Climatic Signals in Tree Rings of Heritiera fomes Buch.-Ham. in the Sundarbans, Bangladesh

Mangroves occur along the coastlines throughout the tropics and sub-tropics, supporting a wide variety of resources and services. In order to understand the responses of future climate change on this ecosystem, we need to know how mangrove species have responded to climate changes in the recent past. This study aims at exploring the climatic influences on the radial growth of Heritiera fomes from a local to global scale. A total of 40 stem discs were collected at breast height position from two different zones with contrasting salinity in the Sundarbans, Bangladesh. All specimens showed distinct tree rings and most of the trees (70%) could be visually and statistically crossdated. Successful crossdating enabled the development of two zone-specific chronologies. The mean radial increment was significantly higher at low salinity (eastern) zone compared to higher salinity (western) zone. The two zone-specific chronologies synchronized significantly, allowing for the construction of a regional chronology. The annual and monsoon precipitation mainly influence the tree growth of H. fomes. The growth response to local precipitation is similar in both zones except June and November in the western zone, while the significant influence is lacking. The large-scale climatic drivers such as sea surface temperature (SST) of equatorial Pacific and Indian Ocean as well as the El Niño-Southern Oscillation (ENSO) revealed no teleconnection with tree growth. The tree rings of this species are thus an indicator for monsoon precipitation variations in Bangladesh. The wider distribution of this species from the South to South East Asian coast presents an outstanding opportunity for developing a large-scale tree-ring network of mangroves.


Introduction
Mangroves are carbon rich ecosystems that lie at the interface between land and sea in the tropical and subtropical regions, providing a wide array of resources and services, such as protection against natural calamities, habitats for wildlife and fisheries, support socioeconomic activities and maintain ecological balance [1][2][3][4][5]. Sundarbans, the largest single

Study area and climate
The study was conducted at Sundarbans which is located in the south-western part of Bangladesh (Fig 1). The forest is characterized by a complex network of branching and meandering distributaries and rivers with a width that varies from a few meters to few kilometers. The main rivers are connected to the Ganges river through the Gorai river and they enclose a collection of low-lying, shifting islands. The tidal amplitude throughout the forests is 3-4 m [47]. Moreover, based on tidal amplitude, the forest can be divided into four zones: inundated by all tides (newly accretions), inundated by normal high tides (covers most of the area), inundated only by spring high tides (mostly in the northern part), and inundated by monsoon high tides (north-eastern part) [9]. The drainage pattern shifts eastward along with the west-to-east tilt, resulting in a substantial reduction in fresh water flow into the western zone, and a natural east (low) to west (high) [9,47] and north (low) to south (high) direction salinity gradient within the forests [14]. Freshwater discharge within the forest decreases from November to May [48], resulting in an increase of salinity from February to May [14]. Average tree height within the forest also varies from the east (10-20 m) to west zone (<10 m) [9].
The study areas are characterized by a monsoonal climate with a unimodal distribution of precipitation (Fig 2). The monsoon ranges from June to September which is preceded by a hot and muggy pre-monsoon (March-May) with sporadic rainfall. The monsoon is followed by a post-monsoon (October-November), and a dry winter (December-February). The climate data is available since 1948 in both stations (Khulna and Satkhira). Distance between both stations is approximately 50 km and climate condition is similar due to close association between precipitation (r = 0.48, p<0.01) and temperature (r = 0.23, p<0.05) of both stations. The average temperature of the two stations ranges from 18 to 22°C during winter, and from 27 to 31°C rest of the year. The average relative humidity throughout the year ranges from 69 to 83%. The average annual precipitation is 1600 mm.

Site variable analysis
Twenty soil samples were collected (during January) in polythene bags from each sampling site (Fig 1) over a depth of 15 cm and brought to the laboratory. The sand, slit and clay percentages were measured using hydrometer method [49]. The electrical conductivity (EC) was determined in a solution of 1:5 soil-water mixtures using a conductivity meter (Extech 341350A-P Oyster) and EC was converted to salinity (ECe) [50]. Salinity variation among the sampling sites was analyzed separately for the eastern zone (site 1, 2, 3, 4 and 5) and the western zone (site 6, 7 and 8) using ANOVA test. In addition, variation between the zones was also analyzed using t-test. The inundation classes were denoted as I, II, III and IV as inundated by 100-76%, 75-51%, 50-26%, 25-5% of the high tides, respectively [51].

