The need to account for Argos Doppler errors when estimating habitat use from Argos Doppler locations: evidence and solution

doi:10.21203/rs.3.rs-2826242/v1

Download PDF

Method Article

The need to account for Argos Doppler errors when estimating habitat use from Argos Doppler locations: evidence and solution

https://doi.org/10.21203/rs.3.rs-2826242/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background.

Due to the large errors in Argos Doppler location estimates, Argos-based satellite transmitter data are rarely used in studies of fine-scale habitat selection by animals. Novel state-space models (SSMs) for path reconstruction from animal movement data improve location estimates, delivering refined estimations of an animal’s most likely path and, also, re-estimating the uncertainties for each location. However, the SSM-refined uncertainties are still relatively large and the true locations of animals tracked with PTTs (Platform terminal transmitters) remain impossible to determine. We suggest an approach that uses the SSM-refined location uncertainties to quantify the probabilities of an animal’s occurrence in each habitat and infer which of the habitats it most likely visited.

Methods.

We test the performance of our approach against habitat use assays based on most likely locations from raw Argos Doppler estimates and Argos Doppler estimates refined with an SSM. For this, we combine a GPS tracking dataset (2214 location fixes) from one individual and an Argos-PTT tracking dataset (1708 location points) from 14 individual Continental Black-tailed Godwits (Limosa limosa limosa) breeding in agricultural grasslands in The Netherlands utilizing both simulations and empirical data to assess habitat use.

Results.

The approach that accounted for location uncertainties on top of a state-space model improved habitat assignments in the simulation study by 5% compared with only the SSM-refined Argos location points and by 23% compared with the raw Argos locations. We provide working code in R that can be reproduced for the analysis of habitat selection of animals followed with PTTs.

Conclusions.

Low-precision tracking data may be suitable to study habitat selection if location uncertainties are taken into account. The approach presented here has the potential to considerably improve the validity of such analyses, opening up new opportunities for the use of Argos Doppler data in analyses of habitat selection by animals. Since Argos Doppler location uncertainty parameters are required for the inference of the most likely used habitat, it is crucial that users acquire this information from Collecte Localisation Satellites (CLS) when initiating a new study.

Argos Doppler error ellipse

location uncertainty

habitat use

CTMM state-space model

path reconstruction

semi-major

ellipse orientation

location accuracy

tracking

home range

Platform terminal (or Argos Doppler) transmitters (hereafter, ‘PTTs’) operated by Collecte Localisation Satellites (CLS, www.argos-system.org) are among the oldest and most widely used tracking devices for collecting animal movement data [1–3]. PTTs offer many benefits, such as their relatively light weight, near real-time data transmission, and durability. That said, the Argos Doppler locations reported by PTTs come with an elliptical error that extends hundreds of meters away from the estimated location [4, 5]. For this reason, there may be a large discrepancy between the reported location and the (unobserved) true location.

The low precision of Argos location estimates argues against their use in fine-scale spatial analyses such as those of animal habitat use [6, 7]. Even though the most likely location estimated for an individual at a point in time (Fig. 1; large, filled orange dot) may fall within one habitat (e.g., habitat A), the probability that the animal was instead in a different habitat (e.g., habitat B) can be fairly high. The anisotropy of Argos Doppler error can additionally — and significantly — affect the representation of habitats within the occurrence probability surface. Thus, it is critical to account for the elliptical location error when making inferences on a scale smaller than the scale of the location error.

Despite their low precision, raw Argos locations have long been used as an approximation of true locations [3]. This practice can be acceptable for the analysis of larger scale movements, such as the seasonal movements of migratory animals [8–11]. However, when the scale of inference starts to fall within the range of the location uncertainty, using raw Argos locations is no longer appropriate. For example, assigning the habitat use of tracked animals when the distance between different habitat types is smaller than the Argos error cannot be done with certainty (Fig. 2).

