Exploring the environmental drivers of waterfowl movement in arid landscapes using first-passage time analysis

Dominic A. W. Henry; Judith M. Ament; Graeme S. Cumming

doi:10.1186/s40462-016-0073-x

Movement Ecology

December 2016, 4:8 | Cite as

Exploring the environmental drivers of waterfowl movement in arid landscapes using first-passage time analysis

Authors
Authors and affiliations

Dominic A. W. Henry
Judith M. Ament
Graeme S. Cumming

Open Access

Research

First Online: 01 April 2016

3 Shares
1.6k Downloads
6 Citations

Abstract

Background

The movement patterns of many southern African waterfowl are typified by nomadism, which is thought to be a response to unpredictable changes in resource distributions. Nomadism and the related movement choices that waterfowl make in arid environments are, however, poorly understood. Tracking multiple individuals across wide spatiotemporal gradients offers one approach to elucidating the cues and mechanisms underpinning movement decisions. We used first-passage time (FPT) to analyse high spatial and temporal resolution telemetry data for Red-billed Teal and Egyptian Geese across a 1500 km geographical gradient between 2008 and 2014. We tested the importance of several environmental variables in structuring movement patterns, focusing on two competing hypotheses: (1) whether movements are driven by resource conditions during the current period of habitat occupation (reactive movement hypothesis), or (2) whether movements are structured by shifts in the magnitude and direction of environmental variables at locations prior to occupation (prescient movement hypothesis).

Results

An increase in rainfall at a 32 day lag (i.e., prior to wetland occupancy), along with tagging site, were significant predictors of FPT in both waterfowl species. There was a positive relationship between NDVI and FPT for Egyptian Geese during this 32 day period; the relationship was negative for Red-billed Teal. Consistent with findings for migratory grazing geese, Egyptian Geese prioritised food quality over food biomass. Red-billed Teal showed few immediate responses to wetland filling, contrary to what one would predict for a dabbling duck, suggesting high dietary flexibility. Our results were consistent with the prescient movement hypothesis.

Conclusions

Using FPT analysis we showed that the proximate drivers of southern African waterfowl movement are the dynamics of rainfall and primary productivity. Waterfowl appeared to be able to perceive and respond to temporal shifts in resource conditions prior to habitat patch occupation. This in turn suggests that their movements in semi-arid landscapes may be underpinned by intimate knowledge of the local environment; waterfowl pursue a complex behavioural strategy, locating suitable habitat patches proactively, rather than acting as passive respondents.

Keywords

Waterfowl NDVI Rainfall Southern Africa Nomadic First-passage time Alopochen aegyptiaca Anas erythrorhyncha

Background

Processes that drive movement occur on a wide range of spatiotemporal scales and are important for the structure and dynamics of populations, communities and ecosystems [1, 2]. In order to adequately link movement patterns and changes in landscape conditions it is necessary to track multiple individuals across broad geographic and seasonal gradients, while simultaneously accurately quantifying the dynamics of landscape resources of interest [3]. The development of lightweight tracking devices with the ability to record high resolution movement data, coupled with broad scale remote sensing data has in many cases made this possible [4, 5, 6]. However, an important challenge lies in detecting phases of movement within the complete path, as well as revealing the environmental factors that drive the emergence and persistence of these phases [7].

Under certain circumstances theory suggests that animals should move slowly and tortuously through habitats containing high quality resources [8] — a behaviour analogous to Area-Restricted Search (ARS; [9]). Patches in which movement is tortuous should be profitable habitats which provide adequate resources that increase fitness through energy acquisition, reproduction and survival. Animals should avoid areas that have negative fitness consequences by moving more quickly and linearly through them. Identifying landscape characteristics in which movements are clustered can provide insight into the factors that shape an animal’s movement through a landscape. It is important to note that these movement patterns can be confounded by several factors which may obscure the relationship between environmental resources and habitat use – including the high levels of individual variation in animal movements; competition; predation; social factors and life-stage requirements.

Resources in most ecosystems are heterogeneously distributed across space and through time [10]. They are organised within a scale-dependant hierarchy, with aggregations (patches) at smaller scales nesting into those at larger scales [11, 12]. The density and dynamics of available resources therefore depend on the scale(s) at which an animal interacts with the landscape [13]. To maximise fitness, mobile animals should be able to alter their behaviour to exploit resources at different scales [8]. Animal movement is a potentially vital mechanism for dealing with heterogeneous landscapes and movement patterns should therefore provide evidence for spatial responses [14].

The presence of water is an essential habitat resource for all waterfowl. Wetlands in semi-arid landscapes are dynamic entities and are usually in a state of flux. The landscape that southern African waterbirds inhabit is generally arid, with unpredictable timing and duration of rainfall events [15]. The dynamics of filling and drying cycles are primarily driven by the stochastic nature of rainfall events. In southern Africa dry periods are common and can last years, but these can be followed by unpredictable periods of above average rainfall [16]. This creates a spatially and temporally variable mosaic of ephemeral wetlands. Many waterbirds occurring in such areas have adapted to this variability by employing nomadic movements during parts of the year [17, 18, 19]. Egyptian Geese Alopochen aegyptiaca and Red-billed Teal Anas erythrorhyncha are two species of southern African waterfowl that adopt widespread nomadic movements [20]. This makes them ideal study species for investigating ARS behaviour in response to environmental heterogeneity.

If waterfowl perform broad scale movements and adopt ARS behaviour opportunistically when suitable resources are encountered, then variation in first-passage time should be best explained by resource conditions in the period during which waterfowl occupy a given area – termed here as the “reactive movement (RM)” hypothesis. This would suggest that movement decisions are a response to current local and immediate environmental conditions. This hypothesis implies the following two predictions: RM₁) If forage availability is an important environmental driver of movements, FPT would be positively correlated with local food biomass, measured by the normalised difference vegetation index (NDVI; [21]). Egyptian Geese are primarily grazers and thus are reliant on vegetation which may surround wetlands, while Red-Billed Teal are dabbling ducks and so rely on food resources located within the water column. We would thus expect the effect of vegetation greenness (NDVI) to be stronger for geese than for teal. RM₂) Wetlands are a primary abiotic resource for waterfowl and provide habitats for foraging, roosting, safety from predators, and moult sites. If the extent of a wetland is an important environmental driver of movements, FPT would be positively correlated with either rainfall or wetland area or a combination of both. Many ephemeral wetlands in southern Africa are shallow rain-fed depressions in which inundation, and hence wetland area, is closely tied to local precipitation events. However, an increase in wetland area is not necessarily associated with higher local rainfall. For example, floodplains can inundate following rainfall events in more distant regions of the catchment basin.

Alternatively, waterfowl may be able to structure their movements in response to changes in the magnitude and direction of resources states leading up to habitat patch occupation. There are a number of potential mechanisms that may drive this behaviour. For instance, well developed spatial memory of the landscape coupled with ability to incorporate information about local weather conditions could allow waterfowl to make movement decisions which are distinctly different to those described in the RM hypothesis. Under this hypothesis, FPT would be best explained by shifts in the magnitude and direction of resource states between the current time of occupation and a lag period prior to bird arrival - here termed the “prescient movement (PM)” hypothesis. We investigated the hypothesis with both 16 and 32 day lag periods. A positive increase in NDVI between two time periods, which reflects changes in vegetative growth, indicates an increase in food quality [22]. It has been found that younger plants have higher nutritional quality (higher nitrogen concentration) and lower levels of secondary plant chemicals. Following the Green Wave Hypothesis (GWH) it has been demonstrated that northern hemisphere geese do not select habitats with the highest biomass, but instead time their migration to take advantage of successive peaks of plant nutrition and digestibility [23, 24]. Our second hypothesis implies the following two predictions: PM₁) If waterfowl movements are a response to food quality, we would expect FPT to be higher in areas that experienced a positive change in NDVI in the 16 or 32 days prior to occupation of a patch. We would expect this effect to be more important for geese, which are grazers, than for teal, which are traditionally thought to be more reliant on invertebrate and macrophyte food resources. If support for the PM hypothesis emerged we also expected that (PM₂) the first-passage time of Red-billed Teal should be longer in areas that experienced positive changes in wetland cover and/or rainfall prior to bird arrival. For Egyptian Geese we also predicted a positive correlation between FPT and increases in rainfall and wetland area, but we expected this response to be more prominent at the 32 day lag period at which water levels start to recede and vegetation starts to grow on previously submerged shorelines.

