Taylor Diagram Module¶

The monet_plots.taylordiagram module provides functionality for creating Taylor diagrams, which are commonly used in meteorology and climate science to evaluate model performance against observations.

Overview¶

Taylor diagrams display the standard deviation and correlation coefficient between model predictions and observations, providing a comprehensive view of model performance in a single plot.

TaylorDiagram Class¶

class TaylorDiagram:
    """Create a Taylor diagram for model evaluation."""


    def __init__(self, obsstd, scale=1.5, fig=None, rect=111, label='OBS', **kwargs):
        """Initialize a Taylor diagram.


        Args:
            obsstd (float): Standard deviation of observations
            scale (float): Scale factor for the diagram
            fig (matplotlib.figure.Figure, optional): Figure to use
            rect (tuple or int): Position and size of subplot
            label (str): Label for observations
            **kwargs: Additional parameters
        """
        pass

Constructor Parameters¶

Parameter	Type	Default	Description
`obsstd`	`float`	Required	Standard deviation of observations
`scale`	`float`	`1.5`	Scale factor for diagram size
`fig`	`matplotlib.figure.Figure`	`None`	Existing figure to use
`rect`	`tuple` or `int`	`111`	Subplot position and size
`label`	`str`	`'OBS'`	Label for observations
`**kwargs`	`dict`	`{}`	Additional parameters

Public Methods¶

`add_sample(stddev, corrcoef, marker='o', label='', **kwargs)`¶

Add a model sample to the Taylor diagram.

td = TaylorDiagram(obsstd=1.0)
td.add_sample(stddev=0.8, corrcoef=0.9, marker='s', label='Model 1')

Parameters:

Parameter	Type	Default	Description
`stddev`	`float`	Required	Standard deviation of the model
`corrcoef`	`float`	Required	Correlation coefficient
`marker`	`str`	`'o'`	Marker style
`label`	`str`	`''`	Model label
`**kwargs`	`dict`	`{}`	Additional plot parameters

Example:

from monet_plots import taylordiagram

# Create Taylor diagram
td = taylordiagram.TaylorDiagram(obsstd=1.2, scale=2.0, label='Observations')

# Add model samples
td.add_sample(stddev=1.1, corrcoef=0.95, marker='o', label='Model A', color='red')
td.add_sample(stddev=1.3, corrcoef=0.88, marker='s', label='Model B', color='blue')
td.add_sample(stddev=0.9, corrcoef=0.92, marker='^', label='Model C', color='green')

# Add contours and legend
td.add_contours(levels=[0.5, 0.8, 0.9, 0.95])
td.finish_plot()

`add_contours(levels=None, **kwargs)`¶

Add contour lines to the Taylor diagram.

td.add_contours(levels=[0.5, 0.8, 0.9, 0.95], colors='gray', linestyles='--')

Parameters:

Parameter	Type	Default	Description
`levels`	`list`	`None`	Contour levels (correlation coefficients)
`**kwargs`	`dict`	`{}`	Additional contour parameters

Example:

# Add custom contours
td.add_contours(
    levels=[0.6, 0.7, 0.8, 0.9, 0.95],
    colors='gray',
    linestyles=['-', '--', ':', '-.', (0, (3, 1, 1, 1))],
    linewidths=1.5,
    alpha=0.7
)

`finish_plot()`¶

Finalize the plot by adding legend and adjusting layout.

td.finish_plot()
td.save('taylor_diagram.png')

Parameters: - None

Example:

td = taylordiagram.TaylorDiagram(obsstd=1.0)
td.add_sample(0.8, 0.9, 'o', 'Model 1')
td.add_contours([0.5, 0.8, 0.9])

# Finalize and save
td.finish_plot()
td.fig.savefig('taylor_diagram.png', dpi=300, bbox_inches='tight')

`set_xlabel(label, **kwargs)`¶

Set the x-axis label.

td.set_xlabel('Standard Deviation', fontsize=12)

`set_ylabel(label, **kwargs)`¶

Set the y-axis label.

td.set_ylabel('Standard Deviation', fontsize=12)

`set_title(title, **kwargs)`¶

Set the plot title.

td.set_title('Model Performance Comparison', fontsize=14, pad=20)

Properties¶

`figure`¶

Access the underlying matplotlib Figure object.

@property
def figure(self):
    """Get the matplotlib Figure object."""
    return self._figure

`axes`¶

Access the matplotlib Axes object.

