Population trends and effects of local environmental factors on waterbirds at Tanguar Haor freshwater wetland complex in northeast Bangladesh

,


INTRODUCTION
Identifying and managing important biodiversity sites is a conservation priority for preventing the decline of global biodiversity. Population trends of birds are widely used as surrogates of biodiversity because they are easily detected and counted in complete systems, compared to other taxa (Burger 2006, Larsen et al. 2012). In addition, long-term datasets on bird populations are increasingly used for answering important ecological and management questions that are useful for conservation management (Magurran et al. 2010, Hansen et al. 2021. Attempts to use long-term datasets to explore bird population trends have rarely been used to inform conservation decisions in South Asia and have not been done in Bangladesh so far, which is located at the intersection of two major flyways, the East Asian-Australian Flyway and the Central Asian Flyway. Wetlands in Bangladesh provide a diversity of habitats for many species of migratory waterbirds during the boreal winter. The topography of the country and its location within the Ganges, Brahmaputra, and Meghna River systems resulted in a network of rivers intersecting the country. These streams feed an abundance of wetlands, such as haors (shallow depressions or seasonally flooded freshwater wetlands), baors (freshwater oxbow lakes), beels (low-lying depressions, often located within the Haor basins), water reservoirs, and lakes, which are seasonally inundated to a depth of 30 cm or more. Different rivers crisscrossing the land have shaped Bangladesh into its current form with internationally significant transboundary rivers and one of the world's most active deltas (Byomkesh et al. 2009).
One of the key freshwater wetlands is Tanguar Haor, located in northeast Bangladesh. Tanguar Haor was declared an ecologically critical area (ECA) in 1999, a Ramsar site in 2000 (Ramsar Convention Secretariat 2007), an important bird and biodiversity area in 2004 (BirdLife International 2021a), and a flyway network site in 2010 (EAAFP 2011) because of its ecological, biological, and economic values (Byomkesh et al. 2009). Tanguar Haor is known to support a total of 206 species of birds including the critically endangered Baer's Pochard Aythya baeri, endangered Pallas's Fish-eagle Haliaeetus leucoryphus, vulnerable Common Pochard Aythya ferina, Greater Spotted Eagle Clanga clanga, Bristled Grassbird Schoenicola striatus, and seven near-threatened species (Alam et al. 2012, Round et al. 2014. Each winter, a large number of waterfowl congregate at Tanguar Haor, ranging from 60,000 to 100,000 individuals in recent years with the highest congregation of 280,000 waterbirds in 2004 (Li et al. 2009, Alam et al. 2012. In addition to the avian diversity, Tanguar Haor supports the livelihoods of approximately 70,000 local people (Hossain et al. 2017) and is home to 19 species of mammals, 27 reptiles, 11 amphibians, 141 freshwater fish, 107 genera of phytoplankton, and around 200 plant species (Bangladesh National Herbarium 1997, Giesen and Rashid 1997, Muzaffar and Ahmed 2007, Alam et al. 2012. Considering the national and global significance of Tanguar Haor for migratory waterbirds, fisheries, local livelihoods, and conservation needs, knowledge gaps on how local environmental factors influence waterbirds at species level and their population trends in South Asia should be addressed to improve the management of the site. We (1) summarize and compare a decade long (2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020)(2021) waterbird census data, (2) determine population trends of 39 species of waterbirds, (3) discuss relationships between local abiotic factors (e.g., rainfall, water-depth) and waterbird population trends, (4) identify conservation priority sites within the Haor complex, (5) discuss key threats, and (6) recommend future studies and conservation measures.