These predictions require some additional explanation. Wetlands are dynamic entities and in many cases are either filling or drying down. These two states represent different opportunities for Egyptian Geese and Red-billed Teal, and a successional response by waterbirds to rainfall events and wetland filling has been demonstrated in arid zone systems [25, 26]. In Australia, Kingsford et al. [25] found that dabbling ducks arrive first to take advantage of the boom period, when nutrients are mobilised and dormant invertebrates emerge and reproduce. Grazing birds, conversely, have a lagged response to rainfall events and may arrive as wetlands start to dry down, utilizing terrestrial plants that colonize the drying shorelines [25]. Studies in southern Africa have recorded Red-billed Teal arriving at inundated wetlands within days of rainfall events, with numbers peaking after a couple of weeks [27, 28]. Large variation in response time appears to exist (e.g., Red-billed Teal abundance peaked 4 months following the inundation of a large river system in Namibia, [26]).

We addressed the interaction between external factors, characterised by landscape attributes, and the navigation capacity of two species of southern African waterfowl. Navigation capacity describes the ability of organisms to decide when and where to move. Effective navigation requires the ability to detect and respond to the spatial and temporal dynamics of underlying environmental conditions [7]. We first used FPT analysis to determine the scale of movement of waterfowl over yearly temporal scales across a 1500 km geographical gradient. We then explored the spatiotemporal dynamics and relative importance of abiotic and biotic variables associated with habitat resources required by waterfowl. We aimed to identify key environmental variables that influence movement behaviour.

Methods

Sites and study populations

The birds in our study population were captured at three wetland sites in South Africa and one in Zimbabwe: Strandfontein wastewater treatment works (34°05′ S, 18°20′ E); Barberspan Nature Reserve (26°33′ S, 25°37′ E); Jozini Dam (27°20′ S, 31°54′ E) and Lake Manyame (17°49′ S, 30°36′ E), respectively (see Appendices 1 and 2 for capture sites and movement paths of all individuals). Strandfontein experiences a Mediterranean climate with the majority of rainfall occurring in the winter months. Barberspan, Jozini Dam and Lake Manyame fall within summer rainfall areas, although the timing and amount of precipitation is highly variable (see [29] for further details of sites). Semi-arid conditions are common over most of southern Africa; mean rainfall over the entire region is 475 mm.

Movement data

The telemetry data were derived from Egyptian Geese and Red-billed Teal tagged with satellite GPS platform transmitter terminals (30 and 22 g PTTs respectively; Microwave Telemetry Inc., Columbia, MD, USA). PTTs were set to record a GPS location every 2 h for geese and 4 h for teal (for details of transmitter attachment methods and success rates, see [30]). Birds were tagged immediately after they had completed moult, which allowed us to confirm the wetlands as moulting sites. Birds tracked for less than 90 days were excluded from the analysis. The resulting sample size for Egyptian Geese was n = 19 and Red-billed Teal n = 14 (Table 1). Note that no teal were tagged at Jozini Dam and so data was only available for the three remaining populations. The duration between fixes in the tracks of each bird were inspected and tracks were split if the time between fixes was greater than 1 week (split tracks of each individual are denoted as either a, b or c dependent on the number of gaps detected – see Table 1). Note that all split tracks had a duration of greater than 90 days.

Table 1

Details of individual GPS-tagged Egyptian Geese (EG) and Red-billed Teal (RBT)

PTT	Spp	Site	Start	End	ND	TF	FD	NS	UDs	UDA (km2)	UDD (days)
77092	RBT	STR	3/12/2008	3/26/2009	379	1804	4.8	5	31	2.24 ± 2.42	15.8 ± 23
77093	RBT	STR	3/12/2008	9/7/2008	179	993	5.5	3	12	1.95 ± 1.24	22.1 ± 20.7
77098	RBT	STR	3/14/2008	11/24/2009	620	3550	5.7	5	44	2.26 ± 2.18	20.6 ± 30.8
77099	RBT	STR	3/14/2008	5/15/2009	427	1859	4.4	4	23	3.41 ± 3.27	20.8 ± 25.2
77100	RBT	STR	3/14/2008	4/16/2009	398	2046	5.1	6	42	1.95 ± 1.32	14.5 ± 16.9
77101	RBT	BAR	4/9/2008	9/28/2008	172	740	4.3	3	11	3.42 ± 2.19	22.2 ± 22.6
77102	RBT	BAR	4/10/2008	4/20/2010	740	4155	5.6	4	82	1.54 ± 0.98	13.5 ± 24
77103	RBT	MAN	5/5/2008	8/24/2008	111	610	5.5	3	33	2.57 ± 2.73	6 ± 7.6
77104	RBT	MAN	5/5/2008	1/25/2009	265	1431	5.4	3	43	1.79 ± 1.74	10.3 ± 21.3
77106	RBT	MAN	5/6/2008	7/25/2009	445	2587	5.8	7	34	2.3 ± 2.09	19.4 ± 25.2
77108	RBT	MAN	5/6/2008	8/29/2008	115	644	5.6	4	22	2.67 ± 3.1	10.5 ± 11.9
77109	RBT	MAN	5/7/2008	12/24/2008	231	1307	5.7	4	37	1.88 ± 1.12	12.3 ± 23.2
77112	RBT	BAR	6/7/2008	5/15/2009	342	1843	5.4	3	23	2.15 ± 2.49	21.8 ± 43.6
77115	RBT	BAR	10/11/2008	7/15/2009	277	1429	5.2	4	20	2.16 ± 1.37	11.6 ± 15
77094	EG	STR	1/12/2008	5/9/2008	118	1218	10.3	4	15	1.99 ± 1.14	12.1 ± 9.6
77094a	EG	STR	8/20/2008	5/1/2009	254	2686	10.6	3	9	2.59 ± 2.29	21.2 ± 34.3
77095	EG	STR	1/12/2008	1/3/2009	357	3397	9.5	5	32	2.29 ± 1.93	15.5 ± 22.3
7711702	EG	JOZ	5/4/2012	9/20/2012	139	1669	12.0	5	13	2.25 ± 1.24	13.3 ± 13.9
7711802	EG	STR	1/17/2009	10/11/2010	632	6453	10.2	5	58	2.56 ± 2.93	18.1 ± 31.3
7712002	EG	JOZ	5/4/2012	5/24/2013	385	4317	11.2	2	86	3.12 ± 3.06	11.5 ± 17.7
7712002a	EG	JOZ	6/9/2013	1/31/2014	236	2592	11.0	3	43	1.93 ± 1.6	10.9 ± 15.9
7712102	EG	JOZ	5/5/2012	9/3/2012	121	1309	10.8	3	4	2.21 ± 1.19	30.5 ± 19.8
7712202	EG	BAR	10/23/2008	5/30/2009	219	2123	9.7	5	37	1.87 ± 1.7	8.5 ± 13.4
7712302	EG	STR	12/5/2008	6/2/2009	179	1756	9.8	4	7	2.53 ± 1.26	23.9 ± 30.7
77125	EG	MAN	5/7/2008	2/21/2010	655	6965	10.6	5	120	2.29 ± 2.98	11.6 ± 24.1
77125a	EG	MAN	4/17/2010	5/31/2011	409	3689	9.0	3	60	2.68 ± 2.76	15.6 ± 21.4
77126	EG	MAN	5/7/2008	12/26/2008	233	2682	11.5	5	29	2.33 ± 2.86	12.6 ± 16.5
77127	EG	BAR	6/7/2008	5/10/2010	702	6351	9.0	6	54	2.7 ± 3.32	17.3 ± 26.9
77128	EG	BAR	6/22/2008	6/6/2009	349	3551	10.2	3	31	1.81 ± 1.9	10.4 ± 25.4
77128a	EG	BAR	8/15/2009	5/25/2010	283	2551	9.0	3	64	1.81 ± 1.9	6.5 ± 14.3
77128b	EG	BAR	9/25/2010	5/6/2011	223	1998	9.0	3	28	1.67 ± 0.96	5.7 ± 12.1
77128c	EG	BAR	7/31/2011	12/2/2011	124	653	5.3	4	16	2.09 ± 1.4	8.8 ± 11.5
77129	EG	BAR	6/7/2008	5/15/2009	342	3491	10.2	5	61	2.13 ± 1.73	9.6 ± 17.5
77130	EG	BAR	11/9/2008	9/19/2009	314	2489	7.9	5	76	2.2 ± 1.76	8.6 ± 12.6
77130a	EG	BAR	10/4/2009	6/4/2010	243	2140	8.8	5	78	1.86 ± 1.5	6.5 ± 10.8
77132	EG	BAR	6/7/2008	5/30/2009	357	2601	7.3	3	35	2.01 ± 1.94	15.5 ± 25.4
77132a	EG	BAR	8/13/2009	4/14/2010	244	2090	8.6	2	12	2.15 ± 0.46	14.8 ± 27.3
7713301	EG	STR	12/4/2008	4/27/2009	144	1506	10.5	5	40	1.83 ± 1.2	4 ± 5.1
7713302	EG	JOZ	5/4/2012	2/19/2013	291	3009	10.3	3	27	2.81 ± 2.17	15.7 ± 23
77134	EG	STR	12/1/2008	7/29/2010	605	5330	8.8	4	19	3.28 ± 2.47	24.2 ± 28.8
77134a	EG	STR	8/19/2010	5/2/2011	256	2561	10.0	5	17	2.04 ± 2.16	19.2 ± 29.9
77134b	EG	STR	7/22/2011	4/12/2012	265	2401	9.1	3	8	2.08 ± 1.66	47.1 ± 46.2
77135	EG	STR	12/1/2008	2/8/2011	799	8522	10.7	4	107	2.5 ± 2.82	11.5 ± 18