In recent years, state-space models (SSMs) [12] have appeared as a tool for refining Argos locations. SSMs reconstruct movement paths by linking an observation model that accounts for location error with a movement process model. McClintock et al. [13] were the first to incorporate Argos error ellipses from the multimodel Kalman filter algorithm [14, 15] into an SSM for animal movement. Over time, this led to the development of R-packages [16] that implement SSMs capable of handling the elongated structure of Argos error. These packages include ‘aniMotum’ [17] and its predecessor ‘foieGras’ [18], ‘crawl’ [19], and ‘ctmm’ [20], most of which were reviewed by Joo et al. [21]. This class of models has proved to be a powerful tool for the analysis of imprecise movement data, considerably improving location estimates and providing refined location uncertainties [5, 22].

Despite the fact that SSMs account for Argos location uncertainties, their refined posterior location estimates still include errors resembling the anisotropy of the original Argos Doppler error. As a result, SSM-improved Argos locations cannot be considered true locations and need to be analysed in a way that takes into account their specific error structure [5, 7]. This presents several challenges, particularly for analyses of habitat use. Home range size, for example, is highly sensitive to location error [23–25], yet SSM-improved Argos locations are often used directly in home range estimations [26–28]. Even error-informed approaches to estimating home ranges may assume a circular error distribution around locations [29], calling into question their compatibility with Argos tracking data [30].

To improve habitat use analyses that rely on Argos locations, we have developed a ‘most likely habitat’ approach that quantifies how likely it is that an animal used a certain habitat by taking into account the uncertainty of the measurements provided by the Argos system. We demonstrate the performance of our most likely habitat approach with a simulation study and illustrate a recommended workflow and reproducible code with a case-study of Continental Black-tailed Godwits (Limosa limosa limosa). Our insights into Argos location errors also raise several practical points with respect to the collection and implementation of PTT data, such as with online services like Movebank (www.movebank.org). In combination, we believe that our approach can help expand the robust use of Argos tracking data in the study of animal movements.

Study system

In this study, we used multiple approaches, including our most likely habitat approach, to infer the habitat use of Continental Black-tailed Godwits tracked with PTT’s during their breeding season in southwest Friesland, The Netherlands. Continental Black-tailed Godwits (hereafter, ‘godwits’) mainly spend the nonbreeding season in West Africa [31]. In spring, they return to Western Europe to breed, mostly in agricultural grasslands. The Dutch godwit population is declining due to high levels of nest failure and brood mortality related to the intensification of agricultural practices [32–35]. We inferred the habitat use of godwits by interpolating tracking data with an agricultural land-use intensity map of southwest Friesland (52°55’N, 5°60’E; see [36]). This map is based on satellite imagery from 2016 and summarizes the land-use intensity of these grasslands – low, intermediate, and high intensity – based on their agricultural use.

Most likely habitat approach

Our most likely habitat approach uses SSM-refined Argos locations and their uncertainties to quantify the likelihood of occurrence in the available habitats. First, for each tracked individual, we fit SSMs to their Argos tracking data with the R package ‘ctmm’ [20]. These state-space models separate the continuous movement process from the discrete-time sampling process, which makes them robust to irregular sampling and autocorrelation [20]. We then used the best movement process model to re-generate (i.e., simulate) an individual’s path 100 times while taking the location uncertainties into account. These 100 path estimates vary slightly and for each original Argos point provide a random sample of 100 points that follows the probability density estimated by the SSM.

By overlaying these locations with a habitat map, we obtained a sample of habitats for each estimated location. Finally, based on this sample, we inferred the most likely habitat used by an animal at each time point for which we had an Argos location estimate.

Quantifying the imprecision of Argos location estimates

Nowadays, along with the traditional location quality class, Argos location data include parameters of the elliptical Argos Doppler location error : semi-major axis, semi‐minor axis, and orientation (0-180° angle from North toward the East; Fig. 2). The axes’ lengths usually range within several hundred meters. The location error follows a bivariate normal probability distribution [14, 15], meaning that the likelihood of occurrence of an animal is normally distributed along the two axes, semi-major and semi-minor, with the highest likelihood in the centre of the ellipse (i.e., the reported Argos location).