Study area
The 9500 ha Tanguar Haor (Fig. 1) is located (25°08′49.1″ N 91°0 4′44.9″ E) at Sunamganj district in northeast Bangladesh, bordering the Garo Hills (c. 1500 m) of Meghalaya State, India and is situated in the catchment area of the world's highest annual rainfall occurring in the area of Cherrapunji. During the rainy season (May-September), the entire area is submerged (Giesen et al. 2000) with an average annual rainfall of over 5000 mm. Water levels drastically fall from 6-10 m to 2-6 m during the dry season (October-April) resulting in numerous shallow waterbodies (mean ± SD, 7.34 ± 1.88 m), channels, reedbeds, other wetland grasses, and emergent herbage (IUCN Bangladesh 2015). These round-shaped floodplain depressions and the surrounding vegetation are known as "beels" and offer important habitat for local fisheries, migratory waterbirds, and passerines (Giesen et al. 2000, Muzaffar and Ahmed 2007, Alam et al. 2012, Chowdhury 2013, Round et al. 2014. We conducted our surveys at six beels in Tanguar Haor comprising Berberia, Chotainna, Hatirgatha, Lechuamara, Rowa, and Rupaboi beels.

Waterbird census
A series of waterbirds (as recognized by the Ramsar Convention, including grebes, cormorants, pelicans, herons, egrets, storks, ibises, spoonbills, flamingos, ducks, geese and swans, cranes, rails, jacanas, shorebirds, gulls, and terns) censuses were undertaken annually during the boreal winters, between 2008 and 2021, from 0600 in the early morning to 1700 hours in the afternoon. Counts were undertaken during the dry season, January and February each year, when birds are less likely to move between sites (Li et al. 2009), by 3 to 5 members with a minimum of 10 years of waterbird survey experience. The count data as well as other information including site name, date and time of the count, and information relating to weather and habitat were recorded in specialized and standardized data sheets developed for the Asian waterbird census (see Delany 2010 for further details on survey methodology).
These censuses were primarily conducted at six beels ( Fig. 1) that hold most of the birds (Alam et al. 2012, Mundkur et al. 2017, which represent the Haor as a whole. Because of the inundation of the habitat, we used medium-sized wooden trawlers to travel and often counted from the boat when nearby land was further than 500 m from waterbird flocks. We counted waterbirds individually and estimated aggregations (e.g., > 1000 individuals) following the widely used block count method (Li et al. 2009). Each flock of birds was counted multiple times to ensure accuracy of the count as well as increase the chance of detecting rare or smaller species. Priority was given to ensure the birds were not disturbed during the counts. During the census week, neighboring villages were requested to minimize their activities in the core Haor areas to further reduce local disturbance (Alam et al. 2012). We identified all waterbirds to species level and counted them using multiple spotting scopes and binoculars. In addition, we took photographs of species we could not identify immediately. Priority was given to not disturb feeding or roosting birds while moving between sites.

Identifying priority sites
Repeated censuses, following the same counting method, by the same observers, and covering the same sites (beels within the Hoar complex), over fourteen years, allowed us to determine temporal and spatial variation across the sites. This aided with the identification of sites regularly used by large numbers and a high diversity of waterbirds throughout the wintering period, including species of global concern, and thus identified priority sites for conservation.

Waterbird habitats
We used the modified normalized difference water index (MNDWI) in ArcMap (v.10.8) to classify LandSat images in habitat categories relevant to waterbirds. These include LandSat 5 for 2010 and 2011, LandSat 7 for 2012 and 2013, and LandSat 8 for (2014-2020) satellite images from January and February, downloaded from the United States Geological Survey' (USGS) official website (earthexplorer.usgs.gov). We classified the habitats into four broad habitat categories of Tanguar Haor: (1) deep water, which held water during the driest months, including rivers, (2) shallow water, including muddy areas, wet banks, which held little water during the driest months, (3) vegetation, e.g., grasses, herbs, shrubs, and trees, and (4) human settlements, e.g., villages, infrastructure, etc.
We determined the area covered by these four habitat classes within Tanguar Haor for each year (sum of all pixels for each class) because these habitat classes change annually based on factors, such as rainfall and water level management (Alam et al. 2012). Because the area remained constant, the sums of these counts were constant from year-to-year. Consequently, the resulting counts are compositional in nature, i.e., the relative frequencies of all land use classes sum up to one and are not independent of the relative frequencies of the other land use classes. We used the R-package "compositions" (van den Boogaart and Tolosana-Delgado 2013) to convert land use class frequencies to their relative compositional positions on the Aitchison Simplex (Pawlowsky-Glahn and Buccianti 2011). This allowed us to calculate the log-ratio of land use classes relative to each other while maintaining the independence of frequencies.
We calculated the log ratio of shallow water compared to deep water areas (hereafter shallow/deep), and the log ratio of deep water, shallow water, and vegetated areas compared to human settlements (hereafter habitat/other).

