Advertisement

Journal of Ornithology

, Volume 159, Issue 3, pp 851–866 | Cite as

Distribution, movements, and survival of the critically endangered Bengal Florican Houbaropsis bengalensis in India and Nepal

  • Rohit R. S. JhaEmail author
  • Jyotendra Jyu Thakuri
  • Asad R. Rahmani
  • Maheshwar Dhakal
  • Ngulkholal Khongsai
  • Narendra Man Babu Pradhan
  • Nikhil Shinde
  • Bridesh Kumar Chauhan
  • Rahul K. Talegaonkar
  • Ian P. Barber
  • Graeme M. Buchanan
  • Toby H. Galligan
  • Paul F. Donald
Original Article

Abstract

We undertook field surveys of the critically endangered Bengal Florican throughout its range in India and Nepal and used the results to develop models of distribution to identify new populations. We also tagged 11 birds with satellite transmitters to assess their distribution and habitat use during the non-breeding season, about which nothing is known. The models suggest that the distribution is now extremely fragmented in the west, although potentially sizeable populations may await discovery in eastern parts of the species’ range, particularly along the Brahmaputra River. After breeding, tagged birds left their breeding sites inside protected areas (PAs) and moved up to 80 km into landscapes characterised by relatively low-intensity agriculture along rivers with lower than average human population densities. Many birds spent more than six months away from their breeding grounds outside PAs. Models of non-breeding distribution suggest that some breeding areas become unsuitable at that time of the year, and therefore that birds may be forced to leave PAs at that time. Tagged birds had high annual adult survival rates (0.92, 95% confidence interval 0.858–0.953). Home ranges during the non-breeding season were significantly larger than breeding season home ranges. The results suggest that the conservation of this species needs to account for the species’ use outside the breeding season of less intensively cultivated floodplain-agricultural landscapes, as well as protecting breeding grassland sites.

Keywords

Subtropical grassland Bird conservation Bustard Species distribution models Terai Satellite telemetry 

Zusammenfassung

Verbreitung, Bewegungsmuster und Überleben der stark gefährdeten Barttrappe ( Houbaropsis bengalensis ) in Indien und Nepal

Wir untersuchten die sehr stark gefährdete Barttrappe in ihrem gesamten Verbreitungsgebiet in Indien und Nepal und entwickelten anhand der Ergebnisse Verbreitungsmodelle, die Aufschluss über neue Populationen geben könnten. Außerdem versahen wir 11 Vögel mit Satellitensendern, um mehr Informationen über ihre Verbreitung und Habitatnutzung außerhalb der Brutsaison zu bekommen, da hierüber bislang nichts bekannt ist. Unsere Modelle lassen vermuten, dass die Verbreitung im Westen zur Zeit extrem fragmentiert ist, wobei es im Osten des Verbreitungsgebiets der Barttrappe, vor allem entlang des Brahmaputra-Flusses, möglicherweise noch potentiell ausreichend große Populationen gibt. Nach der Brutzeit verließen die mit Sendern versehenen Vögel ihre in Naturschutzgebieten liegenden Brutgebiete und wanderten bis zu 80 km weit in Flussgebiete mit relativ wenig Landwirtschaft und besonders geringer menschlicher Besiedlung. Viele Vögel verbrachten mehr als sechs Monate außerhalb der Naturschutzgebiete, fernab ihrer Brutgebiete. Modelle der Verbreitungsmuster außerhalb der Brutzeit lassen vermuten, dass einige der Brutgebiete zu der Jahreszeit ungeeignet sind und die Vögel deshalb gezwungen sein könnten, die Naturschutzgebiete in dieser Jahreszeit zu verlassen. Die markierten Vögel zeigten für Adulte hohe jährliche Überlebensraten (0,92; 95% CI 0,858–0,953). Der normalerweise genutzte Lebensraum war außerhalb der Brutzeit signifikant größer als während der Brutzeit. Unsere Ergebnisse legen nahe, dass neben dem Schutz der Wiesen und Weiden während der Brutzeit für den Schutz dieser Art auch ihre Nutzung der weniger intensiv kultivierten Flussauen sowie landwirtschaftlich genutzter Flächen außerhalb der Brutzeit berücksichtigt werden sollten.

Introduction

The Bengal Florican Houbaropsis bengalensis is a threatened bustard of South and Southeast Asia. It has a peculiarly disjunct distribution, with one centre of population in the Tonle Sap of Cambodia (subspecies Houbaropsis bengalensis blandini), and another more widely spread in the alluvial grasslands (the Terai) of southern Nepal and northern India along the foothills of the Himalayas, and the Brahmaputra Plain in northeastern India (subspecies Houbaropsis bengalensis bengalensis) (Rahmani et al. 1991; Collar et al. 2001; Rahmani 2012; Donald et al. 2013). Both populations are fragmented and declining, probably as a result of the loss of the species’ grassland habitats to agriculture, grassland degradation due to poor management and overgrazing, and maybe hunting and collisions with power lines (Inskipp and Collar 1984; Baral et al. 2003; Poudyal et al. 2008a, b; Gray et al. 2009; Packman et al. 2013, 2014; Mahood et al. 2018). As a result, the species is listed by the International Union for Conservation of Nature as critically endangered (BirdLife International 2017).

Developing conservation solutions for both populations is hampered by a lack of understanding of the species’ ecology and distribution. During the breeding season, males are conspicuous by their display, but at other times of the year Bengal Floricans effectively disappear. An analysis of sightings and museum specimens of the species prior to 2001 yielded virtually no records of the species of either population between June and December (Collar et al. 2001; Donald et al. 2013). Furthermore, most of the records from outside these months came from a small number of relatively well-known sites within the species’ much larger area of distribution (Rahmani et al. 1991; Baral et al. 2003). It is possible that potentially substantial populations may remain undiscovered, particularly along large rivers such as the Brahmaputra, where flooding events regularly reshape the landscape (Dwivedi 2016). Hydrological alterations and changes in grassland condition can bring about rapid changes in the numbers of Bengal Floricans, as at Koshi Tappu Wildlife Reserve in Nepal, where a significant increase was recorded between 1982 and 2011 (Baral et al. 2012).

The population at Tonle Sap has been the focus of an intensive study of the species’ seasonal movements using data from birds tagged with satellite transmitters (Packman 2011). This work showed that the Cambodian population is forced by seasonal flooding to leave breeding sites in open grassland and move up to 42 km to grassy and lightly wooded habitat. Birds avoided agricultural areas, which are mostly rice fields, and closed forest. While a similar displacement due to flooding is expected for birds in northeast India where the Brahmaputra River regularly inundates a large swathe of land during the monsoon, nothing to date is known of the movements or non-breeding habitat use of the populations in the Terai of northwestern India and southern Nepal. The Terai, being well drained, does not generally undergo prolonged monsoonal inundation.

In this study, we modelled the breeding range of the species in the Terai of India and Nepal, in order to evaluate the extent to which known populations are likely to be fragmented and also to identify areas that might hold previously unknown populations. We assessed seasonal movements of the species so that conservation policies can be developed that encompass the species’ entire life cycle. We used satellite tracking data to estimate adult survival rates and home ranges, in order to assess likely demographic causes of decline and to quantify the amount of habitat needed by each individual.