@property
def axes(self):
    """Get the matplotlib Axes object."""
    return self._axes

Usage Examples¶

Basic Taylor Diagram¶

import numpy as np
from monet_plots import taylordiagram

# Create sample data
obs_std = 1.2
model_data = [
    (1.1, 0.95, 'Model A'),
    (1.3, 0.88, 'Model B'),
    (0.9, 0.92, 'Model C'),
    (1.0, 0.97, 'Model D')
]

# Create Taylor diagram
td = taylordiagram.TaylorDiagram(
    obsstd=obs_std,
    scale=1.8,
    label='Observations'
)

# Add model samples
for stddev, corrcoef, label in model_data:
    td.add_sample(
        stddev=stddev,
        corrcoef=corrcoef,
        marker='o',
        label=label
    )

# Add contours and finalize
td.add_contours(levels=[0.5, 0.8, 0.9, 0.95])
td.finish_plot()

# Save the plot
td.save('basic_taylor_diagram.png')

Advanced Taylor Diagram with Custom Styling¶

import numpy as np
from monet_plots import taylordiagram

# Create Taylor diagram with custom styling
td = taylordiagram.TaylorDiagram(
    obsstd=2.5,
    scale=2.2,
    label='Reference Observations'
)

# Add multiple model categories
# High-performance models
high_perf_models = [
    (2.4, 0.97, 'Advanced Model', 'red', 'o'),
    (2.6, 0.95, 'Optimized Model', 'darkred', 's')
]

# Medium-performance models
med_perf_models = [
    (2.8, 0.88, 'Standard Model', 'blue', '^'),
    (2.3, 0.85, 'Baseline Model', 'lightblue', 'v')
]

# Poor-performance models
poor_perf_models = [
    (3.2, 0.72, 'Experimental Model', 'orange', 'D'),
    (3.5, 0.68, 'Prototype Model', 'darkorange', '*')
]

# Add models with different colors and markers
for models, color in [(high_perf_models, 'red'),
                      (med_perf_models, 'blue'),
for models, color in [(high_perf_models, 'red'),
                      (med_perf_models, 'blue'),
                      (poor_perf_models, 'orange')]:
    for stddev, corrcoef, label, c, marker in models:
        td.add_sample(
            stddev=stddev,
            corrcoef=corrcoef,
            marker=marker,
            label=label,
            color=c,
            markersize=8,
            zorder=10
        )

# Add custom contours
custom_levels = [0.5, 0.7, 0.8, 0.9, 0.95, 0.98]
td.add_contours(
    levels=custom_levels,
    colors='gray',
    linestyles=['-', '--', ':', '-.', (0, (3, 1, 1, 1)), (0, (5, 5))],
    linewidths=1,
    alpha=0.6
)

# Customize plot appearance
td.set_title('Comprehensive Model Performance Analysis', fontsize=16, pad=25)
td.set_xlabel('Standard Deviation', fontsize=12)
td.set_ylabel('Standard Deviation', fontsize=12)

# Add legend and finalize
td.finish_plot()

# Save high-resolution plot
td.fig.savefig('advanced_taylor_diagram.png', dpi=600, bbox_inches='tight')

Integration with MONET Plots¶

from monet_plots import TaylorDiagramPlot
import pandas as pd

# Load model evaluation data
data = pd.DataFrame({
    'model': ['Model A', 'Model B', 'Model C', 'Model D'],
    'stddev': [1.1, 1.3, 0.9, 1.0],
    'corrcoef': [0.95, 0.88, 0.92, 0.97],
    'obs_std': 1.2  # Reference standard deviation
})

# Create Taylor diagram using MONET Plots
plot = TaylorDiagramPlot(
    obsstd=data['obs_std'].iloc[0],
    scale=1.5,
    label='Observations'
)

# Add all models
for _, row in data.iterrows():
    plot.add_sample(
        df=pd.DataFrame({'obs': [1], 'model': [1]}),  # Dummy data for corrcoef
        col1='obs',
        col2='model',
        marker='o',
        label=row['model']
    )

# Update standard deviations manually
plot.dia.samples[-1].set_stddev(row['stddev'])
plot.dia.samples[-1].set_corrcoef(row['corrcoef'])

# Add contours and finish
plot.add_contours([0.5, 0.8, 0.9, 0.95])
plot.finish_plot()

plot.save('monet_taylor_diagram.png')

Advanced Features¶

Custom Contour Styling¶

# Create gradient contour colors
import matplotlib.pyplot as plt
from matplotlib.colors import LinearSegmentedColormap

# Create custom colormap
colors = ['#ffffcc', '#ffeda0', '#fed976', '#feb24c', '#fd8d3c', '#fc4e2a', '#e31a1c', '#bd0026', '#800026']
n_bins = 100
cmap = LinearSegmentedColormap.from_list('custom', colors, N=n_bins)

# Add styled contours
td.add_contours(
    levels=[0.5, 0.7, 0.8, 0.9, 0.95, 0.98],
    cmap=cmap,
    linewidths=2,
    alpha=0.8
)

Interactive Taylor Diagrams¶

# Enable interactive features
import matplotlib.pyplot as plt
plt.ion()  # Turn on interactive mode

td = taylordiagram.TaylorDiagram(obsstd=1.0)
td.add_sample(0.8, 0.9, 'o', 'Interactive Model')

