Differential responses to habitat structure and degradation by two Grassbirds (Fam. Locustellidae) syntopic in floodplain grassland

Wet grassland birds world-wide face mounting threats due to habitat destruction especially in the Indo-Gangetic floodplains, where in addition to intense human use, primarily for agriculture, there is a paucity of baseline information to inform conservation and management. We examined the response of two co-occurring floodplain grassland specialists, the globally threatened Bristled Grassbird Schoenicola striatus (VU) and widespread but little-studied Striated Grassbird Megalurus palustris (LC), to habitat structure and human use in floodplain grassland at two sites, one on the lower Ganges River and one on the Brahmaputra River, Bangladesh through point counts combined with vegetation surveys. Responses from the breeding seasons of 2018 and 2019 were assessed by linear models and bird densities estimated using distance sampling. Results indicate that the occurrence of Bristled Grassbird increased with homogeneous tall Saccharum grass cover but decreased with increased crop cover, which may be the main driver of its decline. Striated Grassbird showed a more generalized pattern of positive association with tall grassland vegetation including taller crops and exotic plants, but was also positively associated with grass harvesting indicating its avoidance of homogeneous tall grassland. Although the density of Striated Grassbirds at the Brahmaputra site was 3–4 times greater than that of Bristled Grassbird, it was nearly absent from the Ganges site, revealing previously unappreciated micro-habitat preferences. Comparison with a more common species provided a clearer understanding of habitat and human influences on the more threatened grassbird. Long-term conservation of Bristled Grassbird requires careful management of grass harvesting and cattle grazing.