Methods

Field surveys

In 2013, we collated data on recent and historical sightings of the species throughout India and Nepal to target field surveys. Field surveys were conducted across the species’ range in India and Nepal from 2013 to 2016 inclusive during the Bengal Florican’s breeding season (February–June, although some males are seen displaying up to and including early July) (Fig. 1). While most surveys were done in grass-dominated habitats, surveys in the Koklabari area (n = 10) near the Manas Tiger Reserve in Assam (India) covered an active seed farm with little or no grass [Electronic supplementary material (ESM) 1, Fig. S1]. A small proportion of all surveys (n = 162 or 17%) were of grasslands containing moderately dense scattered trees, mostly commercial species such as eucalyptus Eucalyptus tereticornis, sal Shorea robusta and teak Tectona grandis in Uttar Pradesh (India) as a result of tree plantation activities undertaken by the local Forest Department in the past (ESM 1, Fig. S2). Since the Bengal Florican shares its grassland habitat with many potentially dangerous large mammals, such as the Asian Elephant Elephas maximus, Greater One-horned Rhinoceros Rhinoceros unicornis, Wild Buffalo Bubalus arnee and Tiger Panthera tigris, surveys were done from safe vantage points, such as observation towers (machaans), vehicle rooftops, the forest department’s elevated camps and treetops. Each survey session lasted for 1–3 h during the early morning (slightly before sunrise onward) and late evening (till slightly after sunset). At Koshi Tappu Wildlife Reserve, sweep counts were done by groups of observers to detect Bengal Floricans. At each grassland location, teams of one to four observers who had previously been trained in Florican identification and survey protocols continuously scanned for the presence of Bengal Florican. Upon each sighting of a bird, its distance from the observer was recorded by visual estimation and the bearing from the observer was recorded by a magnetic compass. These data helped us to project near-exact bird locations to input into a species distribution model (SDM). In about 17% of cases, distance and bearing were not recorded. As the mean distance to birds was 215 m where distance was recorded, we used the location of the observer in these 17% of cases in our models, as it is likely to fall within the same 333-m pixel, the resolution used in these SDMs. At the end of each season of field surveys, we built updated SDMs of the species (see below) using the most recently collected data to identify areas to be surveyed in the following breeding season that were predicted to have a high probability of occupancy, but which had not yet been surveyed.
Fig. 1

Summary of Bengal Florican surveys across India and Nepal. Green circles indicate where surveys detected Bengal Floricans, crosses where surveys were undertaken but no birds were recorded. The red stars indicate sites where birds were fitted with satellite transmitters. The one large and two small boxes show areas within which modelling was restricted, from which areas above 400 mean sea level were additionally excised. Models were not extrapolated to Bangladesh, Myanmar or Tibet (China). Note: the international borders depicted herein may not be accurate (colour figure online)

Satellite telemetry

Birds were trapped under licence from the Ministry of Environment, Forest and Climate Change (MoEF&CC) in India and from the Department of National Parks and Wildlife Conservation in Nepal during the breeding season when they were more conspicuous. For trapping, we used leg nooses, or drove birds slowly towards single-shelf nets using a vehicle. Trapped birds were fitted with 35-g Argos PTTs (Microwave Telemetry, Maryland), which were attached to the back with a Teflon tape harness using a standard backpack configuration. Tag attachment took less than 30 min and each PTT and harness weighed c. 3% of the bird’s weight, slightly less than the general rule-of-thumb of 4% (Bander and Cochran 1991). The longevity of the tagged birds (see “Results”) and subsequent re-sightings of tagged birds in the field confirmed that tag attachment had no deleterious long-term effects on the birds. Tags were configured to transmit for 10 h and then to re-charge for 24 h. Dates and locations of capture of each of the 11 birds caught and tagged are given in Table 1. In the analyses of species distribution, we used only locations with Argos location class 3, which is claimed by the manufacturers to equate to a location accuracy of around 150 m. Although it has been suggested that the accuracy may be considerably lower (e.g. Nicholls et al. 2007), our further analysis was conducted at 333 m resolution, which should minimise the impact of location errors. We estimated great circle distances between successive locations using the haversine formula. Visualisation of tracking locations in Geographic Information System (GIS) indicated clear movements out of the breeding areas (where birds were tagged) to discrete non-breeding areas, where birds established non-breeding home ranges, and the breeding and non-breeding seasons of the tagged birds were defined accordingly.
Table 1

Date, sex and location of tagged birds, number of days tracked (to 31 January 2018) and the maximum distance between level-3 location classes across the period of tracking

Tag no.

Sex

Date tagged

Date of presumed mortality

Site

Days tracked

Maximum extent (km)

123070

F

3 June 2014

Shukla (N)

1338

36.3

123071

M

2 April 2013

Koshi (N)

1765

9.5

123072

M

22 May 2015

Shukla (N)

985

10.4

123074

F

19 April 2014

17 August 2017

Koshi (N)

1214

33.9

123075

M

1 April 2013

 

Koshi (N)

1675

12.7

123076

M

3 June 2014

11 September 2017

Shukla (N)

1185

34.6

123078

M

9 May 2014

 

Pilibhit (I)

1363

26.8

123079

M

12 May 2014

17 August 2014

Pilibhit (I)

97

23.2

123080

M

16 May 2014

 

Pilibhit (I)

1356

31.7

136675

M

27 June 2015

 

Dudhwa (I)

949

80.3

136678

M

15 May 2016

 

Chitwan (N)

626

7.2

Shukla Shuklaphanta National Park, Koshi Koshi Tappu Wildlife Reserve and Koshi barrage area, Pilibhit Pilibhit Tiger Reserve, Chitwan Chitwan National Park, N Nepal, I India

Home range estimation

We estimated home range using 50 and 95% minimum convex polygons (MCPs) as well as 50% (core) and 95% (outer) kernel utilisation distributions (KUDs) for both breeding and non-breeding seasons. Locations were recorded in latitude and longitude WGS84 and were transformed to the appropriate projection zone in the Universal Transverse Mercator for analysis. Home ranges were calculated for each individual (n = 11) for each breeding and non-breeding season over which they were tracked (ESM 1, Tables S1, S2). We delineated the core breeding season as the period between the day of arrival of birds at their respective breeding territories, or the known beginning of the breeding season from previous studies (whichever was later), until the first local movement was exhibited away from the birds’ territories, or the known arrival of monsoon in the region (whichever was earlier), when the breeding season is known to cease (Sankaran 1991).

When examining movement patterns and site fidelity, home ranges were produced for all individuals. However, for further home range analyses, only those individuals that had at least 50 high-accuracy locations in the given period of interest were included to provide reliable estimates (Seaman et al. 1999). The mean number of locations obtained per individual per season was 238 (n = 66, SD = 118.45) with a wide range of 36–677 due to variation in the performance of individual transmitters and the duration of a given season.

