Large marine predator aerial survey data for Hauraki Gulf, New Zealand

說明

資料紀錄

此資源出現紀錄的資料已發佈為達爾文核心集檔案（DwC-A），其以一或多組資料表構成分享生物多樣性資料的標準格式。 核心資料表包含 12,263 筆紀錄。

此 IPT 存放資料以提供資料儲存庫服務。資料與資源的詮釋資料可由「下載」單元下載。「版本」表格列出此資源的其它公開版本，以便利追蹤其隨時間的變更。

版本

以下的表格只顯示可公開存取資源的已發布版本。

如何引用

研究者應依照以下指示引用此資源。:

Constantine R, Stephenson F, Hamilton O, Torres L, Kozmian-Ledward L, Pinkerton M (2024). Large marine predator aerial survey data for Hauraki Gulf, New Zealand. Version 1.2. Southwestern Pacific Ocean Biogeographic Information System (OBIS) Node. Occurrence dataset. https://nzobisipt.niwa.co.nz/resource?r=hauraki_predators&v=1.2

權利

研究者應尊重以下權利聲明。:

此資料的發布者及權利單位為 Southwestern Pacific Ocean Biogeographic Information System (OBIS) Node。 This work is licensed under a Creative Commons Attribution (CC-BY 4.0) License.

GBIF 註冊

此資源尚未向GBIF註冊

關鍵字

Occurrence; Observation

聯絡資訊

地理涵蓋範圍

Hauraki Gulf, New Zealand

時間涵蓋範圍

取樣方法

Twenty-two double observer line-transect aerial surveys were conducted typically twice a month from November 2013 to October 2014, following MacKenzie and Clement (2014). Surveys followed a systematic grid of eight parallel transects evenly spaced apart every 10 km, orientated offshore at 62° from the north. Under this design, it was possible to sample across a wide range of environmental conditions. The survey followed a simple, unstratified design due to a wide range of species with different or unknown distributions (Dawson et al. 2008). To ensure equal coverage probability (i.e., a random sample of the total habitat area in the survey area; Buckland et al. 2004) the start point of each survey was randomly chosen using the striplet method (Fewster 2011). Surveys commenced when the Beaufort sea state (BSS) was ≤3 across the study area, there was no rain or fog, and there was sufficient light, i.e., one hour after sunrise and before sunset.

方法步驟描述:

1. All 22 surveys were conducted in a Cessna 207 fixed-wing aircraft, flying at 500 feet (152.4 m) and 100 knots (185.2 km/h). The aircraft accommodated two observers on each side who operated independently with no communication during the survey. Observers logged all sightings of large marine predator species, i.e., cetaceans, pinnipeds, sharks, rays, and oceanic fish observed during the flights. Observers also recorded all sightings of potential prey patches, i.e., schooling fish such as kahawai (Arripis trutta), pilchards (Sardinops sagax), jack mackerel (Trachurus spp.), saury (Scomberesox saurus) and zooplankton aggregations.
2. Due to the double-observer design, records partly consisted of duplicate sightings, i.e., two records of the same animal or group made by observers seated on the same side of the plane. Duplicate sightings were reconciled post-hoc by comparing all records of the same species made by observers seated on the same side of the plane during the same survey. For cetaceans, sightings made within seven seconds and within 10° of each other were considered a duplicate sighting. Sharks and rays were generally encountered at low densities allowing for easy identification of duplicate records. However, we applied a narrower time frame of five seconds to classify records of sharks and prey to compensate for the reduction in the information available to identify duplicates.
3. Sightings of sharks exceeding an estimated total length of 2.5 m (Last and Stevens 2009) that share a similar ecological role were combined to form “pelagic sharks” as they typically inhabit similar habitats (pers. comm. Clinton Duffy, Department of Conservation). These included rarely sighted species, i.e., shortfin mako (Isurus oxyrinchus), blue (Prionace glauca), and adult smooth hammerhead (Sphyrna zygaena) sharks. Juvenile hammerhead shark (total length < 2.5 m; Last and Stevens 2009) distribution was modelled separately as they exhibit different habitat preferences to adults.
4. To estimate the distributions of the large marine predators, binomial BRT models require locations of both presences and absences (Manly et al. 2007). Presence records consisted of sightings recorded during aerial surveys. Given that our sightings only reflect a portion of the animals’ habitat use (i.e., when the animal is at, or close to, the surface of the water), presences records will likely be an underestimate of true distribution. In line with this, absence information collected from our aerial surveys should therefore also not be considered true absences. Instead, pseudo-absences (artificial absence points) were generated (Lobo and Tognelli 2011; Phillips et al. 2009).
5. When data are collected during systematic surveys (as was the case here), pseudo-absences are distributed within the absence zones, i.e., where no sightings were made (Derville et al. 2016). To capture habitat use at a fine temporal (one-day) scale (Derville et al. 2016), 22 unique sets of pseudo-absences were generated for each species, corresponding to each survey. Pseudo-absences were distributed in the on-effort portions of transects, i.e., where observers were in search mode and within the viewing strip as per distance sampling methods (Hamilton et al. 2018; 420 m transect width for sharks and dolphins; 570 m for Bryde’s whale).
6. To ensure that pseudo-absence points reflected available but unused habitat in a non-biased manner, sightings collected over the entire period were pooled to create a kernel density map for each taxon using the Spatial Analyst kernel density tool in ArcMap (2019, v10.2). Density contours were overlaid on survey transects and cropped to the width and length of transects. An exclusion zone was created around presence points two times the length of the transect width (880 m, 1140 m for Bryde’s whales) to prevent environmental overlap between presences and pseudo-absences – the length approximate to the 1 km spatial resolution of the study (Torres et al. 2008). We generated stratified random pseudo-absence points with numbers inversely related to the density contours. Pseudo-absences were distributed with a minimum of two times the transect width to avoid serial autocorrelation (Derville et al. 2016). To weight pseudo-absences according to the stratified design, we created a standardised, inverted kernel density map and extracted the kernel density estimate (KDE) of the cell within which a pseudo-absence point fell to assign a weight. The process produced approximately 30 times more pseudo-absences than presence points for sub-sampling during the modelling process.

引用文獻

1. Constantine, Rochelle et al. (2023). Large marine predator aerial survey data for Hauraki Gulf, New Zealand [Dataset]. Dryad. https://doi.org/10.5061/dryad.wm37pvmrn doi.org/10.5061/dryad.wm37pvmrn
2. Constantine, R., Stephenson, F., Hamilton, O., Torres, L., Kozmian-Ledward, L., & Pinkerton, M. (2023). Large marine predator aerial survey data for Hauraki Gulf, New Zealand [Data set]. Zenodo. https://doi.org/10.5061/dryad.wm37pvmrn doi.org/10.5061/dryad.wm37pvmrn

額外的詮釋資料