To illustrate the discrepancy between reported Argos Doppler locations and the (unobserved) true location, we have simulated how far from a reported Argos location the true location might be given its elliptical location error. For the simulation we combined locations of one GPS track (obtained in the spring-summer of 2013) with uncertainties extracted from the Argos dataset that contained the tracks of 14 godwits tracked during the breeding season (15 March – 26 July 2020) with solar powered PTT’s (Microwave Telemetry, Inc.). This dataset contained 3443 locations from within The Netherlands. We removed (1) 385 locations that did not contain uncertainty estimates, (2) 300 locations which had an Argos error radius larger than 1500 m (location class (‘LC’) 0; Table 1) and (3) 1050 locations obtained with less than 4 messages (‘LC A’ and ‘B’; Table 1). The final dataset contained 1708 locations. Of the locations with uncertainty estimates, the proportion in each location class is presented in Table 1.

Table 1

**The number and proportion of location points of different quality classes in the data.** The shaded cells contain location classes that were removed for the analysis.
Location quality class (and error radii)	LC 3 (< 250 m)	LC 2 (250–500 m)	LC 1 (500–1500 m)	LC 0 (> 1500 m)	LC B	LC A
Number and proportion of fixes	387 (11.2%)	488 (14.2%)	833 (24.2%)	300 (8.7%)	739 (21.5%)	311 (9.0%)

Simulating Argos locations

We demonstrate the performance of the most likely habitat approach with both a simulation study and a case study. We simulate Argos locations by taking ‘true’ locations from a GPS track of a godwit to which we add Argos location error parameters from the Argos dataset described above.

The GPS track used in this simulation study was of a female godwit captured on her nest on 14 May 2013 and outfitted with an UvA-BiTS GPS transmitter [22, 37]. For the simulation study, we only used the 2214 locations that were collected before the bird left the study area on 16 June 2013. This segment covered the last two days of incubation (the clutch hatched on 16 May) and the post-hatching period. During this time, the bird was observed on several occasions. The presence of a brood was not confirmed by visual observations.

Figure 2 illustrates this process: (1) we take a GPS location (grey square) and (2) add an Argos error probability distribution randomly drawn from an Argos location dataset (grey ellipse). We then (3) mimic the deviation of a reported Argos location from a true location by sampling a random point from the error probability distribution (large orange dot) and then (4) transfer the earlier drawn error ellipse to this location estimate (orange ellipse).

Comparing habitat assignment approaches

For the simulated Argos track, we compared the assigned habitat on the basis of the (1) raw Argos locations, (2) SSM-refined locations, and (3) most likely habitat approach with the habitat type of the original GPS locations. For this, we overlaid the locations obtained with all three methods with a habitat map of the area (Fig. 3). Doing so allowed us to quantify the performance of the different approaches by comparing the proportion of locations for which the inferred habitat type corresponded to the ‘true’ habitat type recorded by GPS fix.

In the raw Argos approach, we overlayed the generated tracking locations with the habitat map to extract the habitat at each location. For the SSM approach, we used the ctmm.select function in the ‘ctmm’ package. This function automatically fits and selects among several continuous-time stochastic process movement models. The Ornstein-Uhlenbeck (OU) anisotropic error model fit our data best. We overlayed the ctmm-predicted locations with the habitat map to extract the habitat at each location. In the most likely habitat approach, we derived a set of up to 100 habitat samples for each location fix because in this approach the path gets simulated by the SSM 100 times. From these 100 habitat samples, we estimated the most likely habitat (Fig. 3). For all three approaches, only location points falling within the mapped area (~ 11,500 ha; [36]) were included in the analysis.

In a case-study, we implemented the three approaches to analyze an Argos Doppler track of an individual godwit (259 location points), obtained in the same area in 2020. This bird was most likely breeding, as it stayed within a small area during the breeding season.

Argos Doppler estimated location imprecision

The predicted error distance ranged from 1 to 92802 m, with a median of 653 m and a mean of 1621 m (Fig. 4). The error distance was > 1500 m for 25% of the reported locations (Fig. 1).

