This study was conducted in Ratargul Swamp Forest (RSF), which was declared to be a Reserved Forest under the Assam Forest Act in 1932 with an area of 118.55 ha (Choudhury et al., 2004). Geographically, located between 25°00.025′×0384;N latitude and 91°58.180′×0384; E longitude, RSF is the only freshwater swamp forest in the Sylhet region of northeast Bangladesh ( Figure 1 ).

As it is a swamp forest, the terrain of RSF is not uniform. Most of the land is submerged underwater throughout the year, while the remaining land is plain or low land (Choudhury et al., 2004). The area is situated 35 meters above sea level (asl). The soil is a gray, solid, silty clay loam with a clayey texture (Choudhury et al., 2004). Agricultural fields surrounding the RSF are loamy in general. At the end of the rainy season and during the dry season, soils in the forest area easily turn dry and hard. The soil moisture varies from 18.7 to 37.1% (Sharmin et al., 2016). The mean maximum temperature is approximately 32°C. On average, RSF receives 1250 mm of rain per annum, and relative humidity is approximately 74% during December, while in July–August, it is over 90% (Choudhury et al., 2004). The forest has a canopy coverage of 80% with approximately 15 m canopy height. RSF has two major strata, where the upper story is covered by Koroch (Pongamia pinnata) and the understory is covered by Murta or patipata (Schumannianthus dichotomus) along with other endemic species (Choudhury et al., 2004).

A total of 50 20 m×20 m square plots were laid in the RSF in December 2019. Species, traits, and environment data for this study were collected through intensive field surveys where the locations of the plots were selected randomly with the help of ArcGIS 10.5 software to avoid any sampling bias ( Figure 1 ). The geographic coordinates of each plot were recorded at the center of each plot using a hand-held Garmin GPS. All the trees with a diameter at breast height (dbh) greater than and/or equal to 10 cm were identified with their appropriate taxonomic identity and counted for each plot. The main trunk of trees that possessed more than a single stem was measured at the narrowest point below the branches. High-buttressed trees were measured at approximately 30 cm above the convergence of the protrusions of the buttresses on the bowl. We recorded the number of cut stems and tree canopy openings in a plot as an anthropogenic disturbance factor although we did not assess the timing and rate of the disruption due to unreliable data. A densitometer was taken in the center of the plot to measure the tree canopy coverage. The canopy coverage of the shrub and herb species was, however, measured via visual estimation (i.e., ocular estimate). For the ocular estimate, the average observation of two people was considered (Karim et al., 2020).

We recorded information on six functional traits, i.e., plant height, specific leaf area (SLA), leaf dry matter content (LMDC), wood density (WD), seed mass, and bark thickness, from each of our study plots. Altogether, 511 individual trees were samples from all of our plots following the protocols described in Pérez-Harguindeguy et al. (2013). Tree height was measured with Suunto PM-5 clinometers with the shortest distance between the upper pick point of the tree and ground level, which is measured by a meter (Garnier et al., 2013). The height of each tree standing at dbh ≥ 10 cm was measured from each sample plot. For SLA measurement, fresh leaves were collected from a separate plot. Relatively young but fully expanded leaves presenting to the sun and leaves in shade were gathered from each plot. More than one leaf was collected from each tree. Wood samples were collected by an increment borer. To calculate leaf area, a scanner and ImageJ software were used (Glozer, 2008). SLA was measured as the leaf area of a fresh leaf divided by oven-dry mass (m²/kg). After collecting green leaves, fresh weight was immediately measured. The leaves were then oven-dried for 72 hours at 65–70°C temperature in the laboratory. Oven-dry weight was measured by a digital weighing scale. LDMC was calculated by dividing the oven-dry weight (mg) of a leaf by its fresh weight (g), expressed in mg/g.

Wood density was measured from the dry weight of wood divided by its fresh volume, and samples were oven-dried for 72 hours at 105°C temperature. Afterward, dry weight was measured in milligrams (mg) with a digital weighing scale. The following formula was used to measure the volume of the wood sample: V = πr²h, where V = volume, r = Radius of the fresh wood sample, and h = length of the fresh wood sample. Similar individuals, i.e., fully grown healthy adult trees that had their foliage exposed to full sunlight were sampled for leaf traits and plant height. For the bark thickness of trees, five random measurements of individual trees are made with a thickness gauge. At least 10 mature and live seeds were collected from individuals of a given species (Pérez-Harguindeguy et al., 2013).

The disturbance index (DI) was calculated according to Saimun et al. (2021):

Where DI is the disturbance index, CS is the percentage of cut stems calculated as the number of cut stems in a plot divided by the total stem of that plot, COT is the canopy openness of the tree, and COSH is the canopy openness of shrub and herb species of the plot in percentage.

This DI measures large-scale biomass that integrates the effects of all types of disturbance through recent history, like fuelwood collection, illicit felling, wind throw, logging, or fire. This is important because disturbance is expected to affect forest plots at relatively longer time scales.

