How to use ChatGPT to learn Hydrology - Part II: theory, reasoning and assessment preparation

diagnosis
→ focused explanation
→ practice
→ correction
→ transfer
→ oral explanation

I want to study evapotranspiration.

First ask me five diagnostic questions.
Do not explain the topic until I answer.

Then identify my two main conceptual gaps,
correct them and give me two new problems.

Finish with a 90-second oral question.

Make me explain what Hydrology studies and how it differs
from Meteorology, Hydraulics and Hydrogeology.

Do not accept a one-sentence definition.
Ask for:
1. object of study;
2. scales;
3. processes;
4. engineering applications;
5. one Chilean example.

Treat a catchment as a control volume.

Make me identify inputs, outputs and storage changes.
Then change the temporal scale from daily to annual
and ask which terms may become small and which need not.

Inputs - Outputs = Change in storage

P - Q - ET = Delta S

Compare a river catchment with an administrative boundary.

Ask why a catchment is useful for hydrological analysis
and why integrated water resources management must also consider
users, institutions, infrastructure and ecosystems.

Use precipitation, soil moisture and streamflow.

Make me compare their expected spatial and temporal variability
in an Andean catchment and explain why they cannot all be observed
with the same network or resolution.

Relate atmospheric circulation, topography and precipitation
across Chile.

Make me compare a northern catchment, a Mediterranean catchment
and a southern catchment before giving the explanation.

Build a causal chain:
radiation → temperature → vapour pressure → evaporative demand.

Make me distinguish direct relationships from those
that depend on other variables.

Give me an hourly storm.

Make me distinguish:
- intensity;
- accumulated depth;
- duration;
- temporal resolution;
- antecedent dry period.

Then change only the temporal resolution and ask what information is lost.

Compare two events with the same total rainfall:
one over a forested catchment and one over sparse vegetation.

Make me reason when interception can be hydrologically important
and when its relative effect becomes smaller.

Make me distinguish evaporation, transpiration,
potential ET and actual ET.

Then give one water-limited and one energy-limited situation.

Compare Thiessen polygons, inverse-distance weighting
and kriging for catchment rainfall estimation.

Do not ask which is 'best'.
Give me a station network and make me choose according to
network density, topography, objective, variability and data availability.

Make me interpret a Budyko diagram.

Ask:
- what the axes represent;
- what water-limited and energy-limited mean;
- what information is required;
- why a deviation does not automatically imply an error.

Make me explain why water can move through unsaturated soil
without reducing the explanation to 'gravity pulls it down'.

Require discussion of gradients and hydraulic properties.

Compare the same storm over an initially dry and an initially wet soil.

Ask me how infiltration, storage, runoff generation
and streamflow response are expected to change.

Do not give me the equation first.

Help me reconstruct Green-Ampt from its assumptions:
a distinct wetting front, changes in water content
and capillary forces.

Then show the equation and make me perform dimensional analysis.

Compare forest, agricultural land and an urbanised area.

Make me separate the effects of:
- texture;
- structure;
- compaction;
- vegetation cover;
- impervious surfaces.

Do not let me attribute everything to 'soil type'.

Make me distinguish stage measurement, gauging
and streamflow estimation from a rating curve.

Then give me a flood beyond the gauged range
and ask what uncertainty is introduced.

Show me a hydrograph and make me identify:
baseflow, rising limb, peak, lag, recession and direct-runoff volume.

Do not accept purely geometric descriptions:
require interpretation of processes.

Make me interpret Q10, Q50 and Q90.
Include a trap involving exceedance and non-exceedance probability.

Give me a storm and hydrograph.
Make me distinguish total rainfall, abstractions,
effective rainfall and direct runoff.

Give me a catchment with P, E and Q time series.

Make me decide what can be estimated directly,
what requires modelling, and which storage components
remain poorly observed.

Make me explain F(x), 1-F(x), exceedance probability,
quantiles and return period using annual maxima.

Do not use equations until the interpretation is clear.

Give me an IDF table.

Make me:
1. select duration and T;
2. convert intensity to depth;
3. explain why intensity and duration are not interchangeable;
4. construct a conceptual design hyetograph;
5. identify one extrapolation limitation.

Make me explain the linearity and superposition assumptions
of the unit hydrograph.

Then give me a situation where those assumptions are weak.

Do not give me only the calculation.
Make me interpret how antecedent condition, land use
and CN affect effective rainfall and runoff.

Explain what each main component of an event-based HEC-HMS model represents
and what information I should check before accepting a simulation.

Theory = 0.15 × Tests
       + 0.30 × Test 1
       + 0.40 × Test 2
       + 0.15 × Individual Oral Assessment

Generate a five-question micro-test:
- two conceptual;
- one on units;
- one interpretation;
- one error-diagnosis question.

Ask one at a time.

Create a Chilean catchment problem combining
P, ET, Q, storage and physical catchment characteristics.

Assess formulation, units, calculation, interpretation and uncertainty.

Create an integrative problem.

It must force me to combine at least three parts of the course
and justify an engineering decision.

Simulate a 6-minute individual oral assessment.

Ask one question at a time.
Require me to:
- explain concepts in my own words;
- relate processes;
- interpret a result;
- justify a decision;
- identify a limitation;
- answer follow-up questions.

Correct me only at the end.

Make me explain why a 100-year return period
does not mean 'once every 100 years'.

Assess separately:
formulation, equations, units, calculation and interpretation.

Construct a solution containing exactly three plausible errors:
one conceptual, one dimensional and one methodological.
Do not identify them.

Make me choose only after defining:
objective, data, scale, assumptions, uncertainty and limitations.

Generate expressions involving P, Q, area, volume,
time and intensity.
Make me demonstrate which are dimensionally consistent.

Is the sign plausible?
Is the order of magnitude plausible?
Which process controls this result?
What additional observation would increase confidence?

Give me a conceptual multiple-choice question with five options
and exactly one correct answer.

Do not identify the answer.
I will answer first.
Then give me two plausible arguments that another student
might use to defend a different option.
Finally make me answer again.

Now double the antecedent soil moisture.
Do not recalculate immediately.
First ask what I expect to change and why.

Keep annual P fixed but change its seasonal distribution.
Which conclusions from the annual water balance still hold?