Home range estimates (MCPs and KUDs) were calculated using the adehabitatHR package (Calenge 2015) in R (version 3.3) (R Core Team 2016) and are reported with ranges. The least-squares cross validation procedure was used to determine the smoothing parameter (Seaman and Powell 1996). The number of points used to generate seasonal ranges varied from 59 to 677 (n = 65, mean = 241, SD = 116.63), providing robust kernel density estimates (KDEs). Although KDEs assume that the data are independent and identically distributed, we did not consider autocorrelation between location fixes as a significant factor influencing our home range estimates, since multiple simulation-based evaluations (e.g. Blundell et al. 2001; Fieberg 2007) have shown that sub-sampling data to achieve serial independence of observations is counter-productive and results in reducing the biological relevance of such derived home range estimates (De Solla et al. 1999). Movement-based maps were generated using the open-source geographic information system software QGIS (version 2.14) (QGIS Development Team 2016).

Home range estimates were compared between the core breeding period and non-breeding period using the non-parametric Wilcoxon signed-rank test in R [package core (Hothorn et al. 2008)]. Results were reported with the relevant W- and z-statistic along with effect size r (Pallant 2016; Sullivan and Feinn 2012). For individuals with multi-year data, home range estimates were averaged across years to avoid pseudoreplication.

Survival estimates

We assumed that birds were alive when the tracking data showed movement and the activity detector in the tag continued to indicate that the bird was moving around. We modelled the survival of tagged birds as a Bernoulli process using a generalised linear model with a logit link function. The number of days of observation was fitted as the binomial denominator. We did not fit any explanatory variable to the model due to the small sample size of just 11 tagged birds.

Distribution models

SDMs were created with MaxEnt version 3.3.3k (Phillips et al. 2006), which uses a machine-learning algorithm to produce niche models. Niche modelling predicts a species’ geographic distribution as a function of occurrence records and environmental data layers (Guisan and Thuiller 2005). MaxEnt has been shown to perform well against other distribution modelling protocols, and across a variety of metrics of model performance (Elith and Graham 2009). We created two sets of SDMs. In the first, we modelled the species’ breeding distribution across its entire historical range in India and Nepal (Bhutan is included to encompass the Royal Manas National Park, which abuts the Manas Tiger Reserve in India to form the larger Manas Biosphere Reserve), using locations of birds collected by the breeding season field surveys from 2013 to 2016 as input. In the second, we modelled the fine-scale breeding and non-breeding distributions within two core areas where our tagged birds were concentrated (Shuklaphanta/Pilibhit/Dudhwa region in western Nepal and Uttar Pradesh, India; Koshi region in eastern Nepal; Fig. 1), using high-accuracy Argos location class 3 data obtained from tagged birds between August 2015 and July 2016 as input. For SDMs in the ‘Pilibhit region’ (comprising a total of six tagged birds at Shuklaphanta, Pilibhit and Dudhwa Reserves), we delineated the core breeding season as April–June, and the non-breeding season as August–February. Correspondingly, for SDMs in the ‘Koshi region’ (comprising a total of three tagged birds—two just outside the Koshi Tappu Wildlife Reserve and one within the reserve), we defined the core breeding season as April–July, and the non-breeding season as September–February.

We delimited the range-wide breeding distribution as a rectangle encompassing all recent and most historical records of the species in the Indo-Nepal Terai and Brahmaputra floodplain (Fig. 1) and clipped out all areas over 400 m a.s.l., above which no historical or recent records of the Bengal Florican exist. Because of the elusive nature of the species, we did not use survey locations at which no birds were recorded as ‘true’ absences, but instead used presence-only modelling. In order to generate a targeted set of pseudo-absences (we used the MaxEnt default setting of 10,000 pseudo-absences), we separated high-intensity agricultural areas from natural and semi-natural grasslands and low-intensity agriculture, as areas in the former category have never been known to be used by Bengal Floricans for breeding. For this, we stacked 12 normalised difference vegetation index (NDVI) images (pixel size 333 m) covering the entire modelled area [one composite image of each month in the middle decad (10-day period)] from PROBA-V sensor (http://www.vito-eodata.be) during a 12-month period (January–December 2014, when the greatest number of field surveys for Bengal Florican detection were undertaken) in ArcMap. We then used the isocluster function in ArcMap 10.4 (ESRI 2016) to classify the various land cover types in the landscape into an arbitrary 35 classes/clusters (estimated with respect to the inherent heterogeneity of the landscape). This approach uses the temporal-NDVI signature (and not only NDVI) of pixels to cluster and differentiate between various land cover categories. We used visual interpretation to identify which of the resultant 35 clusters represented grassland habitats. Ground-truthing compared a representative sample of pixels of each of the 35 clusters with high-resolution images (Google Earth). We identified those classes which showed as grassland on these high-resolution images. This enabled us to produce a layer that mapped natural and semi-natural grassland habitats. We ran the geographic range-wide breeding SDM on this layer that excluded intensively cultivated areas.

We also produced a model of Bengal Florican breeding distribution that was not constrained to the areas identified as grassland by the unsupervised classification. For this, we used point location data from all surveys done for detecting Bengal Floricans during 2013–2016, together with 350 other survey points in the Brahmaputra floodplain potentially suitable for Bengal Floricans provided by Rahmani et al. (2016). As it selected pseudo-absence points from any land cover type, rather than just grassland, this model could potentially also include highly managed grasslands, which might be unsuitable for the species, and could therefore overestimate potential distribution. It thus may be considered as one describing the distribution of habitats that could potentially be suitable for Bengal Florican rather than necessarily the species’ distribution.

The models used eight bioclimatic variables (mean annual temperature, temperature seasonality, mean temperature of driest quarter, mean temperature of warmest quarter, mean temperature of coldest quarter, annual precipitation, precipitation seasonality coefficient of variation, and precipitation of driest quarter) thought to influence and characterise the typical low-lying seasonally flooded grasslands used by the species, downloaded from http://www.worldclim.org at a resolution of 30 arc-s (or approximately 1 km at the equator). The models also used a topographic variable relating to elevation/altitude (Global Multi-resolution Terrain Elevation Data 2010) that was downloaded as raster tiles from http://www.earthexplorer.usgs.gov at a 7.5 arc-s (or approximately 250 m at the equator) spatial resolution. As a non-categorical summary of land cover, we used NDVI from the PROBA-V sensor. One raster (decad 2, the middle 10 days of each month) from each of the 5 months of 2014 during the breeding season (February–June) were downloaded from http://www.vito-eodata.be for this purpose. We also used the estimated human population density raster dataset for 2015 (Gridded Population of the World, version 4) downloaded from http://www.sedac.ciesin.columbia.edu at a 30 arc-s (or approximately 1 km at the equator) spatial resolution as we predicted that the choice of Bengal Florican breeding areas/sites (and the extent of ‘suitable’ habitats themselves) may significantly be influenced by this variable. We acknowledge that the spatial resolution of these data is much cruder than 1 km, but they should represent an approximately accurate and consistent estimate of human population densities. We also used a Euclidean distance-to-river raster (derived from http://www.diva-gis.org/gdata) as a potential determinant of distribution.

For the local fine-scale distribution models based on locations from tracking data, we used only two sets of environmental variables: average NDVI during months (decad 2) constituting the breeding or non-breeding seasons, and the estimated human population density raster dataset for 2015. These models aimed to assess whether any movements out of the breeding areas was associated with a decline in their suitability at other times of year.