To characterize trait dissemination in a community, both the mean and variance of trait values weighted by species abundances are examined. The community-weighted mean (CWM) for each trait t was calculated as the mean of species trait values, ti, of the S species in each plot, with each species, i, weighted by its relative abundance, pi (Violle et al., 2007):

Community-weighted mean values were computed in R-3.5.1 software by using the FD package.

Community-weighted mean trait usually describes the overall functional trait value of the total community by adjusting individual species trait values with the abundance of respective species (Diaz et al., 2007). In the present study, to measure community-weighted trait value, we calculated SLA, LDMC, plant height, WD, seed mass, and bark thickness for all individuals within the sample plots using the “FD” package (Ziout et al., 2017). The empirical Bayesian Kriging method was used to make spatial predictions about the community-weighted mean trait distribution in Ratargul Swamp Forest. To perform kriging spatial analysis, we used ArcGIS software (version 10.4.1).

All statistical analysis was performed in the R-3.6.1 environment (R Core Team, 2019; Pinheiro et al., 2020). Analysis of variance (ANOVA) was used in combination with linear models to test the effects of anthropogenic disturbance on the functional trait diversity under packages “nlme”. The results of the comparison were presented by AB trend line using the “ggplot2” package. Principal component analysis was done by package “factoextra” in R. Pearson correlation was measured in the “data analysis” plugin in Microsoft Excel (version 365).

The distribution, intensity of disturbance, and plant functional traits in RSF are shown in Figure 2 . The intensity of disturbance was higher in the northern and southern sides compared to the eastern and western parts of RSF ( Figure 2A ). Human disturbance is lower in the northern and southern parts of RSF. Among the three resource-acquisitive traits, the intensity of CWM-height and CWM-SLA showed similar patterns in the eastern and western parts of the study area, with the highest values. The northern and southern areas had the lowermost values ( Figures 2B, C ). The eastern region followed by the extreme western area represents comparatively vigorous species present in RSF. Moreover, our study found that Koroch and Hijal trees required less time to regrow due to greater abundance in the area. The intensity of CWM-SeedMass, another acquisitive trait, showed higher values in the northern and southern areas compared to the eastern and western parts of the area ( Figure 2D ). The corner of northern and southern areas represents comparatively higher values of CWM-SeedMass because of the high presence of Barun, Puti jam, and the negligible presence of Koroch. These species are highly stress tolerant and can survive under a high-disturbance regime.

In terms of the resource-conservative traits, CWM-bark-thickness and CWM-WD indicated higher values toward the north–southern part and lower values toward the eastern and western areas of RSF ( Figures 2E, F ), while CWM-LDMC showed the opposite pattern with the eastern and western parts having the lowest values and north–southern part having the highest values ( Figure 2G ). The corner of the north–southern part represents comparatively higher values of CWM-bark-thickness and CWM-WD due to a higher density of Barun and Puti jaam. Human disturbance is also higher there than in the eastern and western parts. Fishing, the movement of tourists, and the presence of a tourist watch tower is the main reason for the disturbance. Again, higher wood density and bark thickness values contribute to enhanced plant survival. CWM-LDMC was comparatively higher in the eastern and western parts compared to the north-southern areas mainly due to a greater presence of Koroch and Hijol trees.

Our results indicated that anthropogenic disturbance is associated with functional traits at the plot level ( Figures 3A, B ). Evergreen species traits, PCA 1 (Dim1) and PCA 2 (Dim2), demonstrated 63.9% and 17.9% total variation of the study findings ( Figure 3A ). This indicated that disturbance accounted for 81.8% variation in functional traits within the studied plots. On the other hand, PCA 1 (Dim1) and PCA 2 (Dim2) demonstrated 67.2% and 17.1% of total variation in functional traits ( Figure 3B ). This indicated that together, disturbance accounted for 84.3% variation in functional traits in the studied plots. In evergreen species ( Figure 3A ), CWM of bark-thickness showed the weakest association with disturbance whereas CWM of wood density showed the closest association with disturbance; similar results were also found in the case of deciduous species ( Figure 3 ).

The anthropogenic disturbance had a significant relationship with deciduous species whereas it showed no significant relationship with evergreen species ( Figure 4 ; Table 1 ). In terms of deciduous species, disturbance had significant positive effects (p<0.05) on CWM-Height (m) and CWM-SLA (kg m⁻²) ( Figures 4A, B ). It was also positively correlated with disturbance and CWM-height (m) and disturbance and CWM-SLA (kg m⁻²) with a correlation coefficient of r = 0.34 and r = 0.35, respectively ( Table 1 ). Disturbance had significant negative effects (p< 0.05) on CWM-bark thickness (mm) and CWM-seed mass (mg) ( Figures 4E, F ). It was negatively correlated with CWM-bark thickness (mm) and slightly correlated with CWM-seed mass (mg) with correlation coefficients of r = −0.36 and r = −0.33, respectively ( Table 1 ). On the other hand, disturbance had no significant effects on CWM-LDMC (mg g⁻¹) and CWM-WD (mg mm⁻³) ( Figures 4C, D ). Disturbance showed no correlation with CWM-LDMC (mg g⁻¹) (r = 0.23, p = 0.15) and CWM-WD (mg mm⁻³) (r = 0.0064, p = 0.96) ( Table 1 ).