Introduction
Globally, grassland birds have experienced major population declines in recent decades due to habitat loss and degradation (Rosenberg et al. 2019;Marques et al. 2020).Hence, efforts to stabilize or increase grassland bird populations require identification of species' ecological requirements as a first step (Fisher and Davis 2010).In addition, vegetation structure and spatial distribution in seasonally flooded grassland ecosystems found in wetlands may vary dramatically, depending on the extent of seasonal inundation (Winter et al. 2005).This adds another level of complexity to understanding how each bird species might respond to vegetation variables (Winter et al. 2005), as well as differing levels of human disturbance (Price 2008) and grassland use (Roth et al. 2005;Hossain and Li, 2020;Porensky et al. 2020).For example, seasonally flooded grasslands at the base of the Himalayan foothills and in the floodplains of the Ganges and Brahmaputra rivers in northern India, southern Nepal and Bangladesh (the Indo-Gangetic grasslands) support a distinctive bird community (Baral 2001;Rahmani et al. 2016), with several bird species restricted to this system (BirdLife International 2003, 2015).However, high levels of human use, land clearance for agriculture, overgrazing, and grass harvesting (Baral 2001;Gopi Sundar 2011;Grimmett et al. 2021;Jahan et al. 2022;Douglas et al. 2023) are likely reducing breeding success and survival of birds of this habitat region (Gopi Sundar 2011).
In Bangladesh, the conversion of alluvial soils to cropland due to their high productivity intensifies human pressure on grasslands, posing an ongoing threat to grassland birds.Within Bangladesh, the least disturbed Indo-Gangetic grasslands are now limited to a small number of wetlands in the north-east and to the main rivers bordering the north-west, where seasonal riverine flooding restricts access.Livestock grazing occurs most of the year (October to June).Grass harvesting for feeding livestock occurs yearround at a small scale, and at a larger scale in June, immediately prior to the onset of flooding, when sale of Saccharum sp., especially, serves as an important livelihood source for villagers.Ecologically, Saccharum sp.provides soil binding services, stabilizing riverbeds and reducing bank erosion (Chandran 2015).Although some areas of floodplain grassland in the Indo-Gangetic Plain fall within protected areas in India and Nepal, no government policies have been implemented to protect wet grassland habitats in Bangladesh.Furthermore, there are a lack of data on habitat factors and their impacts on birds in the Indo-Gangetic grasslands that could inform management actions.
The Bristled Grassbird Schoenicola striatus and the Striated Grassbird Megalurus palustris (Family: Locustellidae) are species from closely related genera (Alström et al. 2018), both reliant on parts of the Indo-Gangetic grasslands for part or all of their life cycle (Khan et al. 2015;IUCN Bangladesh 2015;Madge 2020;Jahan et al. 2022).The Striated Grassbird is widely distributed throughout South-east Asia whereas the Bristled Grassbird is endemic to South Asia, and is patchily distributed in Bangladesh, India, Nepal and Pakistan (Del Hoyo and Collar 2016).In Bangladesh, the Striated Grassbird Megalurus palustris toklao is typically present in grassy floodplains throughout the country and is considered to be a common and widespread resident (IUCN Bangladesh 2015).Bristled Grassbird, formerly described as common in parts of Bangladesh in the nineteenth century (Ali and Ripley 1987), went un-recorded in the country for over a century.It was rediscovered in northeast Bangladesh in 2014, and is considered to be rare (Haque and Tareq 2014;Khan et al. 2015).However, a recent study found that Bristled Grassbird was more widespread along major rivers than previously recognized (Chowdhury et al. 2022).Both species breed during the summer and early monsoon season (March-July) (Khan et al. 2015;Jahan et al. 2022).Bristled Grassbird inhabits tall grassland, particularly Saccharum spontaneum dominated grasslands and initial data suggests that it is probably negatively impacted by grass harvest and increasing agricultural intensification (Baral et al. 2014;Singh and Buckingham 2016;Chowdhury et al. 2022).While Striated Grassbird is a highly visible, year-round resident (IUCN Bangladesh 2015), outside the breeding season Bristled Grassbird remains largely unrecorded probably due to its secretive behaviour when not singing and limited survey effort during the nonbreeding (mid-late monsoon) season.Both species can be regarded as grassland specialists since both are adapted to and reliant on grassland habitats for part or all of their life cycle, whether feeding or breeding (Jahan et al. 2022).However, Bristled Grassbird is an obligate tall grass breeder, and entirely dependent on tall grasses for breeding (Singh and Buckingham 2016;Chowdhury et al. 2022;Jahan et al. 2022).There are limited data on micro-habitat preferences of either species.In particular, the very abundance and widespread distribution of the Striated Grassbird may have discouraged study of its detailed habitat use.Globally, Bristled Grassbird is considered to be threatened (Vulnerable), while Striated Grassbird is regarded as common and widespread, with status Least Concern throughout its range (BirdLife International 2022a, b).Our paper seeks to answer which habitat variables and levels of human disturbance might explain the primary differences in status and habitat use by these two ostensibly similar species, and whether insights gained from this species pair might be used to inform grassland bird conservation in wetlands, perhaps mitigating threats to the wider bird community (Han et al. 2020).
Here we tested the hypothesis that Bristled Grassbird and Striated Grassbird differ in their vegetation structural preferences and responses to habitat disturbance.We predicted that Bristled Grassbird would prefer contiguous taller, homogeneous floodplain grassland cover, principally Saccharum sp.(Baral et al. 2014;Singh and Buckingham 2016;Chowdhury et al. 2022;Jahan et al. 2022) and would respond negatively to increased human disturbance (Singh and Buckingham 2016;Chowdhury et al. 2022).Striated Grassbird, on the other hand would occupy more heterogeneous grassland cover (Ali and Ripley 1987;Round et al. 2014).Because of its possible preference for heterogeneous landscapes, Striated Grassbird should respond positively to increased, but moderate, human disturbance.Consequently, we predicted that the Striated Grassbird would be more abundant than Bristled Grassbird (IUCN Bangladesh 2015).