All model variables were resampled to the same spatial scale (333 m) in which the NDVI data were obtained, projected on the WGS84 projection and processed using ArcMap version 10.4. Variables selected for inclusion in the final models were those that contributed > 3% to the maximal model to avoid over-fitting the models while maximising their predictive power. No attempts were made to omit collinear variables, as machine-learning methods have been shown to perform well with such variables, especially when the study goal is predictive accuracy (Elith et al. 2011). Regularisation values of 5 and 10 were chosen for the local scale and range-wide models, respectively, based on visual interpretation of the ‘smoothness’ in the predictive range as determined by successive models using sequentially higher regularisation multipliers (1, 5, 10 and 15) and the area under the curve (AUC) values. The AUC, in reference to the receiver operating characteristic curve, is a threshold independent measure of model performance (Metz 1978; Elith 2000). When using presence-only data, as in this study, an AUC is calculated using background (or pseudo-absences) data chosen uniformly at random from the study area (Phillips and Dudík 2008). The AUC is then the ‘probability that a randomly chosen presence site is ranked above a random background site’ (Phillips et al. 2006). We used the linear feature function for all predictor variables (all continuous) in the models. All models were built by backwards deletion, retaining variables that continued to explain more than 3% of the variation in the data. We ran ten cross-validations over the data to reduce the chance of models being influenced by outliers (Phillips and Dudík 2008). The continuous outputs from the modelling were binned into binary outputs (potentially occupied/not occupied) using the equal training sensitivity and specificity logistic threshold value.

Results

Field surveys

A total of 934 field surveys were undertaken in India and Nepal during the Bengal Florican breeding season (February–early July) during 2013–2016. Bengal Floricans were detected in 255 surveys (27.3%), although this relatively high rate of detection is biased by the concentration of surveys in areas where the species was expected or predicted to be present, and by sampling in such areas in successive years. In those 255 surveys, a total of 425 Bengal Florican detections were recorded, only 7% of which were of females. We used all the 425 detection events as input to the model of the range-wide distribution.

Home ranges

There were clear movements out of the breeding areas (where birds were tagged) to discrete non-breeding areas in all the three regions in which birds were tracked (Fig. 2). We therefore estimated both breeding and non-breeding home ranges (summary for 11 tagged birds in Table 2). Mean MCP 50% and KUD 50% (denoting the birds’ core use area) home range estimates of 11 tagged birds averaged across one to five breeding seasons (2013, 2014, 2015, 2016 and 2017) were 0.96 (0.08–19.122) km2 and 0.48 (0.12–2.16) km2, respectively. Mean MCP 95% and KUD 95% (denoting the birds’ outer home ranges and discounting outliers) breeding home range estimates were 5.90 (0.81–62.18) km2 and 2.86 (0.85–10.67) km2, respectively. Mean MCP 50% and KUD 50% home range estimates for ten PTT-tagged birds across two to four non-breeding seasons (2013–2014, 2014–2015, 2015–2016, 2016–2017 and 2017–2018 for the Chitwan bird) were 10.15 (0.29–124.7) km2 and 5.62 (0.22–98.52) km2 respectively, whereas mean MCP 95% and KUD 95% home range estimates were 32.20 (1.80–305.04) km2 and 30.59 (1.40–487.08) km2, respectively.
Fig. 2

Movements by tagged birds in a Pilibhit Tiger Reserve, Uttar Pradesh, India (n = 3) and in Shuklaphanta National Park, Nepal (n = 3); b Dudhwa National Park, Uttar Pradesh, India (n = 1); c Koshi area (inside and outside Koshi Tappu Wildlife Reserve), Nepal (n = 3); and d Chitwan National Park, Nepal (n = 1). Symbols indicate different birds (where n > 1). Green: Breeding season, purple : non-breeding season, blue: movements immediately post-breeding. The dashed black line is the international border between Nepal and India. a, b The Sharda River runs diagonally across the centre of the image from northwest to southeast; c the Koshi river runs diagonally across the centre of the image from northeast to southwest; d the Narayani River runs diagonally across the centre of the image from northeast to southwest, and the Rapti River joins it from the west. Note: the international borders depicted herein may not be accurate (colour figure online)

Table 2

Minimum convex polygon (MCP) and kernel home ranges of tagged birds during 2013–2017

Season

Sex

Home range area (km2)

Home range area (km2)

MCP 50%

MCP 95%

50% Kernel core

95% Outer kernel

Mean

Min.

Max.

Mean

Min.

Max.

Mean

Min.

Max.

Mean

Min.

Max.

Breeding (core)

Males (n = 9)

0.376

0.095

1.22

2.651

0.814

9.9

0.394

0.12

1.148

2.394

0.849

5.154

Females (n = 2)

3.374

0.08

19.122

19.38

1.182

62.18

0.859

0.298

2.159

4.808

1.776

10.67

All (n = 11)

0.959

  

5.904

  

0.485

  

2.864

  

Non-breeding

Males (n = 8)

11.203

0.296

124.7

30.461

1.804

305.04

1.3

0.441

8.54

8.095

2.773

56.318

Females (n = 2)

6.121

0.656

25.938

38.864

5.912

90.738

22.19

0.225

98.518

116.82

1.398

487.081

All (n = 10)

10.151

  

32.199

  

5.622

  

30.59

  

The non-breeding season home range estimates were significantly larger than the breeding season home ranges, whether considered as 50% MCP (Wilcoxon sign-rank test: W = 6, z = − 2.191, p = 0.027, r = 0.49), 95% MCP (W = 1, z = − 2.701, p = 0.004, r = 0.60), 50% KUD (W = 0, z = − 2.803, p = 0.002, r = 0.62) or 95% KUD (W = 0, z = − 2.803, p = 0.002, r = 0.62).

Survival estimates

At the end of the exposure period (January 2018), eight of the 11 tagged birds were still alive (Table 1). In the other three cases, tag failure cannot be ruled out, so survival estimates may be underestimated. The total exposure period across all birds was 12,761 days, yielding a daily survival rate of 0.9998. This equates to an annual adult survival rate of 0.918 (0.858–0.953 ± back-transformed SE), although with such a small sample size, the confidence was inevitably low.

Movements

There were 20,273 unique locations of high accuracy (location class 3) obtained throughout the tracking period, with successive locations from each bird being an average of 0.6 days apart. Most tagged birds in Uttar Pradesh, India (Pilibhit and Dudhwa reserves) and western Nepal (Shuklaphanta reserve) moved in the non-breeding season to seasonally inundated areas along the Sharda (or Mahakali) River, involving movements of up to 80 km. Movements to these areas were sometimes preceded by shorter movements to areas just outside the protected areas in which the birds were caught (Fig. 2). The furthest distance moved in a day was 32 km, but this was exceptional. Even when moving between breeding and wintering sites, individual movements tended to be small, suggesting that much of the displacement observed occurred through walking and short flights rather than occasional long-distance movements.

