| Description: | Carbon sampling
We measured nonsoil carbon stocks across three main habitat types: farmland, secondary forest (accounting for variation in age), and old-growth forest. We randomly selected 36 400 × 400 m sampling squares across the three habitats in each of the three landscapes (15, 12, and 9 squares in Kiphire, Phek, and Kohima, respectively). The number of squares in each district varied depending on the availability of fallow sites and adjacent old-growth forest sites (distance between fallow sites to the nearest primary forest across the three landscapes = 2,410.5 ± 1,748 m). Sampling squares were placed at least 300 m apart between different habitats and 400 m apart within the same habitat. Within each sampling square, we located three 10 × 30 m sampling plots (n = 108; 3.24 ha sampled in total) that were at least 200 m apart. To ensure unbiased selection of plots, we walked 100 m perpendicular from the boundary into the focal habitat type. The resultant end point was used as the first corner of the 10 × 30 m carbon-sampling plot and the second point was located 30 m to the left (i.e., roughly 30 m parallel to the habitat edge). The other two axes of the rectangular plot were parallel to these two randomly selected points. We followed this methodology consistently for all plots. Within each sampling plot, we first measured aboveground living biomass (trees and lianas) and dead biomass (deadwood and leaf litter) using a composite plot design and converted these biomass estimates to carbon stocks.
Estimating live biomass
We determined live biomass by measuring the diameter at breast height (DBH) and wood specific gravity of trees. We measured DBH at 1.3 m from ground level in each 10 × 30 m plot for all trees larger than 5 cm DBH. We measured trees with 1–5 cm DBH in three subplots each of 2 × 2 m in size at 5-, 15-, and 25-m distance from the start of the plot, along the plot midline. To calculate wood specific gravity, we extracted tree cores from all trees larger than 5 cm DBH at 1.3 m with an increment borer (two threads, 5.15 mm diameter, 400 mm bit length; Haglöf, Långsele, Sweden). The full core was placed in water for 30 min to fully hydrate it and the fresh volume (i.e., green volume) was then measured using the water-displacement method (Chave 2005). Cores were then oven dried at 101°–105°C (Williamson and Wiemann 2010) for 24 h and weighed. Finally, we calculated wood specific gravity (g/cm3) from the dry mass (g) to green volume (cm3) ratio (Chave 2005). The extraction of cores was not possible for small trees (1–5 cm DBH), so for these individuals, we used the mean wood specific gravity calculated from large trees within the focal 10 × 30 m plot.
We calculated tree biomass as the mean estimate from suitable allometric equations generated from studies of harvested trees. We used five allometric equations generated for similar forest types to those in our study that incorporated information on DBH and wood specific gravity: two equations for trees in old-growth forest (Dung et al. 2012, Chave 2014), and three equations for trees in secondary forest (Ketterings et al. 2001, van Breugel et al. 2011, Chave 2014). We did not use equations that included height as a predictor as this is extremely difficult to measure accurately in closed canopy forests and on steep terrain. We did, however, calculate the biomass by measuring heights and DBH of 39 randomly selected trees (DBH range = 75.7–206.9 cm) for which we were able to accurately measure height using a clinometer. For these trees, we compared biomass from the equation that incorporated height with biomass from the one that did not (both equations from Chave 2014). We found that allometric equations with height generated slightly higher biomass estimates than equations without height (matched paired t test, t = 2.25, P = 0.03, RMSE = 6.07 Mg), suggesting that our estimates of biomass are conservative (lower carbon) across our plots. For trees with a DBH of 1–5 cm, we calculated tree biomass using the same allometric equations as those used for larger trees, because the few equations developed specifically for younger trees did not incorporate wood specific gravity as a predictor variable (Nascimento and Laurance 2002).
We measured the DBH at 1.3 m height of all lianas larger than 2 cm DBH in two 1 × 30 m sampling subplots located on the plot sides (V1-2, Fig. S1E). We converted the liana DBH into biomass using five allometric equations for lianas that have been developed for tropical forests (Putz 1983, Gehring et al. 2005, Schnitzer et al. 2006, Sierra 2007, Addo-Fordjour and Rahmad 2013). We used the mean of these five estimates as a measure of the biomass of each liana. We calculated subplot liana biomass by summing the biomass estimates of all lianas for each subplot. Finally, liana biomass for each plot was calculated as the average of the two subplot biomass estimates.