Stem disc collection and preparation
A total of 40 stem discs of H. fomes were collected at breast height (1.3 m above from the ground) position from eight different sites in the Sundarbans (Fig 1). Because of the complex anatomy (for example, frequent occurrence of wedging and partially missing rings) of this species [40], we used stem disc instead of increment core. All stem discs were stored in the Tervuren xylarium, Belgium (accession numbers Tw64638-Tw64677). The stem discs were frozen for one week to prevent insect infestation. All stem discs were sanded using a sanding machine with gradually increasing grit from 150 to 1200.

Tree-ring analysis
On the sanded discs, tree-ring boundaries were marked with pencil under a stereomicroscope on three radii from pith to bark to check for tree-ring anomalies (Fig 3). Anomalies like wedging and particularly partially missing rings were detected (S1 Fig), and correctly dated the rings in which they formed. Crossdating was performed among the radii of each stem disc by direct comparison of the stem disk and composite skeleton plot was drawn for each tree. If successful, crossdating between trees was done by aligning tree composite skeleton plots [52]. Same procedure was subsequently followed for each site and zone. All pencil marked stem discs were scanned at 2400 dpi and the ring widths were measured using Fiji ImageJ software [53]. Spaghetti plots for all radii were also drawn for comparison (S2 Fig). Correspondence between different trees was checked to ensure that each individual tree ring is assigned to the exact calendar year [54]. In addition, quality of crossdating was verified by considering statistical parameters such as Student's t- [55], Gleichläufigkeit values (GLK) [56] using TSAP-Win software [57]. The crossdating threshold was set at a t-value of 2.0 (p<0.05) and a GLK of 60% [58,59]. These thresholds are lower than the thresholds (t>3.5 and GLK>70%) used in regular dendrochronological practices in temperate and boreal regions where the reference chronologies are available. However, there is no reference chronology available for mangroves or other terrestrial species of nearby area of our study site and therefore individual tree-ring series were compared to other individual tree-ring series rather than to a chronology. In addition, the harvesting date is known and dating of the problematic (wedging or partially missing) rings is possibly an approximation, and the range of possible values is relatively narrow [58]. The thresholds were thus lowered for dating the floating time series because of a narrow range of possible dates is likely [19].
Tree-ring indices were calculated by dividing each of the original tree-ring widths by the value of the fitted spline using the ARSTAN program [60], and developed zonal and regional chronologies. Each growth year was repeated at least four trees to be part of the chronology. Auto-correlation (AC) was calculated to assess the influence of the previous year's growth upon the current year's growth [61]. The expressed population signal (EPS) assesses the degree to which the chronology represents a hypothetical chronology based on an infinite number of samples; an EPS of 0.85 is usually taken to identify the reliable part of a tree-ring chronology [62]. The mean sensitivity (MS) is a measure of the mean relative changes between adjacent ring widths and was calculated for the standardized chronologies [61]. In addition, the Pearson's correlation (r) was calculated.

Climate-growth analysis
The influence of local climate (precipitation, temperature and relative humidity) on tree growth was analyzed using DendroClim 2002 [63] over the common time span from 1948 to 2011. In addition to the individual month, we used the different time periods encompassing annual, pre-monsoon, monsoon and post-monsoon precipitation. The precipitation of these time periods of the previous year was also used in the analysis. Zonal chronologies were correlated with climate data of the nearby meteorological stations: the chronology of the eastern zone with the Khulna and the western zone with the Satkhira station. The regional tree-ring chronology, including both eastern and western zone, was correlated with the average climate data of both climate stations.
Next, we analyzed the relation between large-scale climatic drivers (SST of the equatorial Pacific and Indian Ocean) and local climatic variables (e.g., precipitation) and their direct influences on tree growth. Spatial correlation maps were generated using the KNMI explorer [64,65] and were based on gridded (1x1°) monthly and seasonal SST fields (Hadley Centre HadSST3) [66,67]. Correlations were calculated over the same time span of instrumental climate data (1948 to 2011), and spatial correlation maps were generated to indicate which oceanic regions influence precipitation patterns and tree growth (p<0.10). In this study, we used the SST indices for the Niño3.4 region, as this is the traditional region used to assess the strength of El Niño events [68]. We also generated spatial correlation maps (p<0.10) between the annual Niño 3.4 index and gridded (1x1°) seasonal precipitation fields (CRU T.S3.0) [69]. We tested the influence of El Niño and La Niña events on tree growth using t-test for differences between El Niño/La Niña [70,71] and 'normal' years. For 'normal' years, we used the corresponding next year of each event [24]. Finally, we explored the correlation of the regional chronology and monsoon precipitation with the Indian summer monsoon index [72].