Simulation study

The output of the three approaches, together with the GPS reference, are presented on a map fragment (Fig. 5) and as a box-whisker plot (Fig. 6). Accounting for the location uncertainties improved the outcome of the analysis (Fig. 6). Habitat selection estimates based on the raw Argos approach (with simulated locations) coincided with the true habitat revealed by GPS data for 35% (median value, 1st interquartile range 34.6–36.0%) of location estimates. The SSM approach facilitated the coincidence in 53.8% (52.5–54.7%) of location points. The most likely habitat approach, accounting for the location uncertainties on top of the state-space model, additionally increased the coincidence rate up to the 58% (57.2–59.2%; Fig. 6).

Case study on habitat use by an individual Black-tailed Godwit followed with a PTT

Even though the most likely habitat approach involves a larger number of steps, it remains relatively compact (i.e., less than 200 rows of code, including loading necessary packages and data preparation). Running the analysis for the 259 locations of the example track took tens of minutes.

The analysis of habitat selection by a PTT-tracked godwit with all three approaches identified the same habitat (botanically rich agricultural lands maintained at low intensity) as the most visited (see figures in Supplementary materials). For 18 of the 255 fixes (7.1%) in the track, raw Argos-estimated most likely locations fell within terrain that was not classified as one of the three habitat classes (e.g., deep water or road) and thus resulted in no habitat type assignment. For SSM-improved Argos locations, only one fix had no habitat assignment. The most likely habitat approach provided habitat scores for all the fixes.

Our study introduces a user-friendly, transparent, and methodologically correct approach to estimating the habitat use by animals tracked with PTTs. In this approach, the output of a state-space model applied to Argos tracking data is refined by taking into account the bivariate normal distribution of the Argos error, with the major benefit of quantifying a probability of the assignment of habitat types for each location from an individual track. In the simulation study, our most likely habitat approach outperformed two other approaches to estimating habitat use — with raw Argos-estimated locations and SSM-derived data — by 23% and 5%, respectively. The most likely habitat approach has thus improved estimations of habitat selection over previously available approaches. More importantly, the approach provided the probability of use for each habitat at each location point, making our inferences more robust.

Absolute Performance

The absolute accuracy of the approach in our simulation study, however, was rather low, with the rate of coincidence between the simulated Argos-based habitat estimates and GPS assignments reaching only ~ 58%. This result was obtained using the rule of thumb where the habitat with the highest probability (frequency in a sample) was assigned the most likely used habitat for the location point, even when its frequency was 51% against 49% for another habitat. We, therefore, recommend leveraging the strength of our most likely habitat approach that is, the assessment of probability of an animal’s occurrence in each of the possible habitats. As illustrated below, such conscious and qualified habitat assignments can potentially mitigate the bias from the large location errors and improve results.

Bias from the error ellipse and ways to handle it

Unlike the raw Argos and SSM approaches, the most likely habitat approach assigned a habitat to 100% of the location estimates in our simulation study. This was the case because patches without habitat assignment generally had areas smaller than or comparable to the areas covered by location error ellipses. As a result, even when the most likely location fell within a patch without habitat type assignment, the error ellipse always included locations that were outside the patch and did have a habitat type assignment. There could be three ‘realities’ underlying these outputs: (1) the godwit visited an area adjacent to the habitat without assignment, (2) the godwit was in flight, quickly passing over the habitat without assignment when the location fix was taken, or (3) the godwit was present in the habitat without assignment. In the first case, the outcome of our approach may be considered an improvement over the other two approaches, while the second and third illustrate potential biases.

The third situation occurred in our study when the tracked godwit visited a shallow pond, which is a frequently-used habitat for godwits (Fig. 5) but is unassigned in our habitat classification. This particular pond site contains GPS locations in high density and is the main source of the ‘unassigned’ habitat type (Fig. 6A, red star at the rightmost box). A representative number of locations generated by the most likely habitat approach also fell within this and other ‘unassigned’ habitat types (the rightmost blue box in Fig. 6A). However, because the pond is small relative to the size of the ellipse and the surrounding habitat types, the number of points falling within the pond never exceeded the number of points falling within one of the three assigned habitat types bordering the pond.