There was intermittent overlap (in space and time) in the non-breeding ranges of birds from Pilibhit and birds from Shuklaphanta, and birds from both breeding areas crossed the border between India and Nepal during successive non-breeding seasons. Field surveys of non-breeding areas indicated that they were unmanaged grasslands intermixed with farmlands near settlements. Most such non-breeding areas were characterised by relatively low human population densities (due to seasonal flooding), near-absence of permanent infrastructure, and relatively low-intensity agricultural production of crops such as sugarcane, mustard, lentil, rice and wheat, intermixed with fallow plots and intensively grazed grasslands dominated by Typha angustifolia, Saccharum spontaneum, Phragmites karka, Cynodon dactylon and Imperata cylindrica (ESM 1, Fig. S3). Dominant grass species at breeding sites inside PAs were Imperata cylindrica, Saccharum spontaneum, Erianthus munja, Vetiveria zizanioides, Narenga porphyrocoma, Cymbopogon martini, Sclerostachya fusca and others (in no particular order) (ESM 1, Fig. S4). Birds tracked for more than 1 year showed a strong degree of fidelity to both their breeding and non-breeding sites, although the latter was weaker. The solitary tagged bird at Chitwan National Park spent the non-breeding period in patches of grasslands and farmlands along the Narayani and Rapti rivers close to the reserve (Fig. 2).

Despite the small sample sizes, there was a suggestion that birds in different breeding areas showed different patterns of departure and return dates from their breeding areas, with birds breeding at Shuklaphanta returning to their breeding grounds around 2 months before birds elsewhere (Table 3), and that movements of birds at particular sites were strongly correlated (Spearman’s ρ = 0.85, p = 0.0004; see ESM 1, Table S3 for details). For example, the two tagged birds at Pilibhit that were tracked over successive movements usually left or returned to their breeding sites within a few days of each other. Two males at Koshi behaved similarly, leaving or returning to the breeding site within a week of each other in 5 successive years.
Table 3

Dates of post-breeding departure from the breeding sites and return from the non-breeding areas, and the number of days spent away from the breeding site

Site

Bird

Departure

Return

Days away

Pilibhit Tiger Reserve

123078

29 August 2014

2 April 2015

215

23 September 2015

13 March 2016

204

5 September 2016

28 March 2017

203

1 July 2017

123079

25 July 2014

End: 17 August 2014

123080

20 September 2014

1 April 2015

203

14 September 2015

14 March 2016

180

5 September 2016

23 March 2017

198

27 June 2017

 Site mean

19 August

23 March

200.5

Dudhwa National Park

136675

28 July 2015

1 April 2016

246

13 July 2016

17 March 2017

246

22 July 2017

 Site mean

21 July

24 March

246

Shuklaphanta National Park

123070

15 September 2014

7 December 2014

83

1 September 2015

14 January 2016

135

3 September 2016

12 January 2017

131

15 September 2017

23 November 2017

69

123072

17 August 2015

1 February 2015

167

29 August 2016

27 January 2017

151

7 September 2017

23 January 2018

138

123076

27 August 2014

3 February 2015

159

31 August 2015

23 January 2016

144

16 August 2016

18 January 2017 

154

End: 11 September 2017

 Site mean

31 August

12 January

133.1

Koshi region

123071a

26 October 2013

15 March 2014

139

22 September 2014

21 March 2015

179

21 August 2015

13 March 2016

203

17 August 2016

3 April 2017

228

13 August 2017

….

123075a

25 January 2014b

16 March 2014

50

24 September 2014

22 March 2015

178

17 August 2015

23 March 2016

217

27 August 2016

3 April 2017

218

13 August 2017

 Site mean

2 September

22 March

176.5

Chitwan National Park

136678

1 June 2016

6 February 2017

250c

11 June 2017

1 February 2018

245c

 Site mean

6 June

3 February

247.5

A female bird tagged at Koshi was highly itinerant during the non-breeding season and frequently returned to and then left its breeding site, so is not included. One bird from Pilibhit (123079) presumably died shortly after leaving its breeding area in 2014. Bird 123070 is a female bird

aTagged birds 123071 and 123075 in the Koshi region spend both their breeding and non-breeding seasons outside the nearest protected area (Koshi Tappu Wildlife Reserve) and in fact breed in the Koshi barrage area, thus the Days away column in their case implies the time that they spend outside their specific breeding site in the non-breeding season

bConsidered an outlier and thus not included in site mean calculations

cTagged bird 136678 spent 22–35 days inside Chitwan National Park intermittently during the non-breeding seasons of 2016–2017 and 2017–2018

Distribution models

Range-wide model of breeding distribution

The model of the potential breeding distribution of Bengal Florican across northern India and Nepal had an AUC of 0.968, suggesting it has good predictive power. This model was based on pseudo-absences taken from within areas in the unsupervised grassland layer. The model indicated that there might be up to 22,535 km2 of suitable habitat extending along the length of the India-Nepal international border (in the districts of Udham Singh Nagar and Nainital of Uttarakhand; Pilibhit, Kheri, Bahraich, Maharajganj and Balrampur of Uttar Pradesh; West Champaran, Araria and Supaul of Bihar; Darjeeling, North Dinajpur, Jalpaiguri, Alipurduar and Cooch Behar of West Bengal—all in India; and in the districts of Kanchanpur, Bardiya, Rupandehi, Nawalparasi East, Chitwan, Parsa, Bara, Rautahat, Sarlahi, Mahottari, Dhanusa, Siraha, Saptari, Udayapur, Sunsari, Morang and Jhapa of Nepal) and eastwards on both sides of the river Brahmaputra in Assam (significant areas of districts Kokrajhar, Chirang, Baksa, Dhubri, Sonitpur, Lakhimpur, Golaghat, Nagaon, Dhemaji, Jorhat, Sivasagar, Tinsukia and Dibrugarh) as well as lowland Arunachal Pradesh (districts of East Siang, Lower Dibang Valley, Namsai, Lohit and Changlang) (Fig. 3) (ESM 2). The model that included pseudo-absences from outside grasslands, and so could be considered a suitable habitat map, had an AUC of 0.980. The distribution of suitable habitat predicted by this model was broadly similar to that of Florican distribution predicted by the model focused within unsupervised grassland only (Fig. 3), but the area predicted as potentially suitable was around 33,000 km2. There was a high degree of concordance in the model predictions, but this model predicted substantially more areas of potentially suitable habitats in Nagaon, Darrang, Kamrup, East Karbi Anglong, West Karbi Anglong, and Morigaon districts of Assam (in northeast India), in Shravasti (particularly Soheldev Sanctuary) and Kushinagar districts of Uttar Pradesh, and in Nainital district (particularly Ramnagar Forest Division) of Uttarakhand (northwest India) (Fig. 3) (ESM 2).
Fig. 3

Modelled distribution of Bengal Florican across its historical range in India and Nepal; green indicates potentially suitable areas according to the model using pseudo-absences from unsupervised grassland areas, blue indicates potentially suitable areas/habitats according to the model built using pseudo-absences from all land cover types, red indicates areas predicted as suitable by both models; for GeoTIFF files of areas predicted as ‘suitable’, please see ESM 2. Note: The international borders depicted herein may not be accurate (colour figure online)

