Pick Fishes by Seafood Watch Data
Data Background
This project used Selenium to scrape data from the Monterey Bay Aquarium website www.seafoodwatch.org. This organization collects a range of data on seafood production techniques and fisheries to generate scores for each seafood type. These scores serve to indicate the environmental impact and sustainability of each seafood type.
The goal is to give recommendations to the consumer for choosing seafood that is fished or farmed in the most environmentally sustainable manner. I obtained their data for 97 types of seafood and further analyzed it to identify various trends as to the best/worst seafood to consume, the various production methods utilized and production locations.
Seafood in Danger
- Overharvesting of seafood
- Stocks in peril / need for sustainable production of seafood
- Environmental impact of seafood harvesting and production
- Illicit fishing flaunts regulations and abuses crews
Seafood Watch Program
The Seafood Watch program from the Monterey Bay Aquarium (www.seafoodwatch.org) aids consumers and business in choosing seafood through a scoring system that assigns both a numeric score and a recommendation for each seafood type. This is based on numerous scientific investigations and analyses that determine the status of each seafood by looking individually at the available sources for each. The numeric scores assigned range from 0 to 10 while the recommendation categories are "Best Choice", "Good Alternative" or "Avoid".
Data Objectives
I sought to make the Seafood Watch scoring system more user-friendly for consumers and businesses by the following:
- Get scores for all sources in a given seafood group type, find average and distributions
- Determine best and worst seafood types according to scores
- Determine relationship between source production method and scores
- Determine relationship between source location and scores
Data Science Methodology
Selenium / Python was used to scrape the Seafood Watch website to obtain the following data:
- Seafood type name (group)
- Number of recommendation categories
- Heading text
- Sources
- Type (species common names)
- Catch/production method
- Location
- Overall numerical score
- Seafood Watch subscores for Fisheries (4 - 5 items) or Aquaculture (10 items)
- Source assessment text
- Eco-certifications
- Type (species common names)
- Catch/production method
- Location
Data Science Results
-
- Figure 1. Number of Source Categories by Seafood Group
-
- Figure 2. Distribution of all Scores by Recommendation Category. Mean of all sources is indicated.
-
- Figure 3. Scores by Seafood Group. The recommendation categories for individual sources are indicated by color. Mean of all scores is shown.
-
- Figure 4. Distribution of Scores for the Top 10 and Bottom 10 Seafood Groups. Mean of all scores is shown.
-
- Figure 5. Scores by Production Method. The recommendation categories for individual sources are indicated by color. Mean of all scores is shown.
Conclusions
- Highest scoring seafoods in general were:
- Scallop, Seaweed, Catfish, Sturgeon, Bass, Abalone, Tilapia, Sole, Turbot, Arctic Char
- These have a higher probability to be more sustainable / environmentally conscious choices given no further details.
- Choose seafood labeled as Eco-Certified by one of the certifying organizations. These would be expected to score highly in the parallel Seafood Watch classification system.
- Choose seafood produced by aquaculture / farming techniques as sources using these methods scored higher on average.
- Always avoid the following as these had no sources that scored above the mean:
- Shark, Toothfish, Triggerfish, Brill, Orange Roughy, Sardine, Pike
Caveats
- Assume the integrity of Seafood Watch scoring system. Assume sources studied are rather complete, but likely there are other sources not represented, including unregulated, illegal, etc.
- Seafood types analyzed represent a group of similar species, it is possible that not all species in a given type are the same with regards to scoring
- There are some seafood types that receive low scores due to incomplete or lack of data
- Bias towards seafood types commonly consumed in the US, as well as US sources
Future Work
- Look further into subscores and EDA for best / worst types.
- Further investigation of location text to determine ranking of best / worst locations
- Further investigations into individual species within groups
The skills I demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
About Author