This case illustrates the bias resulting from fine-scale habitat structure combined with uneven representation of habitats in a study area. Similar biases likely also occur in most other studies where animals use finely delineated habitats but employ large step lengths (the distance between consecutive points), i.e. when the step lengths of the movements of the tracked animals are generally large relative to the habitat patch size.

Such biases can be mitigated through: 1) aggregation of data by generating statistics not per location estimate, but for clusters of fixes within an animal’s track (such as during a certain time period across multiple days throughout the breeding season); 2) direct use of the proportional distribution of habitat types in the samples of locations (blue boxes in Fig. 6A), instead of just treating the most represented habitat as the most likely one; and, 3) applying weights to the habitats according to their representation within the study plot or the likely degree to which they are used by a species [38].

While the most likely habitat approach focuses on quantification of habitat use, in many cases, researchers are interested in habitat selection. Selection can best be inferred with resource selection functions, which account for the proportion of habitats in the study area. The simulated possible tracks generated from the ‘ctmm’ package, for instance, can be used for parametrisation of these functions. The task of quantifying habitat selection lays outside of the scope of this paper. However, we recommend paying attention to the resource selection functions that are currently being developed for Argos Doppler data within the ‘ctmm’ package (C. Fleming, pers comm.). In general, no matter the species, habitat matrix, or methodological approach, quantification of habitat use patterns based on Argos telemetry data should be interpreted with caution, keeping in mind the low precision of the locations provided by Argos.

Autocorrelation and sampling bias. While autocorrelation-caused bias is mitigated with ‘ctmm’, unrepresentative sampling due to duty-cycling or patterns in satellite passages has to be addressed in each particular case. This can be done by assigning weights to the habitat information obtained during different phases of a duty cycle: upweighting observations that occur during under-sampled times, while downweighting observations occurring during over-sampled times [30].

The importance of obtaining all necessary Argos data parameters

Although accounting for elliptical location uncertainties is crucial in the analyses of habitat use by animals tracked with PTT devices [13, 39], acquisition of the uncertainty parameters is not a default option in the CLS user contract. Backdated uncertainty parameter information can be ordered from Argos for extra charge, but we encourage scientists using animal tracking to make sure that they acquire this information from the beginning of their new tracking projects. We also recommend incorporation of Argos location uncertainty information in the Environmental-Data Automated Track Annotation (EnvDATA, [40]) system implemented in Movebank [41]. Ensuring both that the initial users, as well as future users, have access to these data will represent a big step for the field.

Our most likely habitat approach is aimed at facilitating the use of location error information provided by the Argos system and increasing the scientific quality of assays of habitat use by animals tracked with PTT transmitters. The main advantage of our approach is that, for each location fix, it provides a sample of potentially visited habitats. This allows the user to infer the likelihood of occurrence of the animal in each of the habitats and determine which habitat was the most likely visited by the animal. This is fundamentally different from, and methodologically more appropriate than, inferences based solely on the habitat associated with an individual’s most likely location. The most likely habitat approach can thus increase the utility of the previously retrieved Argos locations as well as new and diverse location types, which are critical for studying the movement of animals and supplying accurate habitat use information for conservation and management efforts.

PTT – Platform Terminal Transmitter

CLS - Collecte Localisation Satellites

SSM – State-space Model

GPS – Global Positioning System

LC – Location Quality Class

Ethics approval and consent to participate

This study meets the legal requirements of capturing, handling, and attaching tracking devices to black-tailed godwits in The Netherlands. Trapping, flag and transmitter attachment, and handling were overseen by under licence numbers DEC-6350A/C/G/H & CCD- AVD105002017823, following the Dutch Animal Welfare Act Articles 9 and 11.

Consent for publication

Not applicable.