Population trends of selected waterbirds and relative abundance
Trends of 47 species of waterbirds, which regularly occurred at Tanguar Haor, were assessed based on annual counts. We calculated trends in waterbird populations over time, using an ordinary least squares regression, and used the slope of the trend as a measure of change in their populations. Waterbird species encountered on 80-100% of the 14 visits were classified as very common, those seen during 50-79% of the visits as common, 20-49% of the visits as uncommon, and < 19% of visits as rare.

Group-level analysis of environmental effects and overall trends
We estimated how environmental factors affected mean species counts depending on habitat preference and seasonal status while taking into account overall trends. To investigate group-level effects of environmental covariates, we categorized all waterbirds into three broad categories based on their habitat preference (Billerman et al. 2020): (1) diving birds (diving ducks, grebes, cormorants, and darters), (2) dabbling ducks (duck species that forage in shallow water), and (3) wading birds (shorebirds, egrets, and herons) and matched the data for each year with the habitat variables described above and the total annual rainfall for the prior year (data from the Bangladesh Meteorological Department). Because environmental data were only available for the years 2010-2020, we excluded the other years from this analysis.
We aggregated the annual counts for each of the waterbird groups and modeled (generalized linear model) how mean count for each species of group g and migratory status s, N gs , was affected by rainfall, the log ratio between shallow and deep water (shallow/ deep), as well as the log ratio between water and vegetation and other land use (habitat/other). Because the model (model 1) could not detect a discernible effect of the latter, we excluded it from the analyses. We allowed the intercept and effect of the environmental covariates to vary with species groups by including it as an interaction term. We included year as an additional predictor for the mean to assess group-level trends over time. Because population trends might be affected differently between resident birds as opposed to migratory birds that spend only winter at Tanguar Haor, we included a three-way interaction term between group, year, and seasonal status (resident or migratory). We assessed whether the model residuals were in line with the model assumptions by simulating the fitted model (R Core Team 2021), which suggested that the model residuals violated the assumption of homoscedasticity. We therefore modeled (model 2) dispersion alongside the mean by including a dispersion term using group, season, and their interaction as covariates. We assessed Model 2 in the same way as Model 1 and found no significant problems. We further assessed an alternative model (Model 3) assuming a poisson error distribution, which revealed that the residuals from this alternative model violated the assumptions of the error distribution as well as the assumption of homoscedasticity, and so we retained the negative binomial generalized linear model (Zuur et al. 2013). Consequently, our final model predicted the mean count for each species N gs using the following structure: N gs ~ NB(μ gs , σ gs ) with log(μ gs ) = α gs + β gs x 1 + γ gs x 2 + δ gs x 3 + θ gs x 4 and log(σ gs ) = κ gs + λ gs , where N gs is the estimated mean count for a species of group g and seasonal status s, NB stands for the negative binomial distribution with mean μ gs and dispersion parameter σ gs , and α gs is the intercept term of the linear equation. β g is the effect of annual rainfall specific for group g, and γ g the group-specific effect of shallow/deep water, and δ gs is the effect of year since beginning the monitoring. Similarly, κ gs and λ gs are the effect of group and seasonal status on the variance of the mean.
We found that the annual rainfall for the prior year and the ratio of shallow to deep water areas were correlated (Pearson correlation coefficient r = 0.33), and thus computed two alternative models with a single environmental predictor each. We evaluated the model using simulated residuals and found that all model assumptions were met. All analyses were conducted in R v. 4.2 (R Core Team 2021), using the packages glmmTMB (version 1.1.3) and DHARMa (version 0.4.5), and data visualizations were created using the packages ggplot2 (version 3.3.5), sjPlot (version 2.8.11), and interactions (version 1.1.5).