Site variation
The salinity variation was not significant among the sampling sites either in the eastern (F 4,15 = 2.75, p>0.05) or western zone (F 2,17 = 2.90, p>0.05). However, salinity variation between the eastern and western zone was significant (t = 8.42, df = 38, p<0.01), and the western zone showed higher salinity ( Table 1). The inundation category did not show clear relationship with salinity variation but sites with higher inundation category (category I) showed lower salinity.

Tree-ring measurements and chronology building
Concentric tree rings are microscopically visible in all samples (Fig 3), and the tree-ring boundaries are marked by a marginal parenchyma band, predominantly one cell wide but occasionally up to three cells and occasionally mixed with fibers [38,40]. The wedging and partially missing rings were found in the samples of the eastern and western zone (S1 Fig). These problematic rings were detected and dated correctly by checking the whole stem disc. The ring width significantly varied between the zones (p<0.05) and the eastern zone showed higher value (1.17 ± 0.16 mm yr -1 ) than the western zone (0.93 ± 0.15 mm yr -1 ). However, ring-width variation within the zone was not significant in both cases. The overall mean tree-ring width was 1.08 ± 0.16 mm yr -1 ( Table 2).
Despite frequent tree-ring anomalies (wedging and partially missing rings), crossdating between tree-ring series of 28 trees was successful ( Table 2). Two zone-specific tree-ring chronologies were developed (Fig 4A and 4B; Table 2) covering respectively 72 (1940-2011) and 68 (1944-2011) year in the eastern and western zone ( Table 2). The pattern on both zonal chronologies was similar up to 2000 and then the chronology of western zone showed a decreasing trend (Fig 4B). Based on the crossdating criteria and significant correlation (r = 0.60, p<0.01) between two zonal chronologies, we therefore developed a regional chronology covering 77 (1935-2011) year ( Fig 4C; Table 2). The EPS of the east, west and regional chronologies was equal or higher than 0.90 and shows that our samples capture around 90% of the inter-annual growth variation in the population. The auto-correlation (AC) of the three chronologies was low ( Table 2). On the other hand, the mean sensitivity (MS) and the Pearson correlation of three chronologies were similar and showed moderate values (Table 2).

Climate-growth relationships
Increased precipitation during May to September and November enhanced local tree growth in the eastern zone (Fig 5). Trees in the western zone showed similar relationships, except the precipitation of June and November. On the regional scale, precipitation of May to September and November showed positive correlation (Fig 5). The annual precipitation yielded the highest correlations with the three chronologies (east, r = 0.52; west, r = 0.57; regional, r = 0.54) followed by monsoon (east r = 0.51; west r = 0.55; regional r = 0.53) and pre-monsoon (east, r = 0.25; west, r = 0.29; regional, r = 0.29) precipitation (Fig 5). Post-monsoon precipitation was not significantly correlated with the chronologies. Similarly, the precipitation of previous year did not show significant correlation with any of the chronologies. In addition, the monthly temperature and relative humidity did not show significant correlations with chronologies.
We found significant correlations for regional precipitation with gridded sea surface temperature (SST) for the Pacific from September to December and Indian Ocean from August to November (Fig 6). The SST of both oceans during this time period or other periods of the year was not directly correlated with the regional tree-ring chronology. At a global scale, precipitation of May to July and October to December was negatively influenced by ENSO (Fig 7). However, tree growth was not influenced by this force, even in the aforementioned time periods. Moreover, tree growth in strong El Niño and La Niña years did not differ significantly from the corresponding 'normal' years. The Indian summer monsoon index [72] showed poor correlation with local precipitation (r = -0.12, n = 64) and radial growth (r = -0.03, n = 64).