Local-scale models

The AUC value of the species’ breeding distribution model around Koshi for the breeding season (Fig. 4a) was 0.956, and 0.951 for the non-breeding season. The form of the relationship with the variables was the same (ESM 1, Figs. S5a and b). The corresponding AUC values for the area around Pilibhit (Fig. 4b) were 0.960 for breeding and 0.896 for non-breeding season. The form of the relationships with NDVI differed between the seasons, with birds more likely to occupy areas of higher NDVI in the breeding season, with the reverse true in the non-breeding season (ESM 1, Figs. S5c, d). The AUC values for Koshi and Pilibhit in the breeding season were similar, but the value for Pilibhit was lower than that for Koshi in the non-breeding season. Although this might indicate that the model was not as good a fit to the data in Pilibhit, it is difficult to compare AUC values between different models of this sort. However, the form of the relationships with NDVI differed greatly between the regions for the breeding season. In Koshi the birds were more likely to be found where NDVI was low between April to July, but in Pilibhit they were more likely to be found where NDVI was high (ESM 1, Figs. S5a, c). During the non-breeding season though, birds from both regions were more likely to use areas with lower than average NDVI (ESM 1, Figs. S5b, d).
Fig. 4

Breeding (green) and non-breeding (blue) modelled distribution of Bengal Florican around a Koshi region and b Pilibhit region (see Figs. 1 and 2 for locations), red indicates overlap; black polygons are protected areas (represented by area-proportional circles where the precise boundaries are not available); for GeoTIFF files of areas predicted as suitable, please see ESM 2. Note: the international borders depicted herein may not be accurate (colour figure online)

The model indicated that around 1100 km2 were potentially suitable breeding habitat around Koshi, with a similar area of 1055 km2 potentially suitable in the non-breeding season. The area of overlap between breeding and non-breeding seasons was high (Fig. 4a). In the Pilibhit region, around 3360 km2 was predicted to be suitable in the breeding season compared to 5086 km2 in the non-breeding season. The area of overlap between the two seasons here was low (Fig. 4b). The majority of protected area coverage in the Pilibhit region captured only the breeding areas, whereas in the Koshi region the protected areas captured an area used for both breeding and non-breeding. However, the overall coverage of predicted suitable areas was poor in both regions (ESM 2).

Discussion

We used a combination of field surveys, satellite telemetry and remote sensing to model the distribution and assess the movements, survival and home ranges of the critically endangered Bengal Florican across its distribution range in India and Nepal for the first time.

The range-wide SDMs support what is known from recent field surveys in suggesting that the species’ range becomes increasingly fragmented from east to west. SDMs also suggest that there may be large areas of potentially suitable habitat in Assam and Arunachal Pradesh where surveys have not been undertaken. Furthermore, surveys in these areas may reveal the presence of hitherto unknown but potentially sizeable populations. In Uttar Pradesh and western Nepal, the models predicted few areas that are not already known to hold birds. A priority conservation action in these areas is to restore and connect surviving fragments of habitat, for example at Bardia National Park (Nepal) and Kishanpur Sanctuary (India), which seem to have lost their populations, and at Dudhwa National Park (India), which has seen a considerable decline over the last 15 years (Rahmani et al. 2017). The movements of birds outside the breeding season suggest that the species has the dispersal capacity to re-occupy restored areas of grassland outside its current breeding range.

The telemetry study revealed hitherto unknown movements of birds from July onward out of their grassland breeding habitats and into areas with a mosaic of unmanaged grasslands and low-intensity agriculture, largely along the seasonally flooded Sharda/Mahakali and Koshi rivers. Indeed, some of the tagged birds spent more than half the year outside their breeding sites (Table 3). The distances moved were similar to those recorded from tagged birds in the Cambodian population (Packman 2011; Mahood et al. 2018), but the reasons for these movements are less clear. In Cambodia, birds are forced to move by seasonal flooding of the Tonle Sap, but there is no prolonged inundation in the sites at which breeding birds were caught in India and Nepal. Our preliminary exploration of changes in grassland in the breeding areas around the time the birds leave has revealed substantially different trends (see ESM 1, Fig. S6 for an example). How changes in grassland composition and structure, which may differ among sites, and certain other factors such as food availability, behavioural plasticity in terms of site fidelity, and human disturbances influence Bengal Florican dispersal warrants further research (Rahmani et al. 2017). This has implications for the conservation of the species, which now must take into account not only grassland habitats in what are largely protected areas, but also the wider landscape. Furthermore, the frequency of cross-border movements of birds between India and Nepal indicates that international collaboration is required to conserve this species.

Local-scale breeding season models showed different trends with NDVI in the Koshi and Pilibhit regions where birds are associated with lower and higher NDVI, respectively. A possible reason for this discrepancy could be that grassland habitats where Bengal Florican breeds in the Pilibhit region have been extensively afforested, mostly with eucalyptus and teak plantations by the Forest Department under various schemes in the past, thereby causing an increase in the ‘greenness’ of these areas (ESM 1, Fig. S2). Non-breeding season models suggest that the type of farmland selected by birds is distinctive and spatially limited, since only a relatively small proportion of the largely agricultural region covered by the models is predicted to be suitable. The models also predict that some of this area may also be suitable for breeding birds. Surveys should be undertaken in these areas during the breeding season to confirm this. Bengal Florican now joins a list of globally threatened species for which low-intensity agriculture should be considered an important habitat (Wright et al. 2011). Whether similar movements are undertaken by birds breeding in the species’ strongholds in Assam and Arunachal Pradesh, while expected (Collar et al. 2001), remains to be assessed.

The Bengal Floricans tagged in this study maintained smaller home ranges in the breeding season than in the non-breeding season, as did tagged Bengal Floricans in Cambodia (Packman 2011). In both regions, the seasonal difference in home ranges may reflect a difference in habitat quality, a patchier distribution of resources in the non-breeding season, where only certain crop types may be suitable, and/or varying intensity of anthropogenic disturbance. Further research in the non-breeding areas is required to assess specific habitat requirements at this time. One of the two tagged females had a much larger home range size than any other tagged bird, but it is unclear whether this reflects differences between the sexes or the movements characteristic of the site where this bird was tagged.

Survival rates appeared to be high, as they were in the Cambodian population (Packman 2011; Mahood et al. 2018). This suggests that, at least in the areas where tagging took place, the common problem in other bustard species of low adult survival resulting from collision with overhead power lines (Jenkins et al. 2011; Raab et al. 2012; Silva et al. 2014; Burnside et al. 2015) is not likely to be the main reason for the species’ decline. Instead, efforts should be made to increase productivity, for example by reducing disturbance during the breeding season and ensuring a heterogeneous grassland structure that allows females to nest in taller patches and birds to forage in more open areas. In many protected areas where birds were tagged (e.g. Dudhwa and Pilibhit Reserves in India, Shuklaphanta Reserve in Nepal), annual winter burning of grasslands by the forest administration has been a long-established management practice to stimulate the growth of fresh forage for large herbivores. However, sometimes grass is burnt very late in the season, in February and March, coinciding with the Bengal Florican display season. The effect of such late-season grass burning is not known, but is likely to be negative. All three mortalities occurred at some point between 17 August and 11 September, which coincided with the time birds were leaving their breeding areas. Although based on a small sample size, it might be prudent in protected areas to maintain suitable habitat year-round in order to prevent birds from having to leave them to disperse into more hostile environments.