Availability of data and materials

The datasets and R code supporting the conclusions of this article are available on Github and Zenodo. The habitat map is available as supporting materials of [36].

Competing interests

The authors declare that they have no competing interests.

Funding

The desktop part of work was supported by the Netherlands Ministerie van Landbouw, Natuurbeheer en Voedselveiligheid and the LIFE-IP GrassBirdHabitats project (LIFE/IPE/19/000004), MAVA (Waders of Bijagos). Funding for the data collection Spinoza Premium Award 2014 to TP from The Netherlands Organisation for Scientific Research (NWO), with additional funding from the TOP Grant “Shorebirds in Space” (NWO) 2011 to TP and funding by private benefactors to Global Flyway Network (2018), and the Gieskes-Strijbis Fonds (2018). The funders had no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.

Authors’ Contributions

ER and TP conceived the study. TP acquired funding for the fieldwork and the analyses. JCEWH and TP coordinated the fieldwork on black-tailed godwits. NRS, MAV and AHJL deployed GPS and PTT transmitters on the godwits and acquired the tracking and observation data on the birds. RAH provided the habitat map of the study area. ER has developed the approach. JK and ER designed and coordinated the study. JK implemented and ran the analyses and drafted the manuscript. JL finalized writing and tested the code. All authors contributed to writing and critically revised the manuscript. All authors read and approved the final manuscript.

Acknowledgements
We would like to thank the members of our field crews, students and volunteers for their help with catching and tagging. Special thanks go to Wiebe Kaspersma and Alice McBride, who spent nearly all their free time helping us in the field, and to Bote de Boer, Jan de Jong and Murk Nijdam, who also aided us in times of need. We thank the individual farmers, private land owners, conservation organizations (It Fryske Gea & Staatsbosbeheer) and local farming collective (Collectief Sudwestkust), who granted us access to their fields. Dr. Wouter Vansteelant kindly reviewed and discussed our approach and R code. We also thank Alice McBride for literary editing of several discussion sections.