Study sites
The study islands (locally known as chars) were located in the Ganges-Padma River floodplain of Rajshahi (Char Majardia) and in the Brahmaputra-Jamuna River floodplain of Kurigram (Char Sakahati), Bangladesh and are referred to hereafter as Rajshahi and Kurigram (Fig. 1).The total area of the Rajshahi site covered 19.4 km 2 , and was approximately 1.2 km from the mainland, while the Kurigram site covered 35.2 km 2 and was about 3.4 km from the mainland, based on 2018 Google Earth data.The two sample islands are about 170 km apart.The study areas were chosen to represent conditions along the two rivers and particularly for the presence of Bristled Grassbird (Jahan et al. 2022).These sites are roughly near the eastern edge of the Bristled Grassbird range and roughly in the center of the mainland Asian range for Striated Grassbird (BirdLife International 2022a, b).Both study areas are subject to seasonal flooding during the monsoon (July-October).During the bird breeding season, when surveys were conducted (May-July), wet, inundated areas were scarce.Land-cover of both areas was characterized by floodplain grasslands containing a mosaic of patches of both short grasses (Cynodon sp.) and tall grasses (Saccharum sp.), with a mix of forbs and bushes.Saccharum sp.mostly grows up to 3 m tall and occur in dense and more extensive patches.Forbs and bushes cover included Solanum sp., Tamarix sp., Ipomoea carnea, Phyla nodiflora, Persicaria sp. and Xanthium strumarium, and were usually mixed with grasses (Table S1).Of the two woody species, neither of which is native, Tamarix sp.grows as small to extensive patches of dense shrubby single-species cover of relatively even height (up to 5 m) which excludes grasses.Tamarix sp.therefore lacks protruding emergent perches.The long established but exotic Ipomoea carnea (Akter and Zuberi 2009) has a different growth form, more straggling with taller emergent woody perches, also of up to 5 m.Furthermore, it usually occurs in patches or clumps amid other vegetation and is often mixed with grasses.
Based on 80 vegetation samples for each survey point (see below in Habitat measurements), Ipomoea carnea was more abundant in Kurigram (0.7% land cover) than in Rajshahi (0.08% land cover) while Tamarix sp. was more abundant in Rajshahi (4.8% land cover) than in Kurigram (0.9% land cover).Both Bristled Grassbird and Striated Grassbird were observed using Saccharum sp., Tamarix sp. and trees (> 7 m height) as their song-post perching substrates, but only Striated Grassbird used Ipomoea carnea and other forbs (> 3 m height) as song-posts (Fig. 2).
Croplands and small villages were also characteristic of the landscape.Differences in crop varieties were noted between islands.At Kurigram site a wider range of crops were planted including rice, lentils, jute, maize, sesame, dhaincha (Sesbania bispinosa), berries (Ziziphus mauritiana), pointed gourd, bitter gourd, chili, and other vegetables, while at Rajshahi site mostly rice was planted with small areas of sesame and jute.The Kurigram site supported year-round human occupancy, in villages located on the island, while the Rajshahi site had no human occupants within the study area, although a village named Nobinagar was located at the south edge of the island, adjacent to the study area, close to the border with India (part of the island is located in India, Fig. 1).Livestock grazing (cows Bos indicus, goats Capra hircus, sheep Ovis aries, and water buffaloes Bubalus bubalis), and grass (Saccharum sp.) harvesting by local people occurred at both sites before the onset of the monsoon.Grazing (blue green), various forbs and bushes including mixed vegetation with grasses (light green), cultivated crops (yellow), and an additional 12 points were surveyed in cropland in 2019 (Black dots).Second set of map of 2019 showed areas covered with grazing (grey), harvesting (orange) and both graz-ing and harvesting (blue).Percentage area of land cover is also presented.Bristled Grassbird and Striated Grassbird observations overlaid on the survey points.(Map created using ArcGIS 10.3).In 2018, Kurigram was surveyed at 91 and Rajshahi was surveyed at 94 points.In 2019, Kurigram was surveyed at 91 and Rajshahi was surveyed at 94 points plus 12 points purposefully added in cropland.These counts consisted of total 382 sample points which covered 619 ha in total on the two islands intensity was observed to be higher in Rajshahi compared to Kurigram (Fig. 1).

Survey methods
Abundance data for both grassbird species were collected during their breeding season, when birds were singing and most easily detected.Point-count surveys were conducted from 21 May to 2 July in 2018 and from 22 May to 16 July in 2019.A total of 382 pointcounts were conducted (185 points in 2018 and 197 in 2019) by overlaying a grid of 250 × 250 m on 2018 Google Earth imagery of the study islands.The grid size was decided based on field surveys to minimize double counting the same birds in adjacent points.Based on image interpretation and field checks, grid points that fell on water, bare sand, woodland, forest plantations and human settlements were excluded, and the remaining points were in grassland patches and other open-country vegetation (bushes, reed, swamp, scrub, cultivation).Sample points were chosen randomly from among the sample grid, and within a chosen grid cell, sample points were always placed at the southwest corner (Fig. S1).All sample points were ≥ 250 m apart.One observer (IJ) recorded individual Bristled Grassbirds and Striated Grassbirds seen or heard within 100-m fixed-radius point-counts (Hutto et al. 1986) of 10-min duration without a waiting period (Savard and Hooper 1995).Both species were noted as either heard, or seen in song-flight, or perching.Differences in song flight behaviour of both species were also recorded.Distances between the observer and birds were estimated directly using a rangefinder when birds were visible, otherwise in 5 m bands to 30 m, then 10 m bands to 100 m (Rosenstock et al. 2002).With points spaced 250 m apart, and birds only counted to a distance of 100 m at each point, the 50 m buffer minimized the likelihood of double-counting the same individual.Moreover, the observer also carefully noted the locations of individual birds seen at adjacent points to further avoid double-counting the same individuals.Surveys took place between 06 h and 10 h under good weather conditions (i.e., with no rain or strong winds).Weather conditions (wind and cloud cover) were also recorded.Wind velocity was recorded following the Beaufort scale and cloud cover was estimated by visual percentage approximation; survey time and duration were also recorded.Rajshahi was surveyed at 94 points which were sampled in each of the two survey years and an additional 12 points were surveyed in cropland in 2019 (Fig. 1).Kurigram was surveyed at 91 points which were also sampled twice (once in each survey year).In total, the detection area of the survey points covered about 17% of the Rajshahi site and 8% of the Kurigram site.