RESULTS
A total of 69 waterbird species were recorded between January and February in 2008-2021 at Tanguar Haor, of which 41 (59%) species were migratory and 28 (41%) resident (Table 1). A total of 24 (35%) species were assessed as very common, 13 (19%) species were assessed as common, 10 (15%) as uncommon, and 22 (32%) as rare. During the study period, a maximum of 166,788 waterbirds were counted in 2013 and a minimum of 28,925 individuals in 2010 at all sites of the Tanguar Haor complex (Fig.  2). The overall population trends (Mann-Kendall test, tau = 0.12, p = 0.64) of all waterbirds did not show any particular pattern. Peak counts of 15 species exceeded the 1% threshold of their East Asian-Australasian flyway population estimates including the globally vulnerable Common Pochard, near-threatened Ferruginous Duck, and Black-tailed Godwit.

Priority sites
Among the six sites (beels) of Tanguar Haor, Lechuamara beel regularly supported more than 20,000 waterbirds including a significant number of priority species (Fig. 1) meeting the Ramsar Criterion 5 alone. Berberia beel, Chotainna beel, and Rupaboi beel ( Fig. 2A) supported more than 20,000 waterbirds at least once in recent years, but Chotainna beel regularly held globally threatened and near-threatened species, including 6500 globally vulnerable Common Pochard in 2013 (Fig. 2B). In addition, a maximum number of species (47) were recorded at Chotainna beel, followed by 45 species at Lechuamara beel. Therefore, we consider these two beels as the most important sites within the Tanguar Haor complex for the conservation of resident and migratory waterbirds.

Group-level trends and environmental effects
We found that the full model could not detect consistent environmental effects, likely due to the correlation between both environmental covariates, whereas the models with single environmental covariates were able to detect an effect of both rainfall and the ratio of shallow to deep water areas. Because the results from the two models led to the same interpretation, we reported on the model using only annual rainfall for the prior year to counts because it explained slightly more of the variance in the data (R² = 0.837; Table 2). The model estimated that the mean count was highest for the average migratory dabbling and diving birds, and lowest for migratory wading birds. Group-level trends indicated that over the period from 2010 to 2020, mean per-species counts remained stable for migratory dabbling and diving birds and were on the verge of showing a consistent increase for migratory wading birds and resident diving birds (Fig. 4, Table  1). Resident dabbling birds showed a consistent negative trend throughout the study period ( Fig. 4; see Table 2 for confidence intervals on the estimated effect size). The model also highlighted that habitat had an effect on the observed counts of dabbling and wading birds but not diving birds (Table 2). We found that total annual rainfall had an overall negative effect on the count of wading birds but a positive effect on dabbling birds (Fig. 5).

DISCUSSION
Our findings suggest that waterbird numbers are higher at Tanguar Haor compared to other wetlands sites in Bangladesh and Tanguar Haor is one of the important freshwater wetlands in the whole of Asia. Tanguar Haor supported more than 50% of all waterbirds counted in Bangladesh in 2010-2013 and 2015, and from 2008 to 2015, an aggregated total of 1,209,315 waterbirds were reported from Bangladesh, of which Tanguar Haor itself contributed around 43% of all stated waterbirds (Thompson et al. 2018).
Compared to counts from nearby sites in neighboring countries, it is clear that Tanguar Fig. 4. Group-level (dabbling, diving, and wading) trends of mean estimated counts of species over the study years. Lines and shaded areas indicate the estimated model effect size and 95% confidence intervals for resident (continuous lines) and migratory (dashed lines) of dabbling birds (A), diving birds (B), and wading birds (C). Points indicate the partial model residuals after accounting for species and environmental effects. The plot reflects the estimates from the model using total annual rainfall as environmental covariate.

Fig. 5.
Group-level effects of the log ratio of rainfall (A) and log ratio of shallow to deep water areas on the mean counts of species. The line and shaded areas indicate the mean effect size and 95% confidence interval estimated by the model for each group, whereas points show the partial model residuals after accounting for species and migratory status. Note that the plots were produced from the two alternative models.

