Lab 1: Community Data EDA

Systematic exploration of the National Lake Assessment dataset

Learning Objectives

By the end of this lab, you will be able to:

  1. Download and import complex ecological datasets from public repositories
  2. Systematically assess data quality for community datasets
  3. Create diagnostic visualizations that reveal community structure
  4. Generate hypotheses about community-environment relationships
  5. Document data cleaning decisions with justifications
  6. Communicate ecological patterns to collaborators across disciplines
  7. Synthesize findings across taxonomic groups to generate research questions

Background: The National Lake Assessment

The National Lakes Assessment (NLA) is a collaborative survey between EPA, states, and tribes to assess the condition of lakes across the United States. The survey occurs every 5 years and collects data on:

  • Water chemistry
  • Physical habitat
  • Biological communities (zooplankton, phytoplankton, benthic macroinvertebrates)
  • Algal toxins
  • Sediment characteristics

We’re using the 2022 NLA data for this course. You can explore summary results using the NLA 2022 dashboard.

Lab Structure: Collaborative Research Approach

This course is different from traditional labs. We’re not just learning JSDMs - we’re actively exploring how they can reveal patterns of community assembly across trophic levels in lake ecosystems. You are collaborators in this research, not just students following a recipe.

How this lab works

  1. You will be assigned to one taxonomic group: Benthic macroinvertebrates, zooplankton, or phytoplankton
  2. You become the expert on that group’s patterns in NLA lakes through systematic EDA
  3. You conduct independent analysis following the framework below, documenting findings for the team
  4. We reconvene as a research group to compare findings across groups and generate hypotheses

Why this approach?

By dividing the analytical work across taxonomic groups, we can:

  • Cover more ecological complexity than any individual could alone
  • Compare community patterns across trophic levels
  • Generate hypotheses about cross-trophic relationships and community assembly
  • Build toward multi-group JSDM analyses

This mimics real collaborative ecology: different researchers specialize in different taxa, then synthesis happens through collaboration.

Your job

Document your work clearly. The rest of the research team needs to understand your findings to build on them. This means:

  • Clear code with comments explaining decisions
  • Interpretable visualizations with captions
  • Written summaries of patterns you observe
  • Explicit statements of data quality issues and how you handled them

Think of your lab submission as a technical report to collaborators, not just an assignment.

What Makes This Course Experimental

As everything in this class, this is a work in progress. We (the instructors) are genuinely exploring these data alongside you. If you find resources, identify additional analytical steps we’re missing, or have methodological suggestions, we welcome contributions through GitHub Issues or direct discussion.

This means some uncertainty is built in. We’ll make decisions together, document our reasoning, and learn from what emerges. That’s how real ecology works.

Part 0: Data Acquisition

Download the Data

Data portal: https://www.epa.gov/national-aquatic-resource-surveys/data-national-aquatic-resource-surveys

For JSDM we need at least three things: site info, species data (counts or presence/absence), and environemntal covariates. Navigate to National Lakes Assessment 2022:

EVERYONE downloads:. 1. Site information. 2. Environmental data:
- Physical Habitat - Secchi - Water Chemistry - Landscape Data

YOUR ASSIGNED GROUP (download ONE of these): - Group 1: Benthic Macroinvertebrates - Group 2: Zooplankton
- Group 3: Phytoplankton

Save these files to your data/raw/ folder. Keep the original filenames.

Document Your Download

Add an entry to your data/README.md following the template from Lab 0.

Include:

  • What you downloaded
  • When you downloaded it
  • The URL
  • Citation

Your Taxonomic Group Assignment

[We’ll assign these in class]

What to focus on for your group:

If you have Benthic Macroinvertebrates: info - These are bottom-dwelling invertebrates (insect larvae, worms, mollusks) - Key ecological role: Indicators of water quality and habitat condition - Questions to explore: How do benthos respond to sediment, nutrients, oxygen?

If you have Zooplankton: info - These are tiny drifting animals (copepods, cladocerans, rotifers) - Key ecological role: Link between phytoplankton and fish - Questions to explore: How do zooplankton respond to nutrients, predators, lake size?

If you have Phytoplankton: - These are microscopic algae - Key ecological role: Primary producers, base of food web - Questions to explore: How does phytoplankton composition vary with nutrients, other environmental variables?

Systematic EDA Framework

Before we fit Joint Species Distribution Models, we need to understand what we’re modeling. This lab walks through a systematic EDA framework, not because you don’t know how to explore data, but because having a checklist prevents you from missing critical issues. As everything in this class, this is a work in progress.