Tree-ring characteristics and chronology development
Despite the distinct character of boundaries (Fig 3), tree-ring anomalies, mostly wedging rings might be associated with phases of low growth rates or eccentric growth occurring in this species [40], and also in other tropical species [58,59,73]. Approximately, 70% trees from the both The chronology of eastern and western zone was correlated with precipitation of Khulna and Shakhira district, respectively. The regional chronology was correlated with regional (average of both districts) precipitation. For current year, the monthly, annual and seasonal, such as monsoon (June-September), premonsoon (March-May) and post-monsoon (October-November) precipitation was used for analysis. In case of previous year (PY-), the annual, monsoon, pre-monsoon and post-monsoon precipitation was used. Dotted horizontal lines indicate 95% significance level and solid vertical line is the boundary between current and previous year. sites could be visually and statistically crossdated which confirms the annual nature of tree rings. This result is consistent with the previous study [40], where the annual nature of tree ring of this species was tested by integrating a cambial marking experiment and cambium activity analysis. The aforementioned study shows that cambial dormancy occurs in H. fomes from January to April in the Sundarbans. Because there was no new restored wood and the growth ring boundary began immediately after the cambial marking (in January) and presence of dormant cambium in the samples harvested in March to April. It is noted that a cambial dormancy lasting more than three months is sufficient to trigger the ring formation in tropical species [74,75]. In addition, synchronization with annual precipitation is also an indicator for existence of annual tree ring in this species [38].
In the eastern zone, the mean radial increment of H. fomes is considerably higher than that of the western zone (Table 2). We might speculate that such inter-zone growth differences resulted mainly from variations in growth condition. Within a homogenous climatic region, the eastern zone is characterized by lower soil salinity than the western zone ( Table 1). The higher salinity stress combined with the lower inundation frequency might be reflected as more unfavorable growth conditions [29,35]. However, the radial increments of this study are similar with the previous study on this species [38,40]. The growth of this species is also in agreement with the relatively slow growth rate of other mangrove species studied in Kenya showing a range of 0-5 mm yr -1 in both R. mucronata Lam. and A. marina (Forssk.) Vierh. [36,76], Xylocarpus granatum J. König (0.62-2.51 mm yr -1 ), Bruguiera gymnorhiza (L.) Lam. (0-2.51 mm yr -1 ), R. mucronata (0.94 mm yr -1 ) and S.alba Sm. (0.31-1.25 mm yr -1 ) from Micronesia [77], A. marina 1.19 mm yr -1 from Australia [78]. We compared the radial growth rate in different continents but it appears precarious due to different confounding factors, such as environment, ontogeny and inter-generic differences, which need to be taken into account to compare growth data [29].
The most important challenges encountered during crossdating were detecting problematic (wedging and partially missing) rings (S1 Fig). However, crossdating success in most of the trees in both zones (Table 2), and strong correlation (r = 0.60, p<0.01) between both chronologies indicate that there is a common external factor influencing the growth of this species [60,74], and lead to develop a regional chronology (Fig 4C). The auto-correlation (AC) is considerably low in each case (Table 2), which might be due to stabilizing residuals after averaging individual chronology. Moreover, this is a shade tolerant species [9], which grows slowly during the juvenile stage and thus a strong growth trend is lacking. The mean sensitivity i.e. year-toyear variability in radial growth is moderate and low values of auto-correlation (Table 2) indicating that the species might response to the annual environmental changes in the Sundarbans.