Habitat measurements
In 2018 and 2019, plant species and height were recorded at 5 m intervals along a 100 m radius from each bird survey point in each of the four cardinal directions (Jahan et al. 2022).Vegetation was sampled using an ocular tube (a simple sighting device consisting of a cylinder with crosshairs following James and Shugart, 1970).At each interval point the single dominant vegetation species seen through the ocular tube when pointed downwards from 100 cm above the ground was recorded.This gave 80 vegetation samples per bird survey point.The plants were classified into four categories: Saccharum sp., Cynodon sp., forbs and bushes, or cropland.Bird survey points in which one of these four vegetation types contributed ≥ 80% cover of vegetation cover was labelled accordingly.Forbs and bushes, including mixed vegetation with grasses, were also classified as "Forbs and bushes" when they covered < 80% of the area.Forbs and bushes included Solanum sp., Tamarix sp., Ipomoea carnea, Phyla nodiflora, Persicaria sp. and Xanthium strumarium, among others, and were usually mixed with grasses.Ipomoea carnea and Tamarix sp., were taller and woody, were classed within the forbs and bushes category, and were additionally classified as song-posts.Vegetation species were identified by comparing voucher specimens from a herbarium with photos taken during the survey.All non-crop and crop vegetation stem heights were measured within a circle of 50 cm radius and averaged.These measurements were repeated for 20 such plots at 5 m intervals in each of 4 cardinal directions.The 80 values of height obtained were then themselves averaged for each sample point.This average height per sample point was later used for bird-habitat association.This method was adopted as a time-efficient way of characterizing vegetation after testing several alternative methods.Vegetation heterogeneity was quantified based on the extent of different vegetation heights recorded at each bird survey point.Vegetation heterogeneity was summarized by calculating the coefficient of variance (CV) of the vegetation height measured for each bird survey plot.The coefficient of variance, obtained by dividing the standard deviation of vegetation height by the mean height from 80 vegetation measures for each bird survey point, served as a standardized measure of heterogeneity (Sliwinski et al. 2020).Vegetation heterogeneity was therefore conceptualized as the extent to which short vegetation and tussocks of taller vegetation occurred in proximity.For example, areas with a mix of both short and tall vegetation were more heterogeneous than areas with only tall or only short vegetation.Human use variables (grazing and harvesting) were measured only in 2019.Grazing intensity was based on a count of livestock dung (Sheidai-Karkaj et al. 2022) and grass harvesting intensity was based on counts of cut grass stems within a 1.5 m radius, at 5 m intervals along the four cardinal direction sampling transects.Some changes in habitat were observed between 2018 and 2019: four points previously covered with Saccharum sp. or forbs and bushes in 2018 were bare land in 2019, and the additional five points supporting either Saccharum sp. or forbs and bushes in 2018 were inundated in 2019.Crop cultivation increased in 2019 (29 points) compared to 2018 (12 points).This comparison does not include an additional 12 points which were surveyed in cropland in 2019 in order to increase the sample of cultivated points.