Think of this as the workflow you’d use if a collaborator handed you a new dataset and said “can you analyze this?” You need to quickly but thoroughly understand what you’re working with.

Your task: Work through the phases below with the NLA 2022 data. The specific code is up to you. I’m providing the questions you should answer and the decisions you should make. Document your findings and decisions as you go.

Phase 1: Understand Data Structure

Goal: Know what you have before you manipulate it or model it

Questions to Answer

  1. What is the observational unit?
  2. What is the response unit?
  3. What format is the data in? (long vs wide, presence/absence vs abundance)
  4. What metadata exists? (coordinates, dates, environmental variables)
  5. *How many sites do you have data for?** (Does it match the site info file?)

Load and Inspect

Code 
library(tidyverse)
library(here)

# Load data - ADJUST FILENAMES TO MATCH WHAT YOU DOWNLOADED
sites <- read_csv(here("data", "raw", "nla2022_wide_siteinfo.csv"))
zoop <- read_csv(here("data", "raw", "nla2022_wide_zooplankton_count.csv"))

Your Analysis

Code 
# Dimensions
dim(sites)
dim(zoop)

# Structure
str(sites)
str(zoop)

# What are we working with?
# How many sites?
# How many species?
# What format?

Deliverable: Write 3-4 sentences describing the dataset structure in plain English.

Phase 2: Assess Data Quality

Goal: Find problems before they become analytical nightmares

Questions to Answer

  1. Are there missing values? Where and how many?
  2. Are there suspicious values? (negatives, impossibly high counts)
  3. Is sampling effort consistent across sites?
  4. Are there data entry errors? (typos in species names, impossible dates)

Your Analysis

Code 
# Missing data patterns
summary(sites)
summary(zoop)

# Check for specific issues:
# - Are there NAs?
# - Range of abundance values
# - Species name consistency
# - Duplicate records?

Deliverable: A list of data quality issues found and decisions made about how to handle them. Document these as comments in your code or in a separate markdown section.

Phase 3: Characterize the Response (Community Structure)

Goal: Understand the biological patterns

Questions to Answer

  1. What’s the distribution of species richness across sites?
  2. Which species are common? Which are rare?
  3. How abundant are communities overall?
  4. Are there dominant species or is diversity evenly distributed?

Your Analysis

You’ll likely need to reshape/transform the data depending on whether it’s in long or wide format.

Code 
# Species richness per site
# Hint: you may need to pivot_longer if data is wide

# Species prevalence (how many sites has each species?)

# Total abundance patterns

# Rarity: how many species found at only 1 site? 2 sites?

Key Visualizations to Create

  1. Histogram of species richness across sites
  2. Species abundance distribution (rank-abundance curve or similar)
  3. Occurrence frequency histogram

Deliverable: 2-3 plots with captions describing what they reveal about community structure.

Phase 4: Explore Environmental Covariates

Goal: Understand environmental context

Questions to Answer

  1. What is the range and distribution of each environmental variable?
  2. Are environmental variables correlated with each other?
  3. Are there environmental gradients we should expect to drive patterns?

Your Analysis

Code 
# Identify key environmental variables in sites data

# Univariate summaries

# Pairwise correlations

# Look for gradients

Key Visualizations to Create

  1. Correlation matrix (visual)
  2. Histograms of key environmental variables
  3. Geographic map of sites (if lat/lon available)

Deliverable: Identify 2-3 environmental variables that might be important and explain why.

Save Your Cleaned Data

After going through all these quality checks and decisions, save a cleaned version:

Code 
# Save your processed data
write_csv(cleaned_data, here("data", "processed", "nla_2022_cleaned.csv"))

Reflection Questions

  1. What surprised you? Were there data quality issues you didn’t expect?

  2. How does this inform JSDMs? Based on your EDA, what patterns would you want a Joint Species Distribution Model to capture?

  3. What decisions did you make? List 2-3 major data cleaning decisions and justify them.

Submission - hold off on this, until we figure it out with Mathew.

Submit to Canvas:

  1. Rendered HTML or PDF of your analysis (knit this document)
  2. Your cleaned data file (data/processed/nla_2022_cleaned.csv)
  3. Your updated data/README.md with NLA data documented
  4. (Optional) GitHub repo link if you’re using version control

Self-Assessment

Review the learning objectives. Can you:

  • Systematically approach a new community dataset?
  • Identify and justify data quality decisions?
  • Create visualizations that reveal community patterns?
  • Generate testable hypotheses about community-environment relationships?