PTT, transmitter identity; BAR, Barberspan; STR, Strandfontein; MAN, Lake Manyame; JOZ, Jozini Dam. Start and end date refers to the time period of tracking data used in the study (ND is the total tracking duration in days). Total fixes (TF) are the number of relocations recorded over the study period, while mean fixes per day (FD) are the total number of relocations divided by the number of days the transmitter was active. The remaining columns contain data from the results of the FPT analysis. NS, number of movement segments per track; UDs, number of utilisation distributions derived from all segments; UDA, mean (± sd) area of utilisation distributions; UDD, mean (± sd) number of days spent in each utilisation distribution polygon

First-passage time analysis

A graphical example of the analytical steps for the movement path of an individual bird (Red-Billed Teal 77115 from Barberspan) is shown in Fig. 1. First-passage time is calculated at each GPS fix along a movement path (Fig. 1a) as the time taken to cross a circle of a given radius [31]. The process is repeated over a range of circles with differing radii. The peaks in variance of log transformed FPT at a specific radius (Fig. 1b) indicates the scale at which an animal’s movements are clustered and hence the spatial scale at which ARS behaviour is occurring [31]. In other words, the peaks correspond to a specific circle radius in which more tortuous and intensive movements are performed. As mean FPT is increases with circle size, we applied a common radius to all bird movement paths of each species, to allow for comparisons of individual bird FPTs. For each species we used the radius at which mean variance of log FPT showed a peak. Following Le Corre, Dussault and Côté [32], FPT was calculated along an individual’s path with a given radius r, ranging from 100 to 10 000 m at 80 m intervals, centred on consecutive locations. The radius r _max is the radius at which the variance of log transformed FPT var _fpt reaches a maximum. The mean of variance var _{fpt mean} was calculated by averaging var _fpt of each bird at each radius. The peak in this mean var _{fpt mean} was then taken at a population average and used as the common spatial scale for all subsequent analysis.

Once FPT analysis was applied to each individual, plots were created of GPS fixes against FPT (Fig. 1c). Lavielle’s segmentation method was then applied in order to identify homogenous movement bouts within an individual’s movement path using the lavielle function in the adehabitatLT R package [33, 34]. The method aims to detect breakpoints in the movement path by minimizing a penalised contrast function [35]. Given that a movement path is made up of K segments, the method searches for an optimal number of segments K _opt with which to partition the movement path. There should be a clear break in the decrease of the contrast function after K _opt, which we identified in two ways. Firstly the break and the corresponding K _opt, was visually detected from the plot of the contrast function. Secondly, K _opt was automatically detected by choosing the last value of K at which the second derivative of the standardised contrast function is greater than a threshold S. Following the recommendation of Lavielle [36], S was set to 0.75. These methods were used in conjunction with one another to determine the number of segments for each individual’s movement path (Fig. 1c and d).

Segments from each movement path (i.e. paths from all individuals across each site) were extracted and processed in the following way: GPS fixes within segments were used to create utilisation distribution which defined an area and time over which environmental variables could be measured (Fig. 1e). Utilisation distributions were calculated with Movement-based Kernel Density Estimator (MKDE) methods [37] using the BRB function within the adehabitatHR R package [33]. In order to reduce the effects of autocorrelation inherent in the data, we calculated a mean first-passage time value at the peak radius (mFPT _Rmax) derived from all GPS fixes contained within the home-range polygon. If the home range was made up of multiple polygons, as is common when using MKDE methods, mFPT _Rmax was calculated for each polygon individually. Each polygon was then used as a sampling unit in which mFPT _Rmax was the response variable and the set of environmental variables measured in that polygon were the explanatory variables (Fig. 1e).

Environmental data

In order to evaluate the relationship between FPT and environmental conditions, 12 variables were calculated for each home range polygon for each bird. Extraction of environmental variables was performed using Google Earth Engine (https://earthengine.google.com/), a cloud platform for the analysis of geospatial data. NDVI, rainfall, water surface coverage (modified normalised difference water index; mNDWI), elevation and temperature were calculated as mean values over the time t for which the home range was occupied (Table 2). Two extra sets of NDVI, rainfall and mNDWI variables were calculated: 1) the difference between mean at time t and t-16 days and 2) the difference between mean at time t and t-32 days. A positive change in NDVI between time periods would indicate and increase in vegetative growth and hence food quality, while a positive change in mNDWI would indicate an increase in wetland extent. Time lags were chosen to correspond with the minimum temporal resolution of the predictor variables (constrained by the resolution of NDVI which is based on 16 day composites – see Table 2 for derivations and sources of data). Dynamic variables were first averaged temporally for the duration of home range occupation, followed by spatial averaging. Geographical location of capture sites was added as a predictor variable to evaluate a study area effect on first-passage times. This included four- and three-level categorical variables for Egyptian Geese and Red-billed Teal respectively. Temperature and elevation were added as fixed effects in some candidate models to assess whether there was an effect of thermal stress on movement behaviour. We developed a set of 36 singular and multi-term candidate models to evaluate our competing hypotheses (Table 3).