Vegetation associations
Both male and female Bristled Grassbirds and Striated Grassbirds were recorded at each sample point.However, the females of both species were relatively cryptic, whereas males were more readily detected, particularly in song flight, and when perched and singing.All the aural detections from song were also likely from males.Therefore, models assessing vegetation associations and density estimates were performed for only males in both species.To evaluate the occurrence of singing males in 100-m point counts of each species relative to habitat variables, we used binomial regression models in a generalized linear models (GLM) framework, using R package glmmTMB (Brooks et al. 2017).Models were fitted as a binary response variable: detected = 1 and not detected = 0. Model sets were developed to conduct vegetation structural preference analysis and the effects of grazing and harvesting for Bristled Grassbird and Striated Grassbird combining both study islands and both years.We used the same set of randomly sampled points to replicate the survey in both years (apart from the 12 additional points in cropland in 2019).However, the impact of the differing extent of flooding and changes in vegetation on bird usage during the two years meant that these sets could reasonably be considered to be independent of one another.Two sets of models were developed: the first set of models included vegetation type with vegetation heterogeneity, vegetation height and song-post counts, using data recorded in 2018 and 2019.The second set of models added grazing and harvesting intensity, these data were recorded in 2019 only.Vegetation categories were Saccharum sp., Cynodon sp., forbs and bushes, and cultivated crops.Use of tall, woody song-posts (Ipomoea carnea and Tamarix sp.) was analysed for both grassbird species.Both model sets contained each variable and combination of variables, in addition to null and global models.Collinearity among predictor variables was checked before analysis.Variables with correlations above 0.70 were not included in the same model per Dormann et al. (2013).AIC model comparison methods (Burnham and Anderson 2002) were used to evaluate the relative ability of each model to explain variation in occurrence for Bristled Grassbird and Striated Grassbird.All models were checked for the homogeneity of their residuals.Models were compared and ranked using Akaike's information criterion corrected for small sample size (AICc) following Burnham and Anderson (2002).Models and their relative importance were also assessed using Arnold (2010) to avoid the inclusion of uninformative parameters.AICc weights (w) provided the relative weight of any particular model (Burnham and Anderson 2002).Model coefficients (β) with 95% confidence intervals (CI) from AICc selected models (Burnham and Anderson 2002; Arnold 2010) were reported, and predictor variables were considered to be strongly influential on response variables if the CI for β did not overlap zero (Bolker 2008;Arnold 2010).We also used model-averaged coefficients to make multi-model inferences when model-selection uncertainty was high (Burnham and Anderson 2002).All statistical analyses were carried out in the Program R version 4.1.0(R Core Team 2020) using glmmTMB, tidyverse, ggpubr, AICcmodavg, visreg packages.
Other habitat variables tested included the presence of standing water within 100 m of the survey point (an indicator of 'wet' habitat) as a rough test of a hypothesis regarding the relative preferences in the the two grassbird species for wetter versus drier habitats.Non-metric multi-dimensional scaling (NMDS) was used to make more quantitative interpretations about the vegetation differences between the two study islands.For the NMDS analysis, all crop varieties were treated as one category named "crop".

Grassbird density
Distance sampling was applied to estimate bird densities using Program DISTANCE (Buckland et al. 2001).Four estimators: uniform, half normal, hazard-rate, and negative exponential were employed to determine the detection model of best fit.Hazard-rate models were not considered when the output gave implausible shapes (Thomas et al. 2010).All detection models were run with the cosine adjustment.Models were evaluated using AICc (Buckland et al. 2001).The detection model with the lowest AICc score was considered the best fit and was used to estimate density of individuals within that specific habitat category (McCollum et al. 2018).The post-stratification function was used to estimate year-wise, and study-area-wise densities.Kolmogorov-Smirnov goodness-of-fit (GOF) tests were used to evaluate models with models having P > 0.05 considered well-fitted to the data (Buckland et al. 2001).

Detections
In 2018 and 2019 Bristled Grassbird was found at both study sites, although Striated Grassbird was primarily found at Kurigram and was largely absent from Rajshahi (Table 1).Bristled Grassbirds were mostly (n = 69) detected in song-flight (displaying males), and less often (n = 42) perching on Saccharum sp. (Fig. 2).Most Striated Grassbird detections were of birds singing from song-posts (n = 81).They mostly used Ipomoea carnea and other woody perches as their perching substrate (Fig. 2).Displaying Bristled Grassbird males made long duration song-flights (2-3 min) high (~ 15-21 m) in the air, returning to perch in tall Saccharum sp. for a period of about 30-50 s.Striated Grassbird males, on the other hand, sang and called while perched on songposts, making short period (< 1 min) song-flights during a short ascent and a "parachute" descent to the same or similar nearby perch.

Habitat
Vegetation species diversity differences (excluding differing crop types) were generally low between the two study areas (Fig. S2), but changes in dominant vegetation and land cover types were observed between years (Fig. 1).In Kurigram a near tripling of crop area (from 11.0% in 2018 to 30.8% in 2019) Vol:. ( 1234567890) resulted in a concomitant decrease in the area of Saccharum sp.(from 23.1% in 2018 to 14.3% in 2019) and a reduction in the area of forbs and bushes mixed with other grasses (from 65.9% in 2018 to 53.8% in 2019).In Rajshahi there was no increase in crop cover between years, but the area of forbs and bushes decreased considerably, the area of Saccharum sp.cover increased (from 14.9% in 2018 to 19.8% in 2019) and the area covered by short grass, Cynodon sp., increased (from 5.3% in 2018 to 14.2% in 2019), associated with grazing and harvesting.Both the survey islands were used by local people for cattle grazing (34.1% of the survey area in Kurigram and 53.8% in Rajshahi) and grass harvesting (23.1% in Kurigram and 11.3% in Rajshahi).