# Add click interaction
def on_click(event):
    if event.inaxes == td.axes:
        # Add model at clicked position
        x, y = event.xdata, event.ydata
        r = np.sqrt(x**2 + y**2)
        if r > 0:
            corrcoef = y / r
            td.add_sample(x, corrcoef, 'x', f'Clicked Model {len(td.dia.samples)}')
            td.fig.canvas.draw()

td.fig.canvas.mpl_connect('button_press_event', on_click)

Data Requirements¶

Input Data Format¶

Taylor diagrams require two key metrics for each model:

Standard Deviation: The RMS difference between model and observations
Correlation Coefficient: The temporal correlation between model and observations

Calculation Example¶

import numpy as np
from scipy import stats

def calculate_taylor_metrics(obs, model):
    """Calculate standard deviation and correlation for Taylor diagram.


    Args:
        obs (array-like): Observation data
        model (array-like): Model data


    Returns:
        tuple: (stddev, corrcoef)
    """
    # Calculate standard deviations
    obs_std = np.std(obs, ddof=1)
    model_std = np.std(model, ddof=1)


    # Calculate correlation coefficient
    corrcoef, _ = stats.pearsonr(obs, model)


    return model_std, corrcoef

# Example usage
obs_data = np.random.normal(0, 1, 1000)
model_data = np.random.normal(0.1, 1.1, 1000)

stddev, corrcoef = calculate_taylor_metrics(obs_data, model_data)
print(f"Standard Deviation: {stddev:.3f}")
print(f"Correlation Coefficient: {corrcoef:.3f}")

Best Practices¶

Model Selection¶

Include reference: Always include observations as the reference point
Compare like-for-like: Ensure models are evaluated on the same dataset
Statistical significance: Only include statistically significant correlations

Visual Design¶

Color coding: Use colors to indicate model categories or performance levels
Marker variety: Use different markers for different model types
Clear labeling: Ensure all models are clearly labeled
Contour levels: Choose appropriate contour levels based on your data range

Interpretation¶

Center point: Models closer to the observation point are better
Distance from center: Radial distance indicates overall skill
Correlation: Higher correlation values are better
Standard deviation: Values close to observation standard deviation are better

Common Use Cases¶

Climate Model Evaluation¶

# Compare multiple climate models
climate_models = {
    'GCM Model A': (1.05, 0.96),
    'GCM Model B': (1.15, 0.92),
    'GCM Model C': (0.95, 0.94),
    'Regional Model': (1.25, 0.88),
    'Statistical Model': (0.85, 0.91)
}

td = taylordiagram.TaylorDiagram(obsstd=1.0, label='Reanalysis')
for name, (std, corr) in climate_models.items():
    td.add_sample(std, corr, 'o', name)

td.add_contours([0.5, 0.8, 0.9, 0.95])
td.finish_plot()

Weather Forecast Verification¶

# Compare different forecast lead times
forecast_models = {
    '24h Forecast': (0.8, 0.98),
    '48h Forecast': (1.0, 0.94),
    '72h Forecast': (1.2, 0.88),
    '96h Forecast': (1.4, 0.82),
    '120h Forecast': (1.6, 0.75)
}

td = taylordiagram.TaylorDiagram(obsstd=1.0, label='Observations')
for name, (std, corr) in forecast_models.items():
    td.add_sample(std, corr, 's', name)

td.add_contours([0.5, 0.8, 0.9, 0.95])
td.finish_plot()

Related Resources:

Plot Types API - Other plot implementations
Examples - Practical usage examples
Base API - Core plotting functionality

Taylor Diagram Module¶

Overview¶

TaylorDiagram Class¶

Constructor Parameters¶

Public Methods¶

add_sample(stddev, corrcoef, marker='o', label='', **kwargs)¶

add_contours(levels=None, **kwargs)¶

finish_plot()¶

set_xlabel(label, **kwargs)¶

set_ylabel(label, **kwargs)¶

set_title(title, **kwargs)¶

Properties¶

figure¶

axes¶

Usage Examples¶

Basic Taylor Diagram¶

Advanced Taylor Diagram with Custom Styling¶

Integration with MONET Plots¶

Advanced Features¶

Custom Contour Styling¶

Interactive Taylor Diagrams¶

Data Requirements¶

Input Data Format¶

Calculation Example¶

Best Practices¶

Model Selection¶

Visual Design¶

Interpretation¶

Common Use Cases¶

Climate Model Evaluation¶

Weather Forecast Verification¶

`add_sample(stddev, corrcoef, marker='o', label='', **kwargs)`¶

`add_contours(levels=None, **kwargs)`¶

`finish_plot()`¶

`set_xlabel(label, **kwargs)`¶

`set_ylabel(label, **kwargs)`¶

`set_title(title, **kwargs)`¶

`figure`¶

`axes`¶