Monitor Only vs. Monitor and Model
What additional assumptions and claims are introduced when environmental data collection moves from monitoring toward modeling?
This project used EPA air-quality data to examine the distinction between observing environmental conditions and constructing a model intended to explain or predict them. It emphasizes that visual patterns and measurements do not independently establish causal relationships.
Completed student research project developed during the ASRI Python track. Its scope was educational and exploratory.
Position
Central claim
Monitoring describes observed conditions, while modeling introduces structural assumptions that support different—and potentially stronger—claims.
Approach
Method and evidence
How the argument is currently supported
Current approach
Collection and organization of EPA air-quality data.
Cleaning and transformation with Python and Pandas.
Exploratory visualization.
Conceptual comparison of monitoring and modeling claims.
Supporting observations
Measurements can show temporal or geographic differences.
Visualization can reveal correlation and pattern.
Prediction requires assumptions about relationships and future stability.
Causal explanation requires more than observed association.
Argument
Current structure
The developing argument
Monitoring
Monitoring records conditions through measurements. It can establish what was observed at a location and time within the limits of the instrument and sampling process.
Modeling
A model introduces relationships among variables. These relationships may support estimation, prediction, or explanation, but they also introduce assumptions.
The epistemic boundary
A visualization may reveal that variables move together. It does not by itself establish why they move together.
The distinction protects the analysis from making claims that exceed the evidence.
Boundaries
Epistemic boundaries
What this work does not yet establish
Current limitations
The project did not construct a complete causal environmental model.
Available data may contain missingness and uneven sampling.
Weather, traffic, industry, and other contextual variables were not fully incorporated.
The project duration constrained validation.
What remains unresolved
Which variables are necessary for useful air-quality modeling?
How should missing spatial and temporal data be represented?
What model would be appropriate for causal rather than predictive claims?
Relationships
Connected work