Table 2

Details of environmental variables used as predictors in the analysis of first-passage time (FPT) of two waterfowl species in southern Africa

		Source	Units	Spatial resolution	Temporal resolution	Earth Engine Layer
NDVI	Normalised difference vegetation index NDVI = (NIR – R)/(NIR + R)	MODIS terra		250 m	16-day composite	MODIS/MCD43A4_NDVI
mNDWI	Modified normalised difference water index mNDWI = (G-MIR)/(G + MIR)	MODIS terra		250 m	16-day composite	MODIS/MCD43A4
Precip	Rainfall	TRMM	mm/h	0.25°	3 hourly	TRMM/3B42
Temp	Land surface temperature	MODIS	Kelvin (Convert to °C)	1 km	Daily	MODIS/MYD11A1
Elev	Elevation	NASA	Meters above sea level	0.9 km	N/A	CGIAR/SRTM90_V4

Table 3

Candidate set of generalised linear mixed models used to investigate the relationship between mean first-passage time (mFPT_Rmax) and environmental variables of Egyptian Geese and Red-billed Teal. Individual birds (ID) were added as a random effect to all models. See Table 2 and text in Methods for derivation of environmental predictor variables

Model			Model formula
1		Food quantity	mFPT_Rmax ~ NDVI_t + (1\|ID)
2		Food quality (16 day)	mFPT_Rmax ~ ΔNDVI _t-16 + (1\|ID)
3		Food quality (32 day)	mFPT_Rmax ~ ΔNDVI _t-32 + (1\|ID)
4		Wetland cover	mFPT_Rmax ~ mNDWI_t + (1\|ID)
5		Wetland cover change (16 day)	mFPT_Rmax ~ ΔmNDWI_t-16 + (1\|ID)
6		Wetland cover change (32 day)	mFPT_Rmax ~ ΔmNDWI _t-32 + (1\|ID)
7		Precipitation	mFPT_Rmax ~ Precip_t + (1\|ID)
8		Precipitation (16 day)	mFPT_Rmax ~ ΔPrecip_t-16 + (1\|ID)
9		Precipitation (32 day)	mFPT_Rmax ~ ΔPrecip_t-32 + (1\|ID)
10		Site	mFPT_Rmax ~ site + (1\|ID)
11		Temperature & elevation	mFPT_Rmax ~ Temp + Elev + (1\|ID)
RM hypothesis:^a
12 (16)		Food quantity, wetland cover & precipitation (+ temperature & elevation, site)	mFPT_Rmax ~ NDVI_t + mNDWI_t + Precip_t + (1\|ID)
13 (17)		Food quantity & wetland cover (+ temperature & elevation, site)	mFPT_Rmax ~ NDVI_t + mNDWI_t + (1\|ID)
14 (18)		Food quantity & precipitation (+ temperature & elevation, site)	mFPT_Rmax ~ NDVI_t + Precip_t + (1\|ID)
15 (19)		Wetland cover & precipitation (+ temperature & elevation, site)	mFPT_Rmax ~ mNDWI_t + Precip_t + (1\|ID)
PM hypothesis:^b
20 (24)	16 day Lag	Change in food quality, wetland cover & precipitation (+ site)	mFPT_Rmax ~ ΔNDVI _t-16 + ΔmNDWI _t-16 + ΔPrecip_t-16 + (1\|ID)
21 (25)		Change in food quality & wetland cover (+ site)	mFPT_Rmax ~ ΔNDVI _t-16 + ΔmNDWI _t-16 + (1\|ID)
22 (26)		Change in food quality & precipitation (+ site)	mFPT_Rmax ~ ΔNDVI _t-16 + ΔPrecip_t-16 + (1\|ID)
23 (27)		Change in wetland cover change & precipitation (+ site)	mFPT_Rmax ~ ΔmNDWI _t-16 + ΔPrecip_t-16 + (1\|ID)
28 (32)	32 day Lag	Change in food quality, wetland cover & precipitation (+ site)	mFPT_Rmax ~ ΔNDVI _t-32 + ΔmNDWI _t-32 + ΔPrecip_t-32 + (1\|ID)
29 (33)		Change in food quality & wetland cover (+ site)	mFPT_Rmax ~ ΔNDVI _t-32 + ΔmNDWI _t-32 + (1\|ID)
30 (34)		Change in food quality & precipitation (+ site)	mFPT_Rmax ~ ΔNDVI _t-32 + ΔPrecip_t-32 + (1\|ID)
31 (35)		Change in wetland cover change & precipitation (+ site)	mFPT_Rmax ~ ΔmNDWI _t-32 + ΔPrecip_t-32 + (1\|ID)
36		Null	mFPT_Rmax ~ 1 + (1\|ID)

RM reactive movement; PM prescient movement

^aModels numbers in parentheses comprise of the same set of environmental predictors with extra addition of site, temp and elevation as a predictor variables

^bModels numbers in parentheses comprise of the same set of environmental predictors with extra addition of site as a predictor variable

Statistical analyses

Generalised linear mixed models were used to model the relationship between mFPT _Rmax and environmental variables using the lmer function from the R package lme4 [38]. Data were first screened for normality, and outliers were removed. mFPT _Rmax was log-transformed prior to inclusion into candidate models. Individual birds were added as random effect to allow for estimation of population level regression coefficients while accounting for variation between individuals. All predictor variables were scaled before inclusion into the models. This allowed for standardisation of parameter estimates and comparison of their magnitudes. Spatial auto-correlation in the residuals of the chosen models was examined in two ways. First, we visually examined spatial plots of the magnitude and signs of residuals; and second, we used semi-variograms to quantify variance as a function of distance between points (bubble and variogram function from gstat R package, [39]). Model selection followed evaluation of candidate models from AIC criteria [40]. Variance inflation factors were used to test for the presence of collinearity amongst predictor variables. R² _GLMM was used as a measure of overall fit for the selected models [41].

Results

After calculating var _fpt for each individual against radius (Figs. 2 and 3), the mean radius r _max for Egyptian Geese and Red-billed Teal were identified as 2180 and 2420 m. The results of the subsequent FPT analysis showed that the number of movement segments ranged from 2 to 7 (mean 4). The number of utilisation distribution polygons derived from the segments ranged from 4 to 120 (mean 38.2). The area of those utilisation distribution polygons ranged from 1.5 to 3.4 km² (mean 2.3 km²). The number of days spent in a specific utilisation distribution polygon ranged from 4 to 47 (mean 15). See Table 1 for the above values of each individual.

Fig. 2
Curves of individual Egyptian Geese showing the variance in log first-passage time against circle radius. Panels correspond to individuals tagged at 4 different wetland sites. PTT, transmitter identity; BAR, Barberspan; STR, Strandfontein; MAN, Lake Manyame; JOZ, Jozini Dam

Fig. 3
Curves of individual Red-Billed Teal showing the variance in log first-passage time against circle radius. Panels correspond to individuals tagged at 3 different wetland sites. PTT, transmitter identity; BAR, Barberspan; STR, Strandfontein; MAN, Lake Manyame

In our analysis of environmental predictors of mFPT _Rmax, the most parsimonious model with the highest support included difference in NDVI over a 32 day lag, difference in rainfall over a 32 day lag, and geographical location for both Egyptian Geese and Red-billed Teal (Table 4). Following the AIC criteria and model selection procedure we employed, model 34 had the greatest support and suggested that variation in mFPT _Rmax was best explained by changes food quality, amount of rainfall, and the geographical location of individuals. The variances explained by the fixed effects (marginal R²) of the chosen models were 9.4 and 11.3 %, while the variances explained by both fixed and random effects (conditional R²) were 17.2 and 13.1 % for Egyptian Geese and Red-billed Teal respectively (Table 5). The individual level variability was noticeably higher in Egyptian Geese (7.8 %) compared to that of Red-billed Teal (1.8 %). The candidate models representing the reactive movement hypothesis were noticeably absent from the top four and nine movement models of Egyptian Geese and Red-billed Teal respectively.

