LAST UPDATE:

19 June 2025

Please update if your installation of AISanalyze is prior to this revision.

Last updates:
correction in AIScorrect_speed for stationary boats; correction of time_travelled and distance_travelled in the case of several consecutive interpolated points.

Functionality of the package

AISanalyze is an R-package developed to specifically correct GPS errors in AIS data, interpolate (linearly) the vessel positions at the desired times, and extract the locations and information of vessels around desired locations, at desired times. The advantage of this package is that the computation time is considerably reduced to perform these operations, enabling results to be obtained in few minutes to few hours depending on the size of AIS data and the number of points to extract. Furthermore, this package identifies the vessels from the base stations and aircrafts in the AIS data, based on the distance and speed travelled, allowing to filter only the first if desired. It extracts information on vessel length and type as well, by excluding errors on vessel length and type originally present in the AIS data. First, an all-in-one function was constructed to analyze, correct, and extract AIS data around desired locations, therefore:

• calculating the distance, time and speed travelled by each vessel;
  
• identifying base stations and aircrafts in the AIS;
  
• correcting GPS errors and GPS delays;
  
• interpolating the vessel positions at the customized times;
  
• extracting the vessel positions and their information around the desired locations, at the desired times.

However, each function is available individually as well. Second, a further function allows to interpolate (linearly) all AIS data at the desired temporal resolution, regardless of the location. Finally, a last function extracts the length and type of vessel per MMSI (Maritime Mobile Service Identity), removing the errors present in the AIS data. This information can later be added to the interpolated/extracted AIS data.

AISanalyze R-package

1. Quick overview

You can:

• calculate the distance and time travelled for each vessel (defined by MMSI) between every AIS signal, as well as the corresponding speed (km/h), using AIStravel function;

• interpolate the MMSI locations at the desired times with the function AISinterpolate_at. To restrict the computation time and not interpolate tracks of vessels that are not of interest, the function asks to associate every time with a location (longitude, latitude): the AIS data are then kept only within a defined radius (in meters, that the user can change in the function parameters) around these locations.

  • This function allows, for example, interpolating the vessel locations around your survey / bio-logging data at the time you wish.
    
  • You can enter, as input data, a data frame with several rows, each one containing different combinations of time~location (in order to extract vessel locations along the track of your survey / bio-logging data, for example).
    
  • Vessel tracks are interpolated based on AIS data that are within the radius around all your input locations; this radius (default value: 200 km) must therefore be large enough to account for GPS or AIS gaps, for example in offshore areas.
    
  • The GPS errors and unrealistic speeds are also corrected in the process of interpolation.
    
  • You can filter out the base-stations and aircrafts from the AIS data using filter_station = TRUE and filter_high_speed = TRUE. The first are estimated based on the distance they have travelled in the AIS data (the threshold of distance can be customized for advanced use), the latter are estimated based on their speed in the AIS (threshold can be customized as well).

• Extract the positions of vessels, interpolated or not, at desired times and around desired locations, with the function AISextract. A parameter search_into_radius_m allows extracting vessel positions that are within a certain distance (in meters) from your input location(s).

  • Your input data is a data frame where each row contains a combination of time~location (location around which you want to extract vessel locations). You can, for example, define input data as your survey / bio-logging data.
    
  • The function parameter t_gap defines the time range over which AIS data are considered for extracting the vessel locations at the closest moment to your input times. Default is t_gap = 10×60 = 10 minutes, meaning AIS data at the input times ± 10 minutes are kept for extracting vessel locations at the closest time of your input time.
    
  • If you are also interested in the presence of vessels before your input time (e.g. how many vessels where present every minute in the study area, until 10 minutes before the survey took place): you can set duplicate_time = TRUE, max_time_diff = 10×60 (number of seconds) and t_gap = 1×60 (number of seconds). This extracts the vessel locations every minutes from your input time until 10 minutes before, for later advanced analysis.
    
  • If you want to extract the vessel locations within an area that is not circular, it is recommended to enter, as input data, different locations that allow to cover your study area when they are overlapped by the radius of search_into_radius_m. Duplicated vessel locations must then be removed from the output data, and vessels signals that are within your study area must be filtered using sf intersection, for example. This is not too complicated, but do not hesitate to contact me for more information, and we could consider adding some options in the function to extract directly vessel locations within the non-circular areas, without these additional steps.
    
  • If many input times are defined for extraction, the process will also be long. The computation time can be greatly decreased with accelerate = TRUE and average_at (number of seconds). This average the input times that you want to extract to time intervals that are separated by ‘average_at’ number of seconds.
    
  • The output data are the input data, with the AIS data of surrounding vessels joined using left_join (dplyr R-package). This means that an input data with 2 surrounding vessels will be present 2 times in the output data, each row containing the AIS data of 1 of the vessel, at the closest moment of the time you want to extract. An input data without any surrounding vessel will be present 1 time in the ouput data, but the columns corresponding to the AIS data will all be filled by NA.
    
  • In addition to the columns of the input data and AIS data, the output data contains other columns that add information on the AIS data returned (for example, how many points this MMSI has in the AIS data; whether any speed_kmh value was NA (very unexpected !); or the difference of seconds between the input time and the extracted time). More information on these columns can be found typing ??AIStravel_interpolate_extract or ??AISextract on R console.

