Produce diagnostics that balance individual error against differencing error

[water-e]

Scatter Plot: Individual vs. Difference-Based Accuracy
X-axis: True values from the reference simulation.
Y-axis: Predicted values from the surrogate.
Different colors/markers: Indicate different scenarios.
Additional line: A diagonal reference line (y = x) to see deviations.
🔹 Variation: Use two scatter plots:
One for individual scenario predictions.
One for differences between scenarios (true vs. predicted differences).

3. Residual Boxplots or Violin PlotsPlot residuals (true - predicted) for both individual predictions and difference-based predictions.
   Helps visualize the bias and spread of errors in both approaches.

Heatmap of Error DifferencesX-axis: Scenario 1 values
Y-axis: Scenario 2 values
Color: Magnitude of error in the difference (true diff - predicted diff).
This reveals whether the surrogate is consistently under/overpredicting scenario deltas.

Bland-Altman Plot for Scenario DifferencesPlots the mean of the true and predicted values (X-axis) against their difference (error) (Y-axis).
Good for spotting trends in surrogate errors.