Table 4

Comparisons of the top models (∆AIC_c < 20) of first-passage time as a function environmental variables of two waterfowl species in southern Africa

Model		K	AIC_c	ΔAIC_c	AIC_c Wt	Cum Wt
	Egyptian Geese
32	mFPT_Rmax ~ ΔNDVI _t-32 + ΔmNDWI _t-32 + ΔPrecip_t-32 + site + (1\|ID)	9	1979.25	0.00	0.43	0.43
34	mFPT_Rmax ~ ΔNDVI _t-32 + ΔPrecip_t-32 + site + (1\|ID)	8	1979.39	0.14	0.40	0.83
30	mFPT_Rmax ~ ΔNDVI _t-32 + ΔPrecip_t-32 + (1\|ID)	5	1982.95	3.70	0.07	0.90
28	mFPT_Rmax ~ ΔNDVI _t-32 + ΔmNDWI _t-32 + ΔPrecip_t-32 + (1\|ID)	6	1983.44	4.19	0.05	0.95
19	mFPT_Rmax ~ mNDWI_t + Precip_t + site + (1\|ID)	10	1984.52	5.27	0.03	0.98
16	mFPT_Rmax ~ NDVI_t + mNDWI_t + Precip_t + site + (1\|ID)	11	1986.00	6.75	0.01	1.00
15	mFPT_Rmax ~ mNDWI_t + Precip_t + (1\|ID)	5	1992.27	13.02	0.00	1.00
	Red-billed Teal
34	mFPT_Rmax ~ ΔNDVI _t-32 + ΔPrecip_t-32 + site + (1\|ID)	7	833.87	0.00	0.38	0.38
26	mFPT_Rmax ~ ΔNDVI _t-16 + ΔPrecip_t-16 + site + (1\|ID)	7	834.79	0.92	0.24	0.62
32	mFPT_Rmax ~ ΔNDVI _t-32 + ΔmNDWI _t-32 + ΔPrecip_t-32 + site + (1\|ID)	8	835.21	1.34	0.19	0.82
24	mFPT_Rmax ~ ΔNDVI _t-16 + ΔmNDWI _t-16 + ΔPrecip_t-16 + site + (1\|ID)	8	836.83	2.95	0.09	0.90
30	mFPT_Rmax ~ ΔNDVI _t-32 + ΔPrecip_t-32 + (1\|ID)	5	837.70	3.83	0.06	0.96
28	mFPT_Rmax ~ ΔNDVI _t-32 + ΔmNDWI _t-32 + ΔPrecip_t-32 + (1\|ID)	6	839.54	5.67	0.02	0.98
22	mFPT_Rmax ~ ΔNDVI _t-16 + ΔPrecip_t-16 + (1\|ID)	5	840.40	6.53	0.01	0.99
20	mFPT_Rmax ~ ΔNDVI _t-16 + ΔmNDWI _t-16 + ΔPrecip_t-16 + (1\|ID)	6	842.45	8.57	0.01	1.00
35	mFPT_Rmax ~ ΔmNDWI _t-32 + ΔPrecip_t-32 + site + (1\|ID)	7	852.07	18.19	0.00	1.00

Models are ranked based on differences in the corrected Akaike’s Information Criteria (ΔAIC_c) Akaike weights (AIC_c Wt). K is the number of estimated parameters and Cum Wt is the cumulative weight of sequential models. Individual birds (ID) were added as a random effect to all models

Table 5

Summary of the generalised mixed models with the highest support in the analysis of mean first-passage time (mFPT_Rmax) as a function of environmental variables

	Parameter	β	Lower CI	Upper CI	SE	R² _GLMM(m)	R² _GLMM(c)	VIF (kappa)
Egyptian Geese						9.4 %	17.2 %	3.64
	(Intercept)	1.34	1.24	1.45	0.06
	ΔNDVI _t-32	0.04	0.01	0.06	0.02
	ΔPrecip_t-32	0.12	0.09	0.14	0.02
	Site : JOZ vs. BAR	0.04	−0.14	0.23	0.11
	Site : MAN vs. BAR	−0.13	−0.33	0.07	0.12
	Site : STR vs. BAR	0.25	0.10	0.40	0.09
No. of observations: 1165, random effect groups: ID, 29
Red-billed Teal						11.3 %	13.1 %	4.42
	(Intercept)	1.59	1.47	1.71	0.07
	ΔNDVI _t-32	−0.06	−0.11	−0.01	0.03
	ΔPrecip_t-32	0.16	0.12	0.21	0.03
	Site : MAN vs. BAR	−0.12	−0.27	0.03	−0.11
	Site : STR vs. BAR	0.24	0.08	0.39	0.09
No. of observations: 445, random effect groups: ID, 14

R² values are measures of model fit based on fixed effects only (marginal variance, R² _GLMM(m)) and on the full model including random effects (conditional variance, R² _GLMM(c))

For Egyptian Geese, differences in both NDVI and rainfall in the 32 days prior to arrival were significantly and positively correlated with mFPT _Rmax, supporting predictions PM₁ and PM₂ (Fig. 4 and Table 5). The magnitude of the effect of rainfall was three times higher than that of NDVI. Individuals from the Strandfontein population had a significantly higher mFPT _Rmax than those of the Barberspan, which was the reference category. The parameter estimates for birds from Jozini and Lake Manyame were not significantly different from zero. For Red-billed Teal, the difference in NDVI during the 32 days prior to arrival was negatively correlated to mFPT _Rmax while rainfall was significantly and positively correlated with FPT (Table 5). Again the magnitude of the effect of rainfall was higher than that of NDVI. Individual teal from the Strandfontein population had a significantly higher mFPT _Rmax than those of the Barberspan population, while parameter estimates for birds from Lake Manyame were not significantly different from zero. We found little support for effects of temperature, elevation or mNDWI in explaining variation in FPTs There was little evidence of spatial autocorrelation in the semi-variograms and bubble plots. The kappa statistic was less than 10 for models of both species, indicating an absence of collinearity in the predictor variables.

Fig. 4
Three panels which represent a gradient of mean values of mean first-passage time (mFPT_Rmax) and environmental variables within utilisation distribution polygons. These data illustrate the positive relationship between FPT and a 32-day lag in NDVI and precipitation (e.g. dark red, dark green and dark blue polygons represent sites at which FPT and environmental variables were strongly positively correlated). ΔNDVI_t-32, the difference between mean NDVI within a polygon at time t and *t-32 days*; ΔPrecip_t-32, the difference between precipitation within a polygon at time t and *t-32 days* measured in mm. Note that mFPT_Rmax was measured in hours and has been logged transformed

Discussion

Our findings show little evidence for the reactive movement hypothesis; instead, waterfowl appeared to respond to shifts in resource conditions in a given area based on changes in magnitude and direction of environmental variables between preceding lag periods and current periods of occupation. This suggests that movement decisions were potentially more complex than those that would result from randomly sampling the landscape and ceasing movement when suitable conditions were encountered. While it may be difficult to identify exactly how waterfowl perceive their landscapes, spatial awareness and prior experience may be mechanisms that might allow waterfowl to capitalise on high quality resources. Egyptian Geese spent more time in areas which had increased primary productivity and associated increases in rainfall in the 32 day period leading up to goose arrival; results which were consistent with predictions PM₁ and PM₂. The magnitude of the effect of rainfall was stronger than that of NDVI. By contrast, Red-billed Teal spent more time in areas in which primary productivity had decreased over the previous 32 days and where rainfall had increased, consistent with PM₂ but not with PM₁. Again, rainfall had a stronger effect than NDVI. Geographical location (site variable) of individuals was a significant predictor of waterfowl movement behaviour; individuals of both species had higher mean FPTs in the Strandfontein populations compared with the Barberspan populations. Temperature and elevation had no significant effect on FPT. These findings suggest both that thermoregulatory constraints do not play a role in structuring movements and that movements are not clustered in coastal regions of southern Africa. The environmental variable that represented a change in wetland extent (∆mNDWI) appeared in several of the competing models, but was not included in the final model with the highest support. This indicates that there is a potential effect of either filling or drying of wetlands on FPT, however these dynamics do not dominate the way in which waterfowl movements are structured.