Habitat associations
Model results for Bristled Grassbird (both study areas combined) ranked Saccharum sp. as the top model (Table 2) indicating that the occurrence of male Bristled Grassbirds increased with increasing Saccharum sp.cover (Fig. 3).For Bristled Grassbird there was a positive, statistically significant, beta coefficient for Saccharum sp.; crop cover and vegetation heterogeneity had significantly negative beta coefficients, while vegetation height had a significantly positive coefficient (Fig. 4).
For Striated Grassbird, vegetation height was ranked as the top model (Table 2) with a significantly positive coefficient between height and occurrence; but no clear association with Saccharum sp.(Figs. 3 and 4).Another indicator of vegetation height, the cover of Cynodon sp.(an indicator of short vegetation) also had a significant negative beta coefficent indicating that the occurrence of Striated Grassbird decreased with increasing Cynodon sp.cover (Figs. 3 and 4).Crop cover had a significant positive beta coefficent indicating a positive association between Striated Grassbird and crop cover.Considering the availability of woody song-posts as potentially important, occurrence of Striated Grassbird was found to be strongly positively correlated with Ipomoea carnea cover (β = 0.435, 95% CI = 0.202 to 0.668) (Fig. 5).Ipomoea carnea was much more abundant at Kurigram than at Rajshahi.Bristled Grassbirds showed no association with either Ipomoea carnea or Tamarix sp (Fig. 5).

Effects of grazing and harvesting
Neither grazing nor harvesting intensity had any significant effect on the presence of Bristled Grassbird when combining data from both study areas in 2019.For Striated Grassbird, harvesting appeared to have a significant positive effect, while no impact from grazing could be detected (Figs. 3 and 4, Table 2).

Density
In the site where both species were present (Kurigram), the density of male Striated Grassbird (0.32/ ha in 2018 and 0.72/ha in 2019) was approximately three times higher that of male Bristled Grassbird in 2018 (0.10/ha) and four times the density in 2019 (0.13/ha) (Table 1).Bristled Grassbird densities were not much different between islands in 2018 but were higher at Rajshahi than Kurigram in 2019 (Table 1).
In between years, a greater density of Bristled Grassbird males was detected at Rajshahi in 2019 (0.26/ ha) vs. 2018 (0.11/ha) (Table 1).No other greater differences in density were observed between years for either species.

Bird-vegetation structure associations
Our results suggested that Bristled Grassbird and Striated Grassbird have similar structural preferences for taller grassland vegetation, although Bristled Grassbird had a clear preference for Saccharum, whereas Striated Grassbird did not.In addition, a key response to cropland was contrasting, with Bristled Grassbird avoiding cropland, while Striated Grassbird was positively correlated with it.
The strong association of Bristled Grassbird with Saccharum sp.abundance matches well with previous studies (Baral et al. 2014;Singh and Buckingham 2016;Chowdhury et al. 2022;Jahan et al. 2022).It was also significantly positively associated with taller vegetation in general.Furthermore, Bristled Grassbird appeared to be negatively influenced by agriculture, avoiding cropland.The negative coefficient associated with heterogeneity further indicated this species was likely associated with more continuous cover of Saccharum, but its avoidance of cropland may be the largest reason for concern over its conservation status.While Striated Grassbird similarly tended to be associated with taller grassland vegetation, it was more general in its preference, and showed much greater tolerance of cropland.Our results also indicate that Striated Grassbird benefits from the presence of the woody exotic Ipomoea carnea which thrives in more disturbed areas, and is used as sturdy song-perches in somewhat contrast to the Bristled Grassbird, which largely makes aerial song-flights and appears not to require such perches.Part of the explanation for differences between the two species fortunes in the two  3 study areas appears to be the dependence of Bristled Grassbird on Saccharum sp.(hence when the area of tall grass increased between 2018 and 2019 in Rajshahi so did the population of Bristled Grassbird), compared with the preference of Striated Grassbird for tall crops and woody vegetation (which are common and increasing in Kurigram but rare in Rajshahi).
Observed differences in display behaviour between the species are most likely relevant to their habitat use.Notably Striated Grassbird sang from prominent perches as a component of its display, which involved only a short display flight, whereas Bristled Grassbird undertook much more sustained, high-soaring songflights (Grimmett et al. 2021).Whether their different song-flight performances are related to territorial defence or mate attraction (or both), differences in display and singing habits of Bristled Grassbird and Striated Grassbird, in particular their differential use of song-posts, may contribute to the major differences in their detailed patterns of habitat use and occupancy within similar grassland habitats.