Table 2.
Here we show the model estimates for group-wise mean estimates of species counts based on the data from 2010-2020. (A) shows the mean estimates with confidence intervals broken down by group and seasonal status; (B) highlights the group-level effects of environmental variables on mean counts; and (C) shows the overall trends of mean counts over the study period per group and seasonal status. Estimates are shown as incidence rate ratios, i.e., estimates < 1 suggest that the mean count is lower for a given effect compared to the intercept, and estimates > 1 indicate the mean count is higher compared to the intercept. Effects that could be discerned with confidence by the model (i.e., 95% confidence intervals not overlapping with 1) are highlighted in bold.

Population trends
Our results suggest that the count of dabbling and diving birds remained constant over time, whereas the wading birds showed a slightly positive trend (Fig. 4). Long-term population trends of Baer's Pochard and Common Pochard (Fig. 3) are decreasing globally (BirdLife International 2021b, c) and regionally (SoIB 2020a). Since the 1990s, Baer's Pochard population has declined 99% in Bangladesh. However, the species could pass undetected among large waterbird flocks (Chowdhury et al. 2012). In alignment with our estimate (Fig. 3), the number of Falcated Ducks wintering at critical sites of the Yangtze River floodplain showed an increase in abundance in recent times (Zhang et al. 2020).
Ferruginous Duck populations seem to be fluctuating in the Asian region (BirdLife International 2021d) similar to our observations at Tanguar Haor (Fig. 3), although a declining trend is observed in neighboring Myanmar in the Ayeyarwady Basin over the past decade. The species has disappeared from the Ayeyarwady river, while maintaining a stable population trend in lake Indawgyi (Zöckler and Kottelat 2018).
Long-term monitoring of the wintering population of Blacktailed Godwit is showing a stable trend in the Asian waterbird census (AWC) regions (Mundkur et al. 2017) including India and Myanmar Kottelat 2018, SoIB 2020b). Based on AWC counts (2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015), Northern Lapwing and Grey-headed Lapwing total counts showed stable to increasing trends. We also observed a similar trend for lapwings (Fig. 3) in Tanguar Haor, however, in India the current population trend is uncertain, which might be because of uncertainty or unavailable long-term monitoring data (SoIB 2020c, d).
Overall, populations of resident dabbling ducks appear to be declining (Fig. 4). Despite being a globally least-concern species, the abundance of Cotton Pygmy Goose is sharply declining in Tanguar Haor (Appendix 1). A similar trend has been observed in India (SoIB 2020e, Mukherjee et al. 2022) and Lake Indawgyi, the largest natural freshwater lake in Myanmar and a Ramsar site (Zöckler and Kottelat 2018). The reasons for the continuous decline are not clearly understood (SoIB 2020e) but are presumably related to the degradation of inland freshwater wetlands and lack of suitable nesting trees (Thompson et al. 2018). The opposite trend is evident in the case of Indian Spot-billed Duck whose numbers are increasing in the Ayeyarwady River in Myanmar but are declining in Tanguar Haor (Appendix 1) and in Indian freshwater wetlands. Among the species whose numbers are increasing at Tanguar Haor are, for example, the Glossy Ibis, which is also exhibiting a strong increase in the wetlands of India and the Ayeyarwady River basin in the Myanmar (Zöckler and Kottelat 2018). Understanding the reasons behind population declines of resident waterbirds should be a high priority research.

Effects of local environmental factors
Water depth is a well-known factor that influences the foraging ability of ducks, and feeding areas need to be shallow enough for them to reach the substrate (Behney 2020). Our findings suggest that total rainfall (Fig. 5A) and consequently water depth, influenced by annual rainfall (Fig. 5B), affected the average number of dabbling and wading bird populations. Similarly, wading birds exhibit a negative correlation with higher annual rainfall in the year prior to the count (and conversely, a positive correlation with increased ratio of areas with shallow as opposed to deep water areas), whereas this trend was the opposite for dabbling birds (Tables 2). Populations of diving birds do not seem to be affected by annual rainfall or water depth (Table 2, Fig. 5). Extreme rainfall and flash-flood events in northeast India have increased in the past two decades due to changing climate and are likely to become even more frequent (Guhathakurta et al. 2011), which may have led to increased water depth at Tanguar Hoar in those years and thus influenced the occurrence and abundance of dabbling and wading birds. Therefore, site managers should make informed decisions targeting waterbird groups or species and manage water levels at Tanguar Haor based on annual rainfall patterns, especially considering extreme rainfall events, using the evidence illustrated in our work.

