Number of underway vessels in the Western Mediterranean (2016/01-2020/06)

Terrestrial AIS (T-AIS) data from the Western Mediterranean (map inset Fig 5a) were collated by the Balearic Islands Coastal and Forecasting System (SOCIB) using a real-time operational system connected to a web-service provided by MarineTraffic (https://www.marinetraffic.com/). The database used in this study contained AIS data from 1st January 2016 until 30th June 2020 at 5 minute intervals (> 545 million AIS messages). In addition to the vessel tracks, the database also included information associated with each vessel, such as the vessel type or length. A first pre-processing of the raw data included the removal of duplicates, invalid identification numbers (i.e. Maritime Mobile Service Identity -MMSI- codes without 9 digits) and codes outside the correct numerical range (i.e. MMSI codes with first digits between 2 and 7 are those intended for individual ships). In order to address inconsistencies in the vessel and MMSI combinations (e.g., changes of MMSI across years), we selected the more frequent combination of MMSI and vessel characteristics (e.g. vessel name and vessel type) for each calendar year. We used a similar vessel categorization as the S-AIS dataset, but were able to derive a sixth category from the AIS metadata, separating “recreational” vessels from “other” vessels. Therefore, vessels were classified into six categories: cargo, tanker, passenger (included high speed crafts and passenger vessels), fishing, recreational (included sailing vessels and pleasure crafts), and others (included all other ship types). We excluded ship type codes 20 to 29 (i.e. wing-in-ground-effect and search and rescue aircraft), as well as codes that had an invalid value (i.e. empty or null) or the value was not listed in the previous type codes. We calculated the number of vessels per day taking into account only those that were underway, thus removing moored vessels inside ports that were inactive. T-AIS coverage was not homogenous in the study area (Holmes et al. 2020) due to a non-uniformly distribution of antennas (i.e. few antennas in north Africa, see www.marinetraffic.com). Based on a previous analysis of the AIS coverage (Holmes et al 2020), we filtered vessels within the coastal zone (44.4 km, ~24 nautical miles) of EU countries (i.e. a total area of 164,318.2 km2 comprised by Spain, France and Italy), thus reducing potential bias due to temporal gaps in signal reception.本研究使用的西地中海区域（内嵌地图见图5a）地面自动识别系统（Terrestrial AIS，T-AIS）数据，由巴利阿里群岛海岸与预报系统（SOCIB）通过接入MarineTraffic（https://www.marinetraffic.com/）提供的Web服务的实时业务系统整理得到。本研究使用的数据库涵盖2016年1月1日至2020年6月30日期间、以5分钟为间隔采集的自动识别系统（Automatic Identification System，AIS）数据，总数据量超过5.45亿条AIS报文。除船舶航迹外，该数据库还包含与每艘船舶相关的各类属性信息，例如船舶类型、船舶长度等。原始数据的首轮预处理包括去除重复条目、剔除无效标识编号（例如海事移动服务标识（Maritime Mobile Service Identity，MMSI）代码位数不足9位的情况），以及排除数值范围不合规的代码：首位数字介于2至7之间的MMSI代码为专为个体船舶分配的合规代码。为解决船舶与MMSI代码匹配不一致的问题（例如不同年份MMSI代码发生变更的情况），我们针对每个自然年选取出现频率最高的MMSI与船舶属性（例如船舶名称、船舶类型）的匹配组合。我们采用与卫星自动识别系统（Satellite AIS，S-AIS）数据集相近的船舶分类标准，但本次研究可通过AIS元数据新增第六类船舶，将“休闲船舶”与“其他船舶”区分开来。据此，船舶被划分为六大类别：干货船、油轮、客船（含高速艇与客运船舶）、捕捞船、休闲船舶（含帆船与游艇）以及其他船舶（涵盖其余所有船型）。我们剔除了船型代码20至29对应的船舶（即地效翼船与搜救航空器），同时排除了数值无效（例如为空或空值）或未列入前述船型代码列表的代码。我们仅统计处于航行状态的船舶以计算每日船舶数量，由此剔除了港口内停泊的非活跃船舶。受天线布设分布不均的影响（例如北非地区天线数量极少，详见www.marinetraffic.com），研究区域内的T-AIS覆盖范围并不均匀（Holmes等，2020）。基于此前针对AIS覆盖范围的分析研究（Holmes等，2020），我们将分析范围限定在欧盟成员国的近岸海域（44.4公里，约24海里），该区域涵盖西班牙、法国与意大利，总面积达164318.2平方公里，由此降低了因信号接收存在时间间隙所带来的潜在偏差。