Parallelize:

AISinterpolate_at can be run in parallel (with parallelize = TRUE; you can define the number of cores to use with nb_cores and the file where iterations, warnings and errors are printed with outfile (doParallel does not print results in the console of R by default). outfile can also contain the path of the folder where you want to save the file, in addition to the name of the file (usually in .txt extension, must be specified in outfile).

Careful:

The RAM and CPU needed to run large datasets increases exponentially and parallel should be run only after having investigated the RAM and CPU used in series computation.

Important:

Most of the time, you are interesting running the above 3 steps (calculate speed, interpolate vessel tracks and extract vessel positions in your study area at the desired times). To simplify the process, the function AIStravel_interpolate_extract does the full process for you in a single run: you can enter your AIS data, your input data and define wether you want to extract the presence of vessels prior to your input data as well (duplicate_time = TRUE, max_time_diff = number of seconds until which you are want to extract vessel locations, t_gap = intervals of seconds over which vessel locations are extracted from your input times until ‘max_time_diff’ seconds before). You can also define your radius where vessels signals are considered for interpolation (see AISinterpolate_at above), and search_into_radius_m (distance over which you want to extract vessel locations around your input locations). Due to the high computation time for large datasets (either input or AIS), the function also allows to progressively write the results during the process, avoiding to spend extra time if the function crashes in reason of large computation requirements. With the appropriate set of parameters, the function can be simply re-run to load the saved results and to continue at the step it stopped. Please do not used the files from save_AISinterlate_at outside of the function as they are not complete: they need to be loaded by the function AISinterpolate_at (or AIStravel_interpolate_extract) to be completed. To do this, you have the function parameters:

• Overwrite: if TRUE, the function runs all the steps and overwrite the existing files.

• load_existing_files: if TRUE, the function loads the existing files, when they exist (otherwise the function runs the usual process).

• save_AIStravel, save_AISinterlate_at, save_AISextract_perHour: if TRUE, save the results of the functions during the process. For the functions AISinterlate_at and AISextract, the results are saved for every day and hour processed. The results are saved as .rds files. The file names can be chosen with the parameters file_AIStravel, file_AISinterlate_at, file_AISextract_perHour (do not specify file extension as .rds will be automatically added, but these names can contain the folder path where you want to save the results). If load_existing_files = TRUE, the function try to load the files named as the 3 above parameters (if they exist, otherwise it runs the full function as usual). Please do not used the files from save_AISinterlate_at outside of the function as they are not complete: they need to be re-loaded by the function.

• run_AIStravel: if TRUE, run the function AIStravel. If the function has already been run and results saved but that the function stopped afterwards, you can define it as FALSE and load_existing_files = TRUE to reload the results and continue the calculation of interpolation/extraction.

• run_AISinterpolate_at: same but for the interpolation of vessel tracks. If the function has stopped during AISinterpolate_at, and not during AIStravel, you can simply set load_existing_files = TRUE, run_AIStravel = FALSE and run_AISinterpolate_at = TRUE to run only the latter and skip the AIStravel.

• run_AISextract_perHour: same but for the extraction of AIS data around your desired location~time. If the function has stopped during AISextract, and not before, you can simply set load_existing_files = TRUE, run_AIStravel = FALSE, run_AISinterpolate_at = FALSE and run_AISextract_perHour = TRUE to run only the latter and skip the first steps.

2. For further information:

• MMSI_infos: taking AIS data as input, this function aims to estimate the most probable type, length, draught, width, imo and name of each vessel based on the AIS information. These can indeed contain diverging information for single vessels. The function returns a list of 2 elements:

  • The first are the most probable type, length, draught, width, imo and name for each vessel (MMSI).

  • The second are the detailed information that led to these estimates (type ??MMSI_infos on R for detailed information on the columns).

The information on vessel length, type, and other, can then be joined by MMSI to the results obtained with the above functions.

• AIScorrect_speed: is the function that corrects GPS gaps and unrealistic speeds in the AIS data. It can be used outside of AISinterpolate_at.

  • If you want to keep the high-speed crafts in output data and avoid correcting their speed, you can set correct_high_speed_craft to FALSE.

  • A fixed upper threshold (in km/h) can be defined in parameters to define “unrealistic speed”.

  • Another vessel-specific threshold can be defined as a function of the vessel speed. This function can use the argument speed_kmh (the vector of speed for each vessel) to use the mean, for example, of the vessel speed – you have to use mean(speed_kmh) in this case – and define a upper threshold adapted to the vessel. In the same manner, the median – median(speed_kmh) – or quantile – quantile(speed_kmh, 0.95) for example – can be used to define the upper threshold, as well as any value derived from the vessel speed. You can take a look to the default expression used for this threshold in AIScorrect_speed to see an example.

How to Cite

Pigeault R., Ruser A., Ramírez-Martínez N.C., Geelhoed S.C.V., Haelters J., Nachtsheim D.A., Schaffeld T., Sveegaard S., Siebert U., Gilles A. (2024). Maritime traffic alters distribution of the harbour porpoise in the North Sea. Marine Pollution Bulletin. 208: 116925.