measured and recorded. Subsamples were oven-dried at 65 oC until reaching a constant mass and dried masses were recorded. Then, the dry mass of the original sample from the wet to dry ratio of the subsamples was estimated (Table 1; Equations 10 and 11). The C concentration was considered to be 44.36% of the dry mass of litter [83]. The process was repeated for all 12 original samples collected from homestead forests across three different sites. C stocks in litterfall were calculated and expressed in kg C ha-1 for three different sites. These C concentrations and stocks of litterfall and soils were compared among three different homestead forest sites.
2.4.3 Estimation of tree species richness, diversity and relative frequency and relative density
Tree species richness (Margalef index) and diversity (Shannon-Weiner Index, H) were estimated according to Equations 12 and 13 (Table 1). The greater value of indices of diversity indicates greater species richness and diversity in an area. In addition, the relative frequency of occurrence (RF %) and relative density (RD %) for species were estimated (Equations 14-16). Mean values of tree height (m), DBH (cm), all indices, RF, and RD were compared among three different sites.
2.4.4 Statistical analyses and modelling work
One-way analysis of variance (ANOVA) was used to determine whether there are any statistically significant differences (p ≤ 0.05) between the three homestead forest sites in tree biomass (Mg C ha-1), height (m), DBH (cm), density (individual ha-1), basal area, BA (m2 ha-1), Margalef richness index and Shannon-Wiener diversity index. Tukey´s post hoc tests were performed to determine which site significantly differed from the other sites. Moreover, two-way analysis of variance (ANOVA) was performed to determine whether there are any statistically significant differences (p ≤ 0.05) of soil C stock (Mg C ha-1) against three sites and three soil depths.
Relationship between tree biomass (Mg C ha-1) and a) height (m), b) DBH (cm), c) density (individual ha-1), d) basal area, BA (m2 ha-1), e) Margalef richness index, and f) Shannon-Wiener diversity index were modelled by using linear regression analysis. In addition, multiple regression analysis was used to model the effect of all variables (a-f) to tree biomass.