diff --git a/lars/preprocessing/__init__.py b/lars/preprocessing/__init__.py
index f5a