In conclusion, this study elucidates the breeding and non-breeding distribution of Bengal Floricans in the Asian subcontinent. Promisingly, we revealed potentially new areas that could support populations of this critically endangered species. Future surveys will be needed to corroborate model results. If indeed Bengal Floricans are present at hitherto unknown locations, appropriate conservation actions such as protecting breeding sites, habitat restoration, community awareness/education/action will urgently be needed to preserve them.

Notes

Acknowledgements

For help with trapping, we are extremely grateful to Ali Hussain, Sikandar Hussain, Lotty Packman, Marcus Handschuh and Jacky Judas. For other help in the field and other support in India we thank the MoEF&CC, Rupak De, Sanjaya Singh, V. K. Singh, Mahaveer Koujalagi, M. K. Yadava, Tashi Mize, Kailash Prakash, Sonali Ghosh, Siva Kumar, Narendra Upadhyaya, K. N. Gautam, K. P. Singh, Vinod Tiwari, Sonu Liladhar, Iphra Mekola, Smarajit Ojah, Bibhab Talukdar, Namita Brahma, Bibhuti Lahkar, Biswajit Chakdar, Manabendra Ray Choudhury, Rohit Ravi, Rekha Warrier, the late Carl D’silva, Rajat Bhargava, Mohit Kalra, Rohan Bhagat, Deepak Apte; forest staff and forest departments of Uttarakhand, Uttar Pradesh, Bihar, Assam and Arunachal Pradesh States, and the administrative staff of Bombay Natural History Society. For help in Nepal we thank the Ministry of Forest and Soil Conservation, Department of National Parks and Wildlife Conservation, Krishna Prasad Acharya, Meg Bahadur Pandey, Gopal Prakash Bhattarai, Laxman Poudyal, Rishi Ranabhat, Mahananda Joshi, Eshwar Pant, Badri Chaudhary, Anish Timsina, Arjun Karki, Bashu Bidari, Hum Gurung, Sushila Nepali, Rajendra Dhungana, Ganga Bahadur Singh, Annanath Baral, Bed Khadka, Ramchandra Kandel, Kamaljung Kunwar, Nurendra Aryal, Bishnu Thapaliya, Gopal Ghimire, Bed Bahadur Dhakal, Yubraj Ghimire, Shambhu Ghimire, Hem Bahadur Katuwal and Ishana Thapa. This study was funded by the UK Darwin Initiative (grant no. 19-011) and the Indian MoEF&CC. Rohit Jha is grateful to the Miriam Rothschild Travel Bursary Programme for funding a 4-week internship with the RSPB in Cambridge, UK. We are grateful to Jonathan Handley, Nigel Collar and three anonymous reviewers for useful comments on this paper.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.

Research involving human participants and/or animals

All applicable international, national, and/or institutional guidelines for the care and use of animals were followed and required permissions from national and state authorities as per prevalent legislation of the country were arranged before trapping animals. Raw datasets and/or those analysed during the current study are available from the corresponding author on reasonable request.

Supplementary material

10336_2018_1552_MOESM1_ESM.docx (3 mb)
Supplementary material 1 (DOCX 3088 kb)
10336_2018_1552_MOESM2_ESM.zip (7.4 mb)
Supplementary material 2 (ZIP 7559 kb)