Jouventin P, Weimerskirch H. Satellite tracking of Wandering albatrosses | Nature. Nature. 1990;343:746–8.
McConnell BJ, Chambers C, Nicholas KS, Fedak MA. Satellite tracking of grey seals (Halichoerus grypus). Journal of Zoology. 1992;226:271–82.
Douglas D, Weinzierl R, Davidson S, Kays R, Wikelski M, Bohrer G. Moderating Argos location errors in animal tracking data. Methods in Ecology and Evolution. 2012;3:999–1007.
Nicholls DG, Road N, Robertson CJR, Murray MD. Measuring accuracy and precision for CLS:Argos satellite telemetry locations. Notornis. 2007;54:137–57.
Lowther AD, Lydersen C, Fedak MA, Lovell P, Kovacs KM. The Argos-CLS Kalman Filter: Error Structures and State-Space Modelling Relative to Fastloc GPS Data. PLOS ONE. Public Library of Science; 2015;10:e0124754.
Hooten MB, Johnson DS, McClintock BT, Morales JM. Animal movement: statistical models for telemetry data. Boca Raton: CRC Press; 2017.
Rozylowicz L, Bodescu FP, Ciocanea CM, Gavrilidis AA, Manolache S, Matache ML, et al. Empirical analysis and modeling of Argos Doppler location errors in Romania. PeerJ. PeerJ Inc.; 2019;7:e6362.
Wood SN. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society: Series B (Statistical Methodology). 2011;73:3–36.
Rakhimberdiev E, Duijns S, Karagicheva J, Camphuysen CJ, Dekinga A, Dekker R, et al. Fuelling conditions at staging sites can mitigate Arctic warming effects in a migratory bird. Nat Commun. Nature Publishing Group; 2018;9:4263.
Bom RA, Conklin JR, Verkuil YI, Alves JA, de Fouw J, Dekinga A, et al. Central-West Siberian-breeding Bar-tailed Godwits (Limosa lapponica) segregate in two morphologically distinct flyway populations. Ibis. 2022;164:468–85.
Piersma T, Kok EMA, Hassell CJ, Peng H-B, Verkuil YI, Lei G, et al. When a typical jumper skips: itineraries and staging habitats used by Red Knots (Calidris canutus piersmai) migrating between northwest Australia and the New Siberian Islands. Ibis. 2021;163:1235–51.
Patterson TA, Thomas L, Wilcox C, Ovaskainen O, Matthiopoulos J. State–space models of individual animal movement. Trends in Ecology & Evolution. 2008;23:87–94.
McClintock BT, London JM, Cameron MF, Boveng PL. Modelling animal movement using the Argos satellite telemetry location error ellipse. Methods in Ecology and Evolution. 2015;6:266–77.
Service Argos. Argos User’s Manual [Internet]. CLS; 2016. Available from: https://www.argos-system.org/manual/
Lopez R, Malardé J-P. Improving ARGOS Doppler location using Kalman filtering [Internet]. CLS-DT-MEMO-11-65. CLS Group; 2011. Available from: https://www.argos-system.org/wp-content/uploads/2016/07/r282_9_technical_paper.pdf
R Core Team. R: A language and environment for statistical computing [Internet]. Vienna, Austria; 2021. Available from: https://www.R-project.org/
Jonsen ID, Grecian WJ, Phillips L, Carroll G, McMahon C, Harcourt RG, et al. aniMotum, an R package for animal movement data: Rapid quality control, behavioural estimation and simulation. Methods in Ecology and Evolution [Internet]. John Wiley & Sons, Ltd; 2023 [cited 2023 Feb 10]; Available from: https://besjournals.onlinelibrary.wiley.com/doi/10.1111/2041-210X.14060
Jonsen ID, Patterson TA, Costa DP, Doherty PD, Godley BJ, Grecian WJ, et al. A continuous-time state-space model for rapid quality control of argos locations from animal-borne tags. Movement Ecology. 2020;8:31.
Johnson DS, London JM, Lea M-A, Durban JW. Continuous-Time Correlated Random Walk Model for Animal ℡emetry Data. Ecology. 2008;89:1208–15.
Calabrese JM, Fleming CH, Gurarie E. ctmm: an r package for analyzing animal relocation data as a continuous-time stochastic process. Methods in Ecology and Evolution. 2016;7:1124–32.
Joo R, Boone ME, Clay TA, Patrick SC, Clusella‐Trullas S, Basille M. Navigating through the r packages for movement. Journal of Animal Ecology. 2020;89:248–67.
Rakhimberdiev E, Senner NR, Verhoeven MA, Winkler DW, Bouten W, Piersma T. Comparing inferences of solar geolocation data against high-precision GPS data: annual movements of a double-tagged black-tailed godwit. Journal of Avian Biology. 2016;47:589–96.
Hoenner X, Whiting SD, Hindell MA, McMahon CR. Enhancing the Use of Argos Satellite Data for Home Range and Long Distance Migration Studies of Marine Animals. PLOS ONE. Public Library of Science; 2012;7:e40713.
Christin S, St-Laurent M-H, Berteaux D. Evaluation of Argos Telemetry Accuracy in the High-Arctic and Implications for the Estimation of Home-Range Size. PLOS ONE. Public Library of Science; 2015;10:e0141999.