Estimating dead biomass
We measured deadwood and leaf litter to estimate the carbon stock in dead vegetation in each plot. To estimate deadwood biomass, we recorded all standing and fallen deadwood larger than 5 cm DBH within each 10 × 30 m sampling plot. We measured the diameter at both ends of the fallen dead wood and its total length (in all cases, these measurements were only taken for the section of deadwood inside each plot). For standing deadwood, we measured the diameter at the bottom of the deadwood and its height using either a measuring tape (when the top was accessible) or a clinometer (when the top was not accessible). When possible, we also measured the diameter at the top of the deadwood. We measured deadwood volume using the “frustum of a cone” formula when diameter at the top and bottom could be measured.(Pfeifer et al. 2015). When the top diameter could not be measured, we assessed volume using the formula for the volume of a cone. We assigned each standing and fallen deadwood into one of five decomposition classes ranging from class 1 (recently dead intact wood) to class 5 (almost decomposed) following Pfeifer et al. (2015). When deadwood was class 1, we extracted a wood core to calculate deadwood density. For the rest of the decay classes, we extracted wood density estimates for each class from the literature (Pfeifer et al. 2015) to estimate deadwood biomass.
We collected all leaf litter (fallen leaves, twigs, and grasses) from three 1 × 1 m subplots centered within each 2-m2 subplot for each 10 × 30 m plot. We measured total leaf litter volume in situ using a “compression” cylinder (Parsons et al. 2009) and calculated the dry mass (oven dried to constant mass) of a 1 L subsample to estimate total dry biomass of leaf litter.
Estimating total carbon
We used our four biomass estimates (living tree, lianas, deadwood, and leaf litter) to calculate biomass within each plot (Mg/ha). To derive an estimate of total carbon stock in each plot, we multiplied the plot-level biomass estimate by 0.474, which is the wood carbon to biomass ratio for both living and dead carbon estimated by Martin and Thomas (2011).
Bird sampling
Within each square, three point-count stations were established, spaced 200 m apart from each other (a total of 108 point-count stations across three landscapes). We sampled birds using repeat-visit point counts at each station between 04:45 and 12:30 avoiding sampling in rain or strong winds. We did so in the summer (April-May) breeding season and in winter (January-February) when Palearctic migrant bird species frequent the region. At each station, four point counts of 10 minutes duration were conducted on consecutive days. However, we were only able to make two visits during summer at nine of our point counts in Kohima landscape due to the early onset of the rainy season and associated flooding. Additionally, we were not able to conduct any point count survey in Kiphire landscape in winter season owing to a civil unrest. This resulted in a total of 414 point counts (108 point count stations) in summer and 252 counts (63 point count stations) in winter.
Any bird seen or heard during the point count duration was recorded with care taken to avoid double counting of the same individuals. To allow for interspecific variation in detection, we estimated different distance categories from the centre of the station as A= 0 – 25 m, B= 25 – 50 m, C= 50 – 100 m, and D= >100 m. At every point count station across different habitats, we recorded all species detected within these distance categories. However, for analysis we chose a 50 m radius to avoid bias in detection across different habitats. Similar point count radii have been used in studies conducted in both primary and secondary forests (Bicknell et al., 2015; Gilroy et al., 2014; Socolar et al., 2019). The entire duration of each point count was recorded with a sound recorder (Olympus LS11) to allow unknown vocalisations to be subsequently identified using online reference material (xeno-canto.org) and assistance from regional experts. We randomized the sampling order of the plots to reduce bias due to survey time, while raptors and birds flying over the plots were excluded from the analysis. Nomenclature followed Jetz et al., 2014) which was compiled from Birdlife International world list (version 3), Handbook of the Birds of the World (Del Hoyo et al., 1992) and IOC world list V2.7 (2010). |
