Visualization¶

GeoEquity provides tools for creating publication-ready spatial accuracy visualizations.

Visualization Modes¶

Mode	Description	Use Case
`observation`	Per-station R² scatter	Show raw accuracy at observation locations
`interpolation`	IDW interpolation	Smooth accuracy surface
`average_NxM`	Grid-averaged R²	Aggregate accuracy in spatial bins
`spatial_model`	SVM regression	ML-based spatial pattern
`two_stage`	TwoStageModel prediction	Density-aware accuracy prediction

TwoStageModel Prediction Map¶

from geoequity.two_stage.visualization import plot_predicted_accuracy_map

# Get station network
stations = df.drop_duplicates(subset=['longitude', 'latitude'])
sufficiency = df['sufficiency'].iloc[0]

# Plot density + predicted accuracy
fig, axes = plot_predicted_accuracy_map(
    ts_model,
    stations['longitude'].values,
    stations['latitude'].values,
    sufficiency,
    lon_range=(-10, 35),
    lat_range=(35, 70),
    grid_size=30
)
plt.suptitle('TwoStageModel: Density → Accuracy', y=1.02)
plt.show()

Compare Multiple Modes¶

from geoequity.visualization import plot_accuracy_comparison

fig, axes = plot_accuracy_comparison(
    df.loc[test_idx],
    model_name='linear',
    modes=['observation', 'interpolation', 'average_15x20', 'spatial_model', 'two_stage'],
    accuracy_range=(0, 1),
    lon_range=(-10, 35),
    lat_range=(35, 70),
    # Required for 'two_stage' mode:
    ts_model=ts_model,
    station_lons=stations['longitude'].values,
    station_lats=stations['latitude'].values
)
plt.suptitle('Model Accuracy - Visualization Modes', y=1.02)
plt.show()

Single Mode Plot¶

from geoequity.visualization import plot_accuracy_map

fig, ax = plot_accuracy_map(
    df.loc[test_idx],
    model_name='linear',
    mode='interpolation',
    accuracy_range=(0, 1),
    lon_range=(-10, 35),
    lat_range=(35, 70)
)
plt.show()

Diagnostic Plots¶

# Generate TwoStageModel diagnostics
ts_model.diagnose(save_dir='diagnostics/', show=True)

Outputs: - stage1_gam.png: GAM partial dependence plots - stage2_svm.png: Spatial residual map - diagnosis.txt: Summary statistics

Note on Negative R² Values¶

R² can be negative when the model performs worse than predicting the mean. This happens when: - Sparse regions have insufficient data - Model has poor generalization in certain areas

If negative R² values appear frequently, consider:

Alternative Metric	Description
MSE/RMSE	Always positive, easier to interpret
MAE	More robust to outliers
Correlation (r)	Ranges -1 to 1, captures linear relationship

Styling Tips¶

Use Spectral_r or RdYlGn for accuracy (red=low, green=high)
Set consistent accuracy_range=(0, 1) for R² values
Add land boundaries with geopandas for context
Use grid_size=30 for smooth maps, lower for faster rendering