Influence of climate
The climate-growth relationship in both zones is similar, except June and November when significant influence is lacking in the higher salinity (western) zone. The annual precipitation positively influenced tree growth from local to regional level (Fig 5), suggesting that fresh water availability is one of the important limiting factor for radial growth. A strong monsoonal impact on radial growth is to be expected as 71% of the annual precipitation occurs during this season in the study area. As predicted, the chronologies exhibit strong correlations with monsoon (June-September, Fig 5) precipitation, representing the main growing season of this species. Earlier studies on subtropical mangrove species in the South Florida also reported that diameter growth in mangrove species certainly linked to precipitation, and trees can produce 75 to 90 percent of their annual radial increments during the wet season [79]. The rate of photosynthesis and stomatal conductance also increased in Kenyan R. mucronata during the wet season [80], indicating higher growth rate with higher precipitation. For other terrestrial species, like Teak (Tectona grandis L.), tree growth in a similar monsoonal climate in Myanmar is positively correlated with precipitation during and prior to the monsoon (May-September) [23,28].
Precipitation influences soil salinity in mangrove forests [9]. Hence, this is a promising candidate to explain at least a part of the annual growth variations. Not only precipitation but also river discharge, land run-off and tidal inundation [48] that extends up to 50 km inland [81] influence soil water salinity, resulting in considerable spatial and temporal salinity variations in the Sundarbans [14,82]. Recorded observations also indicate that the up-steam river flow decreases more than 75% from December to April [48] while the water salinity increases around 80-90% within the Sundarbans [14]. The positive correlations between pre-monsoon (March-May) precipitation ( Fig 5) and three chronologies suggest that before starting of the new growing season it might diminish the soil salinity and relieve the saline stress to trees. This could be reflected in the timing of onset of the growing season that indicates by flushing new leaves in May to June [40]. Still, there are some positive effects of November precipitation on the eastern and regional chronology (Fig 5), which probably maintains low saline soil-or stem-water content and thereby creating high water potentials e.g., [83] that might favor tree growth during this season.
Tree-ring data of the Indian sub-continent have been reported to be linked with the SST which could be used for teleconnections among different climate phenomena [20,21,23]. The spatial analysis with the SST data shows a distinct pattern of relationship with local precipitation, dominated by a significant positive correlation with the Indian Ocean from August to November and during September to December for the central equatorial Pacific (Fig 6). However, we did not observe significant influences between tree growth and SST of both oceans, even in the aforementioned time periods. The anomalous ENSO conditions are also not associated with tree growth, even though local precipitation has strong influences on that. Precipitation of May to July and October to December was negatively influenced by ENSO (Fig 7), indicating that El Niño/La Niña years resulting in drier/wetter than normal years. However, instrumental data of local precipitation in the ENSO influencing period (May-July; October-December) is not consistent with all El Niño or La Niña years as observed in previous studies [70,71]. For example, precipitation of the strong El Niño year 1997 is not drier than normal year and similarly La Niña years 1973 and 1975 is not much wetter. Moreover, the correlation patterns also vary between two time periods, such as ENSO from May to July mainly influences the western part whereas during October to December in southern part of the country (Fig 7). The influences of ENSO on precipitation occur in the early and later part of the growing season and could have lesser influence on tree growth. In a similar monsoonal climate (Myanmar), T. grandis correlates negatively with Niño-3 SST, consistent with the tendency for El Niño events to be linked to drought over Southeast Asia [23]. It is exhibited that ENSO events do have impacts on monsoon systems on sub-continental scale [70,71] which suggests that local precipitation variability is related to other oceanic patterns, or only to local/regional atmospheric dynamics e.g., [84]. The Indian summer monsoon index [72] showed poor correlation with local precipitation, an agreement with the earlier studies [68,85], and so the radial growth of trees. This states that the monsoon precipitation over Bangladesh is generally not correlated with India.

Conclusions
For the first time, we present the dendroclimatological potential of H. fomes from Bangladesh mangroves. Its tree rings are well-defined and crossdatable which leads to the development of a regional tree-ring chronology. Despite distinct differences in site conditions such as salinity, both study zones showed similar growth patterns. The tree growth is mainly influenced by the local climate, such as monsoon precipitation. Still there are many high diameter trees (>50 cm) with nearly 200 years life-span [40], which certainly offers the opportunity to construct a longer chronology. The influences of large scale climatic drivers (SST and ENSO) on tree growth were not significant. However, these drivers influenced the local climate only for some periods. Therefore, understanding the climate-growth relationships will be helpful to make inferences about the responses of this ecosystem to future climatic change. Summarized a longer chronology with a good site replication would improve our understanding of the climate/ growth association for this species and support the construction of a tree-ring network within the natural range of H. fomes.