ENEN90028 Chap.2 Causal Questions and Conceptual Models
Causal Questions and Conceptual Models
Intervention monitoring asks what would have happened without the intervention. This chapter uses the counterfactual definition of cause to separate perturbable actions from conditioning attributes such as geology or location. It then turns the intervention, mechanisms and environmental response into a conceptual model whose arrows guide measurement and expose alternative explanations before field work begins.
What this chapter covers
- 01
Counterfactual cause and alternate scenarios
- 02
Conditioning attributes versus interventions
- 03
Intervention mechanism response pathways
- 04
Evidence for individual causal links
- 05
Alternative explanations and confounders
- 06
Directional and temporal predictions
Worked example · free
Build a vehicle access pathway
- 1The intervention is greater vehicle access near the dam.
- 1A plausible mechanism is soil disturbance followed by fine sediment delivery.
- 1The response is a change in macroinvertebrate assemblage condition.
- 1Surrounding land use or hydroperiod is an alternative explanation to match or measure.
Key terms
- Counterfactual cause
- A cause defined through the alternate response expected if the cause were perturbed.
- Conditioning attribute
- A property that changes consequences without itself being the intervention.
- Causal pathway
- A sequence of linked processes connecting intervention and response.
- Mechanism
- The process through which an intervention is expected to alter a response.
- Confounder
- An alternative factor associated with both comparison condition and response.
Causal Questions and Conceptual Models FAQ
What makes a causal question testable?
The action must be something that could conceivably be changed, the environmental response must be observable, and the design must provide a defensible alternate scenario. A statement about association alone does not identify what would happen if the proposed cause were absent or altered.
Does every arrow need a measurement?
Not necessarily. A short project may monitor selected links, but it should state which mechanisms are measured and which remain supported only by literature. Measuring an intermediate process can strengthen interpretation, while skipping every mechanism leaves the conclusion dependent on a broad association.
Why draw a conceptual model before choosing equipment?
The model defines the processes and endpoints that matter, so it guides what should be measured. Choosing devices first encourages a convenience driven project in which available readings are collected without a clear relationship to the intervention or the valued environmental response.
Assessment move
Take an unfamiliar environmental scenario and write intervention, mechanism, response and alternative cause in four columns. Convert each arrow into an if-then sentence and predict its direction and delay. Next identify which link the proposed measurement tests. In quiz practice, reject causal statements that name only a pattern, season, location or fixed attribute without a perturbable intervention.
Redraw one conceptual model from memory each study session, but change the environmental setting so the reasoning transfers. Begin with a perturbable action such as stock exclusion, flow release or fertiliser application. Add no more than three intermediate processes before the valued response. For every arrow, write a supporting observation and an observation that would challenge the link.
This turns the diagram into a test plan rather than a picture. Mark conditioning attributes in a different colour. Soil type, landscape position and prior land use may change the size of a response, but they are not automatically interventions. Decide whether each should be matched, blocked, measured as a covariate or acknowledged as a limitation.
Then practise a counterfactual sentence: if the action had not occurred, what response would the design predict at the same time and place? If no defensible answer is available, the causal question needs a stronger comparison. When reading literature, assign evidence to individual arrows rather than treating one paper as proof of the whole model.
Note the system, scale and intervention studied, then decide which link it supports in the project. A mechanism established elsewhere may justify a prediction without proving that it operated at the field site. Prepare for presentation questions by explaining the model without jargon in thirty seconds. Name the action, the mechanism and the environmental consequence, then state which links the team measures directly.
A strong answer distinguishes field evidence from prior knowledge. For quiz revision, look for options that confuse an association with a cause or treat a fixed background attribute as an intervention. Finish every explanation with the alternative cause that the design addresses or leaves open. This habit makes limitations precise and prevents a broad causal claim from outrunning the comparison.
After drawing the model, assign each proposed observation to an arrow or node. If a measurement cannot be placed, decide whether it is contextual information or unnecessary effort. If an important arrow has no observation, label it as literature-supported, assumed or untested rather than implying that the project measures the whole pathway.
Then challenge the model with timing: identify the earliest response, the delayed endpoint and the observation window required for each. Reversing the predicted direction or shifting the lag should change what the data would look like. Close with a rival-model drill.
Draw one alternative route from the intervention or a background driver to the same response, and name the control, baseline, covariate or limitation needed to address it. Comparing two explicit causal stories is more revealing than listing generic confounders, and it prepares a clear explanation of why the chosen design supports some links more strongly than others.