References

  1. Bander R, Cochran W (1991) Radio-location telemetry. Wildlife Society, Washington, DCGoogle Scholar
  2. Baral N, Timilsina N, Tamang B (2003) Status of Bengal Florican Houbaropsis bengalensis in Nepal. Forktail 19:51–55Google Scholar
  3. Baral HS, Ram AK, Chaudhary B, Basnet S, Chaudhary H, Giri TR, Chaudhary D (2012) Conservation status of Bengal Florican Houbaropsis bengalensis bengalensis (Gmelin, 1789) (Gruiformes: Otididae) in Koshi Tappu Wildlife Reserve and adjoining areas, eastern Nepal. J Threat Taxa 4:2464–2469CrossRefGoogle Scholar
  4. BirdLife International (2017) Species factsheet: Houbaropsis bengalensis. http://www.birdlife.org. 03 Apr 2017
  5. Blundell GM, Maier JAK, Debevec EM (2001) Linear home ranges: effects of smoothing, sample size, and autocorrelation on kernel estimates. Ecol Monogr 71:469–489CrossRefGoogle Scholar
  6. Burnside RJ, Collar NJ, Koshkin MA, Dolman PM (2015) Avian powerline mortalities, including Asian Houbara Chlamydotis macqueenii, on the Central Asian flyway in Uzbekistan. Sandgrouse 37:161–168Google Scholar
  7. Calenge C (2015) Home range estimation in R: the adehabitatHR package. https://cran.r-project.org/web/packages/adehabitatHR/vignettes/adehabitatHR.pdf
  8. Collar NJ, Andreev AV, Chan S, Crosby MJ, Subramanya S, Tobias JA (2001) Threatened birds of Asia: the BirdLife International Red Data Book. BirdLife International, CambridgeGoogle Scholar
  9. De Solla SR, Bonduriansky R, Brooks RJ (1999) Eliminating autocorrelation reduces biological relevance of home range estimates. J Anim Ecol 68:221–234CrossRefGoogle Scholar
  10. Donald PF, Collar NJ, Marsden SJ, Pain DJ (2013) Facing extinction: the world’s rarest birds and the race to save them, 2nd edn. Helm, LondonGoogle Scholar
  11. Dwivedi KK (2016) The Mahabahu Brahmaputra. Flood and Erosion Management Agency of Assam, GuwahatiGoogle Scholar
  12. Elith J (2000) Quantitative methods for modeling species habitat: comparative performance and an application to Australian plants. In: Quantitative Methods for Conservation Biology. Springer, New York, USA, pp 39–58Google Scholar
  13. Elith J, Graham CH (2009) Do they? How do they? Why do they differ? On finding reasons for differing performances of species distribution models. Ecography 32:66–77CrossRefGoogle Scholar
  14. Elith J, Phillips SJ, Hastie T, Dudík M, Chee YE, Yates CJ (2011) A statistical explanation of MaxEnt for ecologists. Divers Distrib 17:43–57CrossRefGoogle Scholar
  15. ESRI (2016) ArcGIS desktop: release 10.4. Environmental Systems Research Institute, RedlandsGoogle Scholar
  16. Fieberg J (2007) Kernel density estimators of home range: smoothing and the autocorrelation red herring. Ecology 88:1059–1066CrossRefPubMedGoogle Scholar
  17. Gray TNE, Collar NJ, Davidson PJA, Dolman PM, Evans TD, Fox HN, Chamnan H, Borey R, Hout SK, Van Zalinge RN (2009) Distribution, status and conservation of the Bengal Florican Houbaropsis bengalensis in Cambodia. Bird Conserv Int 19:1–14CrossRefGoogle Scholar
  18. Guisan A, Thuiller W (2005) Predicting species distribution: offering more than simple habitat models. Ecol Lett 8:993–1009CrossRefGoogle Scholar
  19. Hothorn T, Hornik K, van de Wiel MA, Zeileis A (2008) Implementing a class of permutation tests: the coin package. J Stat Softw 28:1–23. http://www.jstatsoft.org/v28/i08/
  20. Inskipp C, Collar NJ (1984) The Bengal Florican: its conservation in Nepal. Oryx 18:30–35CrossRefGoogle Scholar
  21. Jenkins AR, Shaw JM, Smallie JJ, Gibbons B, Visagie R, Ryan PG (2011) Estimating the impacts of power line collisions on Ludwig’s Bustards Neotis ludwigii. Bird Conserv Int 21:303–310CrossRefGoogle Scholar
  22. Mahood SP, Silva JP, Dolman PM, Burnside RJ (2018) Proposed power transmission lines in Cambodia constitute a significant new threat to the largest population of the critically endangered Bengal Florican Houbaropsis bengalensis. Oryx 52:147–155.  https://doi.org/10.1017/S0030605316000739 CrossRefGoogle Scholar
  23. Metz CE (1978) Basic principles of ROC analysis. Semin Nucl Med 8(4):283–298CrossRefPubMedGoogle Scholar
  24. Nicholls DG, Robertson CJR, Murray MD (2007) Measuring accuracy and precision for CLS: Argos satellite telemetry locations. Notornis 54:137–157Google Scholar
  25. Packman CE (2011) Seasonal landscape use and conservation of a critically endangered bustard: Bengal Florican in Cambodia. PhD Dissertation, University of East AngliaGoogle Scholar
  26. Packman CE, Gray TNR, Collar NJ, Evans TD, Van Zalinge RN, Virak S, Lovett AA, Dolman PM (2013) Rapid loss of Cambodia’s grasslands. Conserv Biol 27:245–247CrossRefPubMedGoogle Scholar
  27. Packman CE, Showler DA, Collar NJ, Virak S, Mahood SP, Handschuh M, Evans TD, Chamnan H, Dolman PM (2014) Rapid decline of the largest remaining population of Bengal Florican Houbaropsis bengalensis and recommendations for its conservation. Bird Conserv Int 24:429–437CrossRefGoogle Scholar
  28. Pallant J (2016) SPSS survival manual: a step by step guide to data analysis using IBM SPSS, 6th edn. Open University Press, MaidenheadGoogle Scholar
  29. Phillips SJ, Dudík M (2008) Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography (Cop) 31:161–175CrossRefGoogle Scholar
  30. Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecol Model 190:231–259CrossRefGoogle Scholar
  31. Poudyal LP, Singh PB, Maharjan S (2008a) The decline of Bengal Florican Houbaropsis bengalensis in Nepal. Danphe 17:4–6Google Scholar
  32. Poudyal LP, Singh PB, Maharjan S (2008b) Bengal Florican Houbaropsis bengalensis in Nepal: an update. BirdingASIA 10:43–47Google Scholar
  33. R Core Team (2016) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/
  34. Raab R, Schütz C, Spakovszky P, Julius E, Schulze CH (2012) Underground cabling and marking of power lines: conservation measures rapidly reduced mortality of West-Pannonian Great Bustards Otis tarda. Bird Conserv Int 22:299–306CrossRefGoogle Scholar
  35. Rahmani A (2012) Threatened birds of India. Oxford University Press, MumbaiGoogle Scholar
  36. Rahmani AR, Narayan G, Rosalind L, Sankaran R, Ganguli U (1991) Status of the Bengal Florican Houbaropsis bengalensis in India. J Bombay Nat Hist Soc 88:349–375Google Scholar
  37. Rahmani AR, Khongsai N, Rahman A, Imran M, Sangwan T, Ojah S (2016) Conservation of threatened grassland birds of the Brahmaputra floodplains. Bombay Natural History Society, MumbaiGoogle Scholar
  38. Rahmani AR, Jha RRS, Khongsai N, Shinde N, Talegaonkar R, Kalra M (2017) Studying movement pattern and dispersal of the Bengal Florican (Houbaropsis bengalensis): a satellite telemetry pilot project. Final Report 2013–2016. Bombay Natural History Society, Mumbai, p 157Google Scholar
  39. Sankaran R (1991) Some aspects of the breeding behaviour of the Lesser Florican Sypheotides indicus (JF Miller) and the Bengal Florican Houbaropsis bengalensis (Gmelin). PhD dissertation, University of BombayGoogle Scholar
  40. Seaman DE, Powell RA (1996) An evaluation of the accuracy of kernel density estimators for home range analysis. Ecology 77:2075–2085CrossRefGoogle Scholar
  41. Seaman DE, Millspaugh JJ, Kernohan BJ, Brundige GC, Raedeke KJ, Gitzen RA (1999) Effects of sample size on kernel home range estimates. J Wildl Manag 63(2):739CrossRefGoogle Scholar
  42. Silva JP, Palmeirim JM, Alcazar R, Correia R, Delgado A, Moreira F (2014) A spatially explicit approach to assess the collision risk between birds and overhead power lines: a case study with the Little Bustard. Biol Conserv 170:256–263CrossRefGoogle Scholar
  43. Sullivan GM, Feinn R (2012) Using effect size—or why the P-value is not enough. J Grad Med Educ 4:279–282CrossRefPubMedPubMedCentralGoogle Scholar
  44. Wright HL, Lake IR, Dolman PM (2011) Agriculture—a key element for conservation in the developing world. Conserv Lett 5:11–19CrossRefGoogle Scholar
  45. QGIS Development Team (2016) QGIS Geographic Information System. Open Source Geospatial Foundation Project. http://qgis.osgeo.org

Copyright information

© Dt. Ornithologen-Gesellschaft e.V. 2018

Authors and Affiliations

  • Rohit R. S. Jha
    • 1
    • 7
    Email author
  • Jyotendra Jyu Thakuri
    • 2
  • Asad R. Rahmani
    • 1
  • Maheshwar Dhakal
    • 3
  • Ngulkholal Khongsai
    • 1
  • Narendra Man Babu Pradhan
    • 2
    • 8
  • Nikhil Shinde
    • 1
  • Bridesh Kumar Chauhan
    • 1
  • Rahul K. Talegaonkar
    • 1
  • Ian P. Barber
    • 4
  • Graeme M. Buchanan
    • 4
  • Toby H. Galligan
    • 4
  • Paul F. Donald
    • 4
    • 5
    • 6
  1. 1.Bombay Natural History SocietyMumbaiIndia
  2. 2.Bird Conservation NepalKathmanduNepal
  3. 3.Environment DivisionMinistry of Forest and Soil ConservationKathmanduNepal
  4. 4.RSPB Centre for Conservation Science, RSPBSandyUK
  5. 5.BirdLife InternationalCambridgeUK
  6. 6.Conservation Science Group, Department of ZoologyUniversity of CambridgeCambridgeUK
  7. 7.Wildlife Institute of IndiaDehradunIndia
  8. 8.IUCN NepalKathmanduNepal

Personalised recommendations