The moderate amount of variance explained by the models could have resulted from several unmeasured factors affecting landscape use. For instance, waterfowl form large aggregations outside of breeding periods, and so social factors such as competition may affect the choice of habitat used. Human disturbance and predation pressure are also likely to significantly influence habitat choice and movements [42]. Another potential issue which might have influenced the explained variance is the choice of method used to delineate the geographic area over which environmental conditions were measured. We used the widely adopted movement-based KDE method which has its foundation in point-based methods. Traditional KDE methods, however, may significantly underestimate the size of utilisation distributions [43, 44]. Measuring environmental variables over a broader spatial extent would, therefore, change the calculated landscape conditions that the waterfowl would have experienced. In turn, this has the potential to alter the outcomes of the first-passage time models.

Our findings that higher FPTs of Egyptian Geese are a response to increases in primary productivity, as opposed to standing biomass, are in accordance with several studies of migratory movements of herbivorous waterfowl occurring at high latitudes in the northern hemisphere. These movements are linked to plant phenology and follow the predictions set out by the GWH, which states that waterfowl time their spring migration to take advantage of successive peaks of forage quality along their migration routes [23, 24, 45, 46, 47]. Although semi-nomadic waterfowl, living in low productivity environments where the distribution of resources is patchier [48], have different constraints in terms of locating resources to those of migrants (i.e. lack of distinct and predictable seasonal changes), they seem to prioritise forage quality in a similar manner. Responding to such changes, however, requires that waterfowl have some sort of prior knowledge of the state of landscape resources and do not simply perform random searches through the landscape to settle where conditions are suitable.

For birds living in semi-arid areas, there are trade-offs between when to stay and when to leave an area [49]. The ability of waterfowl in our study system to possess spatial awareness could allow them to capitalise on highly nutritious food sources and leave areas when nutritional quality starts to decline, providing an adaptive advantage through periods of resource uncertainty. The role of spatial memory in movement has recently received attention [50, 51, 52, 53] and there are indeed fitness benefits of memory in heterogeneous landscapes of intermediate complexity. Waterfowl may employ a similar strategy to that of other nomadic birds (e.g., Snail Kites [54] and Pacific Black Ducks [55]) in that exploratory movements are adopted through periods of high resource abundance. This would allow waterfowl to attain a level of familiarity with high quality resource patches, avoiding the need to search extensively when resource abundance is low. Memory and prior knowledge therefore have the potential to be particularly relevant to waterfowl movement strategies in arid landscapes [56]. Indeed, in this study Egyptian Geese adopted behaviour that allowed them to respond to food quality in a similar manner to that of migrant geese following the green wave.

Kingsford et al. [25] developed a conceptual model of the movement, breeding and feeding response of five arid-zone waterbird functional groups (dabbling and diving ducks, herbivores, piscivores, large waders and small waders). They proposed that grazing and dabbling (or invertebrate) feeders should have different temporal responses to rainfall and wetland filling events. Dabbling ducks should arrive first to take advantage of invertebrates which have hatched following wetland inundation, while grazers should lag in their response to capitalise on terrestrial or aquatic plants which grow and germinate more slowly. Egyptian Geese responded to rainfall at a 32 day lag period which is what we would expect for a grazing bird feeding on emergent vegetation on freshly exposed shorelines.

Observational studies of Red-billed Teal responses to rainfall events in southern Africa support ideas that invertebrate feeders should respond quickly [27, 28]. Data from longer term studies, however, suggest that this peak in abundance occurs at a much longer lag period, with a peak occurring at 4 months post rainfall [26]. Our finding that teal responded to rainfall at a 32 day lag rather than a 16 day lag might be explained by their niche breadth. Petrie [57] showed that teal have considerable dietary flexibility. During energetically demanding periods such as the breeding season, for example, invertebrate consumption was < 14 % of total food items while the majority of remaining energy requirements was satisfied by the consumption of native grass seeds which surround wetlands. The relationship between NDVI and FPT in Red-billed Teal was opposite to that found in Egyptian Geese. First-passage time was higher in areas in which primary productivity had decreased over a 32 day lag period. It is possible that there was a higher abundance of grass seeds available as growth decreased, providing teal with an adequate food resource. These results suggest that responses to rainfall events vary considerably across apparently similar arid zone landscapes and that dietary flexibility may drive changes in movement responses between species.

One question that remains enigmatic is which cues waterfowl use to detect distant rainfall events. It has been proposed that they might be able to sense rain fronts, but the evidence that waterfowl respond to lag variables and not immediate conditions suggests that they have some knowledge of landscape conditions and can make decisions based on environmental cues [55]. There is evidence of a similar response in other species. Red-billed Quelea Quelea quelea, small granivorous passerines in southern Africa, are able to respond to dynamic changes in resources [58]; they appear to move ahead of rainfall events and then track back towards areas in which rain has fallen to take advantage of grass seeds.

Both Egyptian Geese and Red-billed Teal had higher than average FPTs in utilisation distributions in the Strandfontein population than in the Barberspan population. Although there was no significant difference between the other sites, Jozini had slightly higher FPT than Barberspan while the Lake Manyame population had slightly lower average FPT (Table 5). It is interesting to note that the direction and magnitude of these patterns were consistent between both these species, which indicates the important influence of landscape conditions on movement in comparison to differences in life history and ecological traits between the two species. Differing responses to environmental variation by populations of a species has been shown in several instances [59, 60, 61]. This follows a theoretical prediction that increased variation of movement responses within a species range should be associated with increased variability of resources at broad landscape scales [55, 59, 62]. This is indeed evident in our study as Barberspan and Strandfontein occur in noticeably different landscapes. Barberspan lies in an arid summer rainfall region, whereas Strandfontein is a winter rainfall region with less variability in the timing and amount of precipitation (Appendix 3). The landscape surrounding Strandfontein is characterised by a high density of grain producing agricultural land. Associated with these farms are small dams used for storage, many of which have stable water levels throughout the year. On the other hand, areas into which many of the individuals from the Barberspan population moved were more arid, with agricultural land separated by semi-deserts (Appendices 1, 2 and 3). Spatiotemporal correlation of resources was thus higher near Strandfontein and could mean there is less need for waterfowl to move long distances, resulting in higher first-passage times. This illustrates the range in strategies of nomadic movement, which is proposed to be an outcome of spatiotemporal correlation in landscape resources [63]. Differences in movement behaviour (measured by, for example, parameters such as daily movement rates, distance moved, turning angles) in these populations has previously been demonstrated [64], indicating that populations of Egyptian Geese at Barberspan and Strandfontein move in different ways, while little separated the movements of different populations of Red-billed Teal. In our analyses, however, the data showed clear differences in patterns of FPT (Figs. 2 and 3, Table 1). This suggests that analysing movements in the FTP framework can provide new as well as complementary insights into existing drivers of waterfowl movement. Additionally, it is important to recognise the variation of individuals within the same population in understanding population level processes [65, 66].

Conclusions

We were able to undertake the first quantitative analysis of the interaction between external factors and navigational capacity of southern African waterfowl in the context of a current movement ecology framework. More generally, we have shown the utility of linking long term telemetry data over broad geographic scales with environmental conditions experienced by multiple individuals to uncover the proximate drivers of waterfowl movement. The analysis of movement using the FPT method allowed us to conclude that waterfowl movements in southern Africa are a response to the dynamics of rainfall and primary productivity. In addition, our findings suggest that waterfowl movements are not simply reactive but rather involve mechanisms which allow waterbirds to integrate information of the local landscape in order to take advantage of productive habitats. Future research should take the form of a more detailed analysis of movement and changes in resources to further understand the mechanism underlying the prescient movement hypothesis.