Threats
A spatio-temporal study by Haque and Basak (2017) found that anthropogenic impact has converted about 40% of the original Haor Basin into low-lying agricultural land and human settlement over just three decades. Analyzing normalized difference water index (NDWI), the study also indicates that from 1980 to 2010 about 71% of the deep-water body has been degraded into shallow water, then converted to rice cultivation. Other anthropogenic disturbances in the form of increased domestic duck rearing, cattle grazing, motorized and hand paddled boats, and fuelwood collection are some prominent threats to migratory waterfowl at Tanguar Haor (Muzaffar 2004, Alam et al. 2012). In addition, illegal waterfowl hunting using nets and poisoned bait are burgeoning problems in northeast Bangladesh's wetlands, including the Tanguar Haor area during winter (Chowdhury et al. 2012, Datta 2021. Although playing an important role in the conservation of overall biodiversity of the wetland, swamp habitats and waterbirds have received less attention than fisheries management (Round et al. 2014). Five fish sanctuaries and three bird sanctuaries (Lechuamara beel, Chotainna beel, and parts of Hatirgatha and Berberia) were identified under the community-based sustainable management of Tanguar Haor (CBHMTH) project (Chowdhury 2013, IUCN 2016, however, there were few management interventions after the project ended, and there was no reflection in waterbird numbers ( Fig. 2A).

Conservation and management
The first conservation project at Tanguar Haor was initiated by the Government of Bangladesh in the early 1990s and a subsequent management plan was formulated in 2000. Later, IUCN Bangladesh implemented three phases (2006-2016) of a co-management initiative via the community-based sustainable management of Tanguar Haor (CBHMTH) project to promote sustainable use of natural resources (IUCN Bangladesh 2016). The initiative introduced the co-management system and abolished the traditional leasing systems (where rich and influential people would lease out the beels and harvest fish, which proved to be destructive to the wetland ecosystem), ensuring rights to the fishers, to the waterbodies, and establishing a permitbased fishing scheme in Tanguar Haor (IUCN Bangladesh 2016). A three-tiered co-management organization was established encompassing 76 villages with 7089 members, with 28% being women-led households. The co-management system involved various stakeholders, from local stakeholders at the grassroots level to the highest policymakers in the Government of Bangladesh. A three-tiered co-management organization was established encompassing 76 villages with 7089 members, with 28% being women-led households. The co-management system involved various stakeholders, from local stakeholders at the grassroots level to the highest policymakers in the Government of Bangladesh. Through community involvement, the conservation of important habitats was initiated by the afforestation of degraded swamp habitats through the plantation of native Hijol Barringtonia acutangula and Koroch Pongamia pinnata trees inside the Haor as well as in the peripheries and villages, along with homestead seedlings and sapling distribution. Patrolling community guards were formed and trained to stop illegal fishing and bird hunting (IUCN Bangladesh 2016).
We recommend that the 2010 Tanguar Haor management plan be revised following the Tanguar Haor management plan framework and guidelines (IUCN Bangladesh 2015), addressing emerging threats such as cattle herding and domestic duck grazing, deforestation, illegal resource extraction, and water-level management.
Furthermore, our long-term findings recognize Lechuamara and Chotainna as two critically important beels for migratory waterbirds. These beels should thus receive immediate conservation attention by declaring them bird sanctuaries where fishing, cattle herding, domestic duck grazing, and entry by tourists would not be permitted, following the guidelines provided by the Ramsar Convention (Ramsar Convention Secretariat 2007).