The data described in the Information and Data Sources Section were combined to produce an average spatial distribution as presented in the maps on individual taxon pages. Abundance from each station was measured as catch per 10 m2 surface area. The statistical method, which we adopted, was to stratify the data into equal-sized time intervals (e.g., years), which partially controls the bias due to uneven sampling effort in time. This does not completely correct the bias because some time interval strata have no data at all, and stratification does not help in these cases. Bias for the strata that do have data is corrected by giving each stratum equal weight regardless of the number of stations. For the maps, the data within each 625 km2 grid cell were stratified by year. This removes bias due to uneven sampling between years within each cell, but only for those years for which each cell was sampled; this does not correct for uneven sampling for those years where the cell was not sampled at all.
The maps were produced with a geographic information system (GIS) using ArcInfo software (ArcInfo, 2001), a product of Environmental Systems Research Institute (ESRI). Generic base data such as land extents were derived from a combination of Global Land One-kilometer Base Elevation (GLOBE) (GLOBE Task Team, 1999) and Digital Chart of the World (DCW) data. To produce the land relief, GLOBE data were hill-shaded to show relief, missing data values were filled in, and data were clipped to the extent shown with the DCW data and converted to image format. Bathymetry data are from in-house and the General Bathymetric Chart of the Oceans digital atlas (GEBCO) (IOC, IHO, and BODC, 1997).
Fish density data (catch per 10 m2) were read into a GIS, i.e. ArcInfo, software and converted into ArcInfo coverages (data layers). The fish density data layer shows all the sample locations referenced geographically. However, samples were often taken at the same location resulting in points plotting on top of each other. Because we wanted to show the quantity and distribution of the samples, a different symbology was needed. Fish density is continuous in space even though the density may be zero at some locations. So, point locations were aggregated into a surface by overlaying the points onto a regular grid and assigning a mean value to each grid cell. The cell size of the grid is 25 km x 25 km (625 km2), or roughly 15.5 miles x 15.5 miles. The mean value at each grid cell was derived in two stages. First, the mean catch per 10 m2 was calculated per cell per year, and then the mean catch per 10 m2 was calculated per cell averaged over the years. This equalized the contribution of any single year since some years may have had a greater number of samples. The resulting data layer was a polygon showing all cells where samples were taken and their associated abundance based on catch per 10 m2.
A chloropeth map design was chosen to depict abundance. The purpose of the map is to show the general spatial extent of the data and the general trend of average larval abundance over space. chloropeth maps shade statistical units with intensity proportional to the data values. All cells that were sampled, but contained no individuals of the taxon, are symbolized as gray, indicating absence. The remaining data were classified using quantiles: data were ranked, ordered, and divided into four categories, each containing an equal number of observations. The legend shows the range for each class and the colors are hierarchical in that lighter colors connote lower levels of abundance and darker colors connote higher levels.
The occurrences of adults were derived for the most part from unpublished AFSC data residing in RACEBASE, an Oracle database. RACEBASE was developed by the Resource Assessment and Conservation Engineering Division (RACE) and comprises data from assessment, hydroacoustic, and foreign surveys conducted by federal fishery scientists from 1948 to the present. The geographic extent of RACEBASE data covers the continental shelf and slope of western North America and northeastern Asia from the Arctic Ocean (72° 14' N, 167° 52' W) south through the eastern half of the Chukchi Sea, throughout the Bering Sea (including the continental shelf of northeast Siberia), the Aleutian Basin and eastward along the Aleutian Islands, and along the U.S. Pacific coast from the Gulf of Alaska to the southern border of California (32° 28' N, 119° 18' W).
The data for the egg occurrence map was processed similarly to the adult occurrence map. Data were extracted from ICHBASE and exported in ASCII text format. A data layer was created with the data using ArcInfo, which was then overlaid with a polygon grid having a 25 km x 25 km cell size to produce the final data layer showing presence or absence of eggs of the taxon. The result was a layout design consistent with the larval abundance map, which overcame the issue of stacked data points. Gray squares denote where samples were taken, but no eggs were found. Black squares show the presence of the taxon. Cell sizes appear smaller than those in the average larval abundance map because of the change in map scale.