Surprisingly, in our study, disturbance showed no significant effect on evergreen tree species: CWM-Height (m) (r = 0.11, p = 0.43), CWM-SLA (kg m⁻²) (r = 0.12, p = 0.41), CWM-LDMC (mg g⁻¹) (r = 0.064, p = 0.66), CWM-WD (mg mm⁻³) (r = 0.011, p = 0.93), CWM-bark thickness (mm) (r = −0.12, p = 0.47), and CWM-seed mass (mg) (r = −0.16, p = 0.28) ( Figure 4 ; Table 1 ).

The present study showed that the community-weighted mean of SLA and tree height had higher trait values in the eastern and western areas compared to the northern and southern parts of RSF. The eastern and western parts showed more resilience (and better ecosystem health) and required less time to grow during various disturbance events sometimes also leading to increased nutrient availability. Cornelissen et al. (2003) also reported similar results wherein areas under less disturbance contributed more nutrients and were comparatively more robust, whereas Moles et al. (2014) reported a minimum community-weighted mean of SLA and height of trees that survived under extreme disturbance events. The community-weighted mean of LDMC and WD was also higher in the southern and northern parts compared to the eastern and western areas. High values of the community-weighted mean of LDC and WD were directly related to poor habitat constitution and highly disturbed areas, respectively. Cruz et al. (2010) reported similar results where a highly disturbed area contributed to poor nutrients and was comparatively less healthy. Poorter et al. (2008) also reported that higher values of wood density and leaf dry matter content contributed to an enhanced plant survival rate. The community-weighted mean of bark thickness and seed mass was found to be higher in the northern and southern parts compared to the eastern and western areas of RSF. Bark thickness which is directly related to the fire regime, fungal attack, and seed size (mass) is an important indicator of plant fitness (Pausas et al., 2014). However, the IAWA Journal (2007) reported that the dead outer bark protects the tree from fires, pathogens, and herbivores; reduces water loss; and provides structural support. Seemi and Shaukat (2010) showed that seed mass is directly related to plant fitness. Eastern and western parts of RSF were more susceptible to fire breakage and fungal and pathogen attacks (Wright et al., 2004).

In our study, specific leaf area and tree height showed a significantly positive correlation with anthropogenic disturbance which is in accordance with the study by Herben et al. (2018) and Ribeiro et al. (2019). Anthropogenic disturbance affecting the nutrient concentration, photosynthesis, and concentration rates of a plant community explains this relationship pattern between disturbance, tree height, and leaf area (Finegan et al., 2015). On the other hand, we found that bark thickness is negatively correlated with anthropogenic disturbance, which is similar to the findings of Graves et al. (2014). We also found seed mass negatively correlated with disturbance, which is similar to the findings of Castro et al. (2008). Grubb et al. (2013), on the other hand, found a positive relationship between plant functional traits and disturbance. Seed mass takes some time to adapt to anthropogenic disturbances in a forest area. This may explain the negative relationship of seed mass with disturbance in our study which may turn positive in due time or with tree maturity.

Resource-conservative traits such as leaf dry matter content and wood density conserve resources to survive in a stressed environment (King et al., 2006). Contrary to Ribeiro et al. (2019), leaf dry matter content and wood density were not affected by disturbance in the present study.

Our study found that evergreen species were not significantly affected by anthropogenic disturbance, whereas deciduous species were significantly affected. Similarly, Bonfil et al. (2004) found that evergreen species were more resistant to disturbance compared to deciduous species. With the increase in disturbance intensity, the CWM changed from conservative traits to more acquisitive traits (Carreño-Rocabado et al., 2012). The present study also found that disturbance had a significant positive effect on deciduous species’ acquisitive traits (SLA and tree height and a negative significant effect on acquisitive seed mass). We also found that disturbance had a negative significant effect on the conservative trait of bark thickness. Sfair et al. (2018) and Graves et al. (2014) reported similar results, where disturbance had a significant effect on acquisitive traits versus conservative traits. It was also expected that a higher disturbance would lead to a decrease in the CWM of other acquisitive traits but instead, there was a significant increase. Conservative traits survived under stressful conditions compared to acquisitive traits (King et al., 2006). The community-weighted mean of SLA and height increased with disturbance intensity, but bark thickness and seed mass dropped ( Table 1 ; Figure 4 ), implying an increase in primary productivity along with carbon and other nutrient cycling in the short term. Kaul et al. (2010) found that in the early stages of forest regeneration, an increase in fast-growing species was accompanied by an increase in primary productivity. Species with low community-weighted mean of SLA and high WD, on the other hand, were projected to influence primary productivity over longer periods because they increased their abundance at the expense of fast-growing species.