Thomson JA, Börger L, Christianen MJA, Esteban N, Laloë J-O, Hays GC. Implications of location accuracy and data volume for home range estimation and fine-scale movement analysis: comparing Argos and Fastloc-GPS tracking data. Mar Biol. 2017;164:204.
Shaver DJ, Hart KM, Fujisaki I, Bucklin D, Iverson AR, Rubio C, et al. Inter-nesting movements and habitat-use of adult female Kemp’s ridley turtles in the Gulf of Mexico. PLOS ONE. Public Library of Science; 2017;12:e0174248.
Poessel SA, Uher-Koch BD, Pearce JM, Schmutz JA, Harrison A-L, Douglas DC, et al. Movements and habitat use of loons for assessment of conservation buffer zones in the Arctic Coastal Plain of northern Alaska. Global Ecology and Conservation. 2020;22:e00980.
Lamont MM, Johnson D, Catizone DJ. Home ranges and Movements of Two Diamondback Terrapins (Malaclemys terrapin macrospilota) in Northwest Florida. Estuaries and Coasts. 2021;44:1484–91.
Kranstauber B, Kays R, LaPoint SD, Wikelski M, Safi K. A dynamic Brownian bridge movement model to estimate utilization distributions for heterogeneous animal movement. Journal of Animal Ecology. 2012;81:738–46.
Silva I, Fleming CH, Noonan MJ, Alston J, Folta C, Fagan WF, et al. Autocorrelation-informed home range estimation: A review and practical guide. Methods in Ecology and Evolution. 2022;13:534–44.
Verhoeven MA, Loonstra AHJ, McBride AD, Both C, Senner NR, Piersma T. Migration route, stopping sites, and non-breeding destinations of adult Black-tailed Godwits breeding in southwest Fryslân, The Netherlands. J Ornithol. 2021;162:61–76.
Kruk M, Noordervliet MAW, ter Keurs WJ. Hatching dates of waders and mowing dates in intensively exploited grassland areas in different years. Biological Conservation. 1996;77:213–8.
Kruk M, Noordervliet MAW, ter Keurs WJ. Survival of black-tailed godwit chicks Limosa limosa in intensively exploited grassland areas in The Netherlands. Biological Conservation. 1997;80:127–33.
Kleijn D, Schekkerman H, Dimmers WJ, Van Kats RJM, Melman D, Teunissen WA. Adverse effects of agricultural intensification and climate change on breeding habitat quality of Black-tailed Godwits Limosa l. limosa in the Netherlands. Ibis. 2010;152:475–86.
Kentie R, Senner NR, Hooijmeijer JCEW, Márquez-Ferrando R, Figuerola J, Masero JA, et al. Estimating the Size of the Dutch Breeding Population of Continental Black-Tailed Godwits from 2007–2015 Using Resighting Data from Spring Staging Sites. arde. Netherlands Ornithologists’ Union; 2016;104:213–25.
Howison RA, Piersma T, Kentie R, Hooijmeijer JCEW, Olff H. Quantifying landscape-level land-use intensity patterns through radar-based remote sensing. Journal of Applied Ecology. 2018;55:1276–87.
Bouten W, Baaij EW, Shamoun-Baranes J, Camphuysen KCJ. A flexible GPS tracking system for studying bird behaviour at multiple scales. J Ornithol. 2013;154:571–80.
Lele SR, Merrill EH, Keim J, Boyce MS. Selection, use, choice and occupancy: clarifying concepts in resource selection studies. Journal of Animal Ecology. 2013;82:1183–91.
Montgomery RA, Roloff GJ, Hoef JMV, Millspaugh JJ. Can We Accurately Characterize Wildlife Resource Use When Telemetry Data Are Imprecise? The Journal of Wildlife Management. 2010;74:1917–25.
Dodge S, Bohrer G, Weinzierl R, Davidson SC, Kays R, Douglas D, et al. The environmental-data automated track annotation (Env-DATA) system: linking animal tracks with environmental data. Movement Ecology. 2013;1:3.
Kays R, Davidson SC, Berger M, Bohrer G, Fiedler W, Flack A, et al. The Movebank system for studying global animal movement and demography. Methods in Ecology and Evolution. 2022;13:419–31.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

The need to account for Argos Doppler errors when estimating habitat use from Argos Doppler locations: evidence and solution

Status:

Version 1

Abstract

Figures

Background

Methods

Study system

Quantifying the imprecision of Argos location estimates

Simulating Argos locations

Comparing habitat assignment approaches

Results

Argos Doppler estimated location imprecision

Simulation study

Case study on habitat use by an individual Black-tailed Godwit followed with a PTT

Discussion

Absolute Performance

Bias from the error ellipse and ways to handle it

The importance of obtaining all necessary Argos data parameters

Conclusions

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1