Notes

Acknowledgements

This research was supported by a GAINS (Global Avian Influenza Network for Surveillance) sub-contract from USAID, via the Wildlife Conservation Society, to GC. Additional funding was provided by the DST/NRF Centre of Excellence at the Percy FitzPatrick Institute, an NRF Incentive Grant to GC, and the University of Cape Town.

References

1.
Hanski IA. Metapopulation ecology. Oxford: Oxford University Press; 1999.Google Scholar
2.
Turchin P. Quantitative analysis of movement: measuring and modeling population redistribution in animals and plants. Sunderland: Sinauer Associates; 1998.Google Scholar
3.
van Moorter B, Bunnefeld N, Panzacchi M, Rolandsen CM, Solberg EJ, Sæther B-E. Understanding scales of movement: animals ride waves and ripples of environmental change. J Anim Ecol. 2013;82:770–80.CrossRefGoogle Scholar
4.
Avgar T, Mosser A, Brown GS, Fryxell JM. Environmental and individual drivers of animal movement patterns across a wide geographical gradient. J Anim Ecol. 2013;82:96–106.CrossRefGoogle Scholar
5.
Trierweiler C, Mullié WC, Drent RH, Exo K-M, Komdeur J, Bairlein F, et al. A Palaearctic migratory raptor species tracks shifting prey availability within its wintering range in the Sahel. J Anim Ecol. 2013;82:107–20.CrossRefGoogle Scholar
6.
Boettiger AN, Wittemyer G, Starfield R, Volrath F, Douglas-Hamilton I, Getz WM. Inferring ecological and behavioral drivers of African elephant movement using a linear filtering approach. Ecology. 2011;92:1648–57.CrossRefGoogle Scholar
7.
Nathan R, Getz WM, Revilla E, Holyoak M, Kadmon R, Saltz D, et al. A movement ecology paradigm for unifying organismal movement research. Proc Natl Acad Sci U S A. 2008;105:19052–9.CrossRefGoogle Scholar
8.
Fauchald P, Tveraa T. Hierarchical patch dynamics and animal movement pattern. Oecologia. 2006;149:383–95.CrossRefGoogle Scholar
9.
Kareiva P, Odell G. Swarms of predators exhibit “preytaxis” if individual predators use area-restricted search. Am Nat. 1987;130:233–70.CrossRefGoogle Scholar
10.
Johnson A, Wiens J, Milne B, Crist T. Animal movements and population dynamics in heterogeneous landscapes. Landsc Ecol. 1992;7:63–75.CrossRefGoogle Scholar
11.
Wiens J. Population responses to patchy environments. Annu Rev Ecol Syst. 1976;81–120.Google Scholar
12.
Levin SA. The problem of pattern and scale in ecology : the Robert H. MacArthur Award Lecture. Ecology. 1992;73:1943–67.CrossRefGoogle Scholar
13.
Wiens J, Milne B. Scaling of “landscapes” in landscape ecology, or, landscape ecology from a beetle’s perspective. Landsc Ecol. 1989;3:87–96.CrossRefGoogle Scholar
14.
McIntire EJB, Fajardo A. Beyond description: the active and effective way to infer processes from spatial patterns. Ecology. 2009;90:46–56.CrossRefGoogle Scholar
15.
Tyson PB, Preston-Whyte RA. The weather and climate of Southern Africa. Oxford: Oxford University Press; 2000.Google Scholar
16.
Siegfried W. Wildfowl distribution, conservation and research in southern Africa. Wildfowl. 1970;21:89–98.Google Scholar
17.
Roshier D, Klomp N, Asmus M. Movements of a nomadic waterfowl, Grey Teal Anas gracilis, across inland Australia-results from satellite telemetry spanning fifteen months. Ardea. 2006;94:461–75.Google Scholar
18.
Underhill LG, Tree AJ, Oschadleus HD, Parker V. Review of ring recoveries of waterbirds in Southern Africa. Cape Town: Avian Demography Unit, University of Cape Town; 1999.Google Scholar
19.
Oatley T, Prys-Jones R. A comparitve analysis of movements of southern African watefowl (Anatidae), based on ringing recoveries. South Afr J Wildl Res. 1986;16:1–6.Google Scholar
20.
Hockey PAR, Dean WRJ, Ryan P. Roberts Birds of Southern Africa. 7th ed. Cape Town: Trustees of the John Voelcker Bird Book Fund; 2005.Google Scholar
21.
Pettorelli N, Vik JO, Mysterud A, Gaillard J-M, Tucker CJ, Stenseth NC. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends Ecol Evol. 2005;20:503–10.CrossRefGoogle Scholar
22.
Doiron M, Legagneux P, Gauthier G, Lévesque E. Broad-scale satellite Normalized Difference Vegetation Index data predict plant biomass and peak date of nitrogen concentration in Arctic tundra vegetation. Appl Veg Sci. 2013;16:343–51.CrossRefGoogle Scholar
23.
van der Graaf A, Stahl J, Klimkowska A, Bakker JP, Drent RH. Surfing on a green wave – how plant growth drives spring migration in the Barnacle Goose Branta leucopsis. Ardea. 2006;94:567–77.Google Scholar
24.
Shariatinajafabadi M, Wang T, Skidmore AK, Toxopeus AG, Kölzsch A, Nolet BA, et al. Migratory herbivorous waterfowl track satellite-derived green wave index. PLoS One. 2014;9:e108331.CrossRefGoogle Scholar
25.
Kingsford R, Roshier D, Porter J. Australian waterbirds - time and space travellers in dynamic desert landscapes. Mar Freshw Res. 2010;61:875–84.CrossRefGoogle Scholar
26.
Cumming GS, Paxton M, King J, Beuster H. Foraging guild membership explains variation in waterbird responses to the hydrological regime of an arid-region flood-pulse river in Namibia. Freshw Biol. 2012;57:1202–13.CrossRefGoogle Scholar
27.
Herrmann E, Anderson MD, Seaman M. Occurrence and abundance of waterbirds at an ephemeral pan in the Northern Cape Province, South Africa. Ostrich. 2004;75:275–84.CrossRefGoogle Scholar
28.
Simmons RE, Barnard P, Jamieson IG. What precipitates influxes of wetland birds to ephemeral pans in arid landscapes? Observations from Namibia. Ostrich. 1999;70:145–8.CrossRefGoogle Scholar
29.
Cumming GS, Caron A, Abolnik C, Cattoli G, Bruinzeel LW, Burger CE, et al. The ecology of influenza A viruses in wild birds in southern Africa. Ecohealth. 2011;8:4–13.CrossRefGoogle Scholar
30.
Cumming GS, Ndlovu M. Satellite telemetry of Afrotropical ducks: methodological details and assessment of success rates. African Zool. 2011;46(October):425–34.CrossRefGoogle Scholar
31.
Fauchald P, Tveraa T. Using first-passage time in the analysis of area-restricted search and habitat selection. Ecology. 2003;84:282–8.CrossRefGoogle Scholar
32.
Le Corre M, Dussault C, Côté SD. Detecting changes in the annual movements of terrestrial migratory species: using the first-passage time to document the spring migration of caribou. Mov Ecol. 2014;2:19.CrossRefGoogle Scholar
33.
Calenge C. The package “adehabitat” for the R software: a tool for the analysis of space and habitat use by animals. Ecol Modell. 2006;197:516–9.CrossRefGoogle Scholar
34.
R Core Team: R. A language and environment for statistical computing. 2013.Google Scholar
35.
Lavielle M. Using penalized contrasts for the change-point problem. Signal Process. 2005;85:1501–10.CrossRefGoogle Scholar
36.
Lavielle M. Detection of multiple changes in a sequence of dependent variables. Stoch Process their Appl. 1999;83:79–102.CrossRefGoogle Scholar