Impacts of human use
Our results support the hypothesis that the two species are affected differently by human use (agriculture, livestock grazing and/or grass harvesting intensity).Bristled Grassbird presence was negatively correlated with cropland as found in other studies in both Bangladesh (Chowdhury et al. 2022) and Nepal (Singh and Buckingham 2016).On the other hand, the Striated Grassbird had a significantly positive association with cropland indicating an ability to utilize a wider range of vegetation.
There was no clear indication that grazing or harvesting affected Bristled Grassbird.However, Striated Grassbird was positively associated with grass harvesting which points to its avoidance of homogeneous grassland, and association with more disturbed or mosaic habitats.The evidence from our study supports the generalization that grassland bird species respond differently to harvesting and grazing practices and intensities (Coppedge et al. 2008;Kim et al. 2008;Sliwinski and Koper 2015).The Striated Grassbird is much more of a generalist in grassland habitat-use compared to the Bristled Grassbird in floodplain grassland, but our study revealed its songpost requirements.

Density
The estimated density of male Bristled Grassbird was similar on both study islands in 2018: 0.10/ha at Kurigram and 0.11/ha at Rajshahi (Table 1).However, in 2019 density was higher at Rajshahi (0.26/ha) than at Kurigram (0.13/ha).Nonetheless, although there was In both years, the density of Striated Grassbirds at Kurigram (0.32/ha in 2018 and 0.72/ha in 2019) was three to four times greater than the density of Bristled Grassbird (Table 1), indicating the much greater ecological tolerance of this wide-ranging and relatively abundant species compared to Bristled Grassbird.The Striated Grassbird was largely absent from the grassland habitat of our sampled areas in Rajshahi.No specific association of Striated Grassbird with either Saccharum sp. or forbs and bushes was evident.Instead, a generalized preference for habitat mosaics, exemplified by a significant positive association with cropland and with Ipomoea carnea, a colonist of disturbed habitats, much more abundant at Kurigram, explains its high density at the latter site, and its absence from the much less disturbed habitat at Rajshahi.The proportion of cropland was higher in Kurigram compared to Rajshahi in both years of study.However, the population of the Striated Grassbird may decline as agricultural intensification progresses (Douglas et al. 2023).The occurrence of the Striated Grassbird also decreased as the cover of Cynodon sp.increased.Cynodon sp.(indicating short vegetation) was more prevalent at Rajshahi compared to Kurigram in both years.The positive association of Striated Grassbird with grass harvesting also indicated its avoidance of homogeneous tall grassland.Saccharum sp.cover at Rajshahi actually increased during the study period, (from 14.9% in 2018 to 19.8% in 2019) while at Kurigram a substantial increase in crop area (from 11.0% in 2018 to 30.8% in 2019) coincided with a decrease in the area covered by Saccharum sp. and mixed forbs and bushes with grass in 2019.There was no increase in crop cover at Rajshahi during 2018 to 2019.Overall, our data indicated that Striated Grassbird preferred more disturbed, hererogeneous vegetation species mosaic habitats in which tall woody perches were present among open and disturbed areas, but further study of other sites where it is present could reveal whether it is more specialized than previously thought.

Conclusions
As the micro-habitat requirements of both grassbird species differ, even though they occupy similar floodplain grassland habitats, so habitat management plans must take these differences into account, with greater emphasis on the scarcer habitat required by the more threatened species.For Bristled Grassbird in Bangladesh, Chowdhury et al. ( 2022) recommended protection from grass cutting and livestock grazing.However, we were unable to demonstrate any clear significant negative impact from moderate levels of either activity on Bristled Grassbird in either study area.Presumably intensive over-grazing and overharvesting would be a problem.However, patches of tall Saccharum sp.take several years to develop and therefore need protection from outright conversion to cropland.They likely can sustain partial harvesting and grazing, providing neither activity is too intensive.The timing of these activities would also probably be relevant for sustainable management.Importantly, disturbance from grass-cutting during the nesting season could be a source of nesting failure in Bristled Grassbird.This is not something on which we have data, and the detailed response of the species across a range of harvesting and grazing regimes would reward further study.Without disturbance from either seasonal inundation, harvesting or grazing, ecological succession might over time lead to Saccharum being replaced by woody shrubs.Furthermore, multiple uses and /or natural or human instances of disturbance are more likely to lead to maintenance of diverse habitat mosaics (e.g., grassland patches of different ages and heights) which might benefit overall community diversity.
Given that the most likely alternative to the present seasonal use of riverine floodplains for cattle grazing and harvesting of grasses is outright conversion to croplands, we would argue in favour of maintaining the present moderate use regime.Ideally future research should sample grassbirds and wider bird community diversity across a gradient of harvesting and grazing regimes so as to refine our knowledge.This, with care, would help contribute to the sustainable use and management of grasslands, benefiting the conservation of both grassbird species and the livelihoods of local people.Achieving total protection of such wetland grasslands would be a major challenge in river islands that are remote from any conservation authorities, have long established patterns of human use, and are subject to frequent change due to floods and erosion nearly every year.Strategies to limit the extent of agriculture, diversify crops on pre-existing cultivated areas and ensure that other bird-friendly practices such as minimizing pesticide use are adopted would be appropriate to conserve both species (Chandran 2015).In addition, leaving some farm fields fallow each year would be a good additional measure and possible compromise for local farmers.That would help to provide links for all grassland dependent birds to use larger areas and to move between patches of tall grasses, by allowing medium height grasses and forbs to grow, in this way providing usable habitat for a number of grassland bird species of management concern.This management approach will require cooperation among conservationists, local governments and villagers to develop local plans to zone river islands and regulate the uses in different zones to achieve a reasonable balance between bird conservation and human livelihoods.This will also necessitate negotiation and an adaptive approach to monitor and adjust practices based on longer-term monitoring of target wet grassland dependent species.

Fig. 1
Fig. 1 Locations and maps of the two survey areas in the Ganges and Brahmaputra Rivers, Bangladesh with survey points.Multiple colored dots represent areas (ha) covered with different vegetation types: Saccharum sp.(dark green), Cynodon sp.(blue green), various forbs and bushes including mixed vegetation with grasses (light green), cultivated crops (yellow), and an additional 12 points were surveyed in cropland in 2019 (Black dots).Second set of map of 2019 showed areas covered with grazing (grey), harvesting (orange) and both graz-

Fig. 2
Fig. 2 Percentage of observational differences between Bristled Grassbird and Striated Grassbird during point counts (left A: behaviour at detection, right B: substrate used by those seen

Fig. 3
Fig. 3 Relationship between the probability of presence (occurrence) of Bristled Grassbird and Striated Grassbird and different habitat variables at floodplain grassland at two sites combined in the lower Ganges and Brahmaputra Rivers, Bangladesh, based on data recorded in 2018 and 2019.Habitat variables were standardized.Vegetation type was measured as the total number of times (out of 80 samples) each vegeta-

Fig. 4
Fig. 4 Estimated coefficients of each habitat variable for the probability of presence of Bristled Grassbird and Striated Grassbird in floodplain grassland at two sites combined in the lower Ganges and Brahmaputra Rivers, Bangladesh, based on data recorded in 2018 and 2019.Points are the estimated beta coefficients, and horizontal lines present 95% confidence

Fig. 5
Fig. 5 Relationship between the presence (occurrence) of Bristled Grassbird and Striated Grassbird and song-posts (Ipomoea carnea and Tamarix sp.).Plots with an asterisk indicate significant regression coefficients

Table 1
Densities of Bristled Grassbird and Striated Grassbird in the sample floodplain grassland at two study sites All selected detection models were run with cosine adjustment.A total of 197 sample points were surveyed (total effort of 377 point counts), which covered 619 ha in total on the two islands: 286 ha in Kurigram (91 points) while Rajshahi covered 295 ha in 2018 (94 points) and 333 ha in 2019 (106 points).Values reported are: number of birds (M male, F female), number of points detected, estimated density (birds/ha) with 95% confidence intervals (95% CI)

Table 2
Model sets for predicting vegetation structural associations and the effects of human use on the occurrence of Bristled Grassbird and Striated Grassbird in floodplain grasslands of the lower Ganges and Brahmaputra Rivers, data recorded in 2018 and 2019

Table 3
Mean and standard deviation (SD) of habitat variables sampled in 2018 and 2019 combining both islands Total 185 points surveyed in 2018: 91 points at Kurigram and 94 points at Rajshahi.Total 197 points were surveyed in 2019: 91 points at Kurigram and 106 points at Rajshahi.Stem count/point is the number of stems of this vegetation type/species out of a total of 80 stem counts per point.Vegetation height is the average of 80 measures per point.Vegetation height heterogeneity is the coefficient of variation (CV) of averaged height per point.Droppings and cut grass are the number of spots where this was found out of 80 spots per point