37.
Benhamou S, Riotte-Lambert L. Beyond the utilization distribution: identifying home range areas that are intensively exploited or repeatedly visited. Ecol Modell. 2012;227:112–6.CrossRefGoogle Scholar
38.
Bates D, Maechler M, Bolker B, Walker S. lme4: Linear mixed-effects models using Eigen and S4_. R package version 1.1-7. 2014.Google Scholar
39.
Pebesma E. Multivariable geostatistics in S: the gstat package. Comput Geosci. 2004;30:683–91.CrossRefGoogle Scholar
40.
Burnham K, Anderson D. Model selection and multi-model inference: a practical information-theoretic approach. 2nd ed. Berlin: Springer; 2002.Google Scholar
41.
Nakagawa S, Schielzeth H. A general and simple method for obtaining R2 from generalized linear mixed-effects models. Methods Ecol Evol. 2013;4:133–42.CrossRefGoogle Scholar
42.
Cumming GS, Harebottle DM, Mundava J, Otieno N, Tyler SJ. Timing and location of reproduction in African waterfowl: an overview of > 100 years of nest records. Ecol Evol 2016. doi: 10.1002/ece3.18.
43.
Fleming CH, Fagan WF, Mueller T, Olson KA, Leimgruber P, Calabrese JM. Rigorous home range estimation with movement data: a new autocorrelated kernel density estimator. Ecology. 2015;96:1182–8.CrossRefGoogle Scholar
44.
Fleming CH, Calabrese JM, Mueller T, Olson KA, Leimgruber P, Fagan WF. From Fine-Scale Foraging to Home Ranges: A Semivariance Approach to Identifying Movement Modes across Spatiotemporal Scales. Am Nat. 2014;183:E154–67.CrossRefGoogle Scholar
45.
van der Graaf AJ, Stahl J, Veen GF, Havinga RM, Drent RH. Patch choice of avian herbivores along a migration trajectory–From Temperate to Arctic. Basic Appl Ecol. 2007;8:354–63.CrossRefGoogle Scholar
46.
Bos D, Drent R, Rubinigg M, Stahl J. The relative importance of food biomass and quality for patch and habitat choice in Brent Geese Branta bernicla. Ardea. 2005;93:5–16.Google Scholar
47.
van Wijk RE, Kölzsch A, Kruckenberg H, Ebbinge BS, Müskens GJDM, Nolet BA. Individually tracked geese follow peaks of temperature acceleration during spring migration. Oikos. 2012;121:655–64.CrossRefGoogle Scholar
48.
Roshier D, Robertson A, Kingsford R, Green D. Continental-scale interactions with temporary resources may explain the paradox of large populations of desert waterbirds in Australia. Landsc Ecol. 2001;16:547–56.CrossRefGoogle Scholar
49.
Dean W, Barnard P, Anderson M. When to stay, when to go: trade-offs for southern African arid-zone birds in times of drought. S Afr J Sci. 2009;105:24–8.CrossRefGoogle Scholar
50.
Mueller T, Fagan WF. Search and navigation in dynamic environments - from individual behaviors to population distributions. Oikos. 2008;117:654–64.CrossRefGoogle Scholar
51.
Fagan WF, Lewis MA, Auger-Méthé M, Avgar T, Benhamou S, Breed G, et al. Spatial memory and animal movement. Ecol Lett. 2013;16:1316–29.CrossRefGoogle Scholar
52.
Seidel DP, Boyce MS. Patch-use dynamics by a large herbivore. Mov Ecol. 2015;3:1–10.CrossRefGoogle Scholar
53.
Van Moorter B, Visscher D, Benhamou S, Börger L, Boyce MS, Gaillard JM. Memory keeps you at home: a mechanistic model for home range emergence. Oikos. 2009;118:641–52.CrossRefGoogle Scholar
54.
Bennetts RE, Kitchens WM. Factors influencing movement probabilities of a nomadic food specialist: proximate foraging benefits or ultimate gains from exploration? Oikos. 2000;91:459–67.CrossRefGoogle Scholar
55.
Mcevoy JF, Roshier DA, Ribot RFH, Bennett ATD. Proximate cues to phases of movement in a highly dispersive waterfowl, Anas superciliosa. Mov Ecol. 2015;3:13–5.CrossRefGoogle Scholar
56.
Roshier DA, Asmus M, Klaassen M. What drives long-distance movements in the nomadic Grey Teal Anas gracilis in Australia? Ibis (Lond 1859). 2008;150:474–84.CrossRefGoogle Scholar
57.
Petrie SA. Red-billed teal foods in semiarid South Africa: a north-temperate contrast. J Wildl Manage. 1996;60:874–81.CrossRefGoogle Scholar
58.
Cheke R, Tratalos J. Migration, patchiness, and population processes illustrated by two migrant pests. Bioscience. 2007;27:145–54.CrossRefGoogle Scholar
59.
Singh N, Börger L, Dettki H, Bunnefeld N, Ericsson G. From migration to nomadism: movement variability in a northern ungulate across its latitudinal range. Ecol Appl. 2012;22:2007–20.CrossRefGoogle Scholar
60.
Singh NJ, Grachev IA, Bekenov AB, Milner-Gulland EJ. Tracking greenery across a latitudinal gradient in central Asia - the migration of the saiga antelope. Divers Distrib. 2010;16:663–75.CrossRefGoogle Scholar
61.
Mandel JT, Bohrer G, Winkler DW, Barber DR, Houston CS, Bildstein KL. Migration path annotation: cross-continental study of migration-flight response to environmental conditions. Ecol Appl. 2011;21:2258–68.CrossRefGoogle Scholar
62.
Roshier DA, Doerr VA, Doerr ED. Animal movement in dynamic landscapes: interaction between behavioural strategies and resource distributions. Oecologia. 2008;156:465–77.CrossRefGoogle Scholar
63.
Jonzén N, Knudsen E, Holt RD. Uncertainty and predictability: the niches of migrants and nomads. In: Milner-Gulland EJ, Fryxell JM, Sinclair ARE, editors. Animal migration: a synthesis. Oxford: Oxford University Press; 2011. p. 91–109.Google Scholar
64.
Cumming GS, Gaidet N, Ndlovu M. Towards a unification of movement ecology and biogeography: conceptual framework and a case study on Afrotropical ducks. J Biogeogr. 2012;39:1401–11.CrossRefGoogle Scholar
65.
Hawkes C. Linking movement behaviour, dispersal and population processes: is individual variation a key? J Anim Ecol. 2009;78:894–906.CrossRefGoogle Scholar
66.
Doerr ED, Doerr VAJ. Dispersal range analysis: quantifying individual variation in dispersal behaviour. Oecologia. 2005;142:1–10.CrossRefGoogle Scholar

Copyright information

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Authors and Affiliations

Dominic A. W. Henry
- 1
Email author
Judith M. Ament
- 2
- 3
Graeme S. Cumming
- 1
- 4

1.Percy FitzPatrick Institute, DST/NRF Centre of Excellence, University of Cape TownRondeboschSouth Africa
2.Centre for Biodiversity and Environment Research, Department of Genetics, Evolution and EnvironmentUniversity College LondonLondonUK
3.Institute of Zoology, Zoological Society of LondonLondonUK
4.ARC Centre of Excellence for Coral Reef Studies, James Cook UniversityTownsvilleAustralia

Exploring the environmental drivers of waterfowl movement in arid landscapes using first-passage time analysis

Abstract

Background

Results

Conclusions

Keywords

Background

Methods

Sites and study populations

Movement data

First-passage time analysis

Environmental data

Statistical analyses

Results

Discussion

Conclusions

Notes

Acknowledgements

References

Copyright information

Authors and Affiliations

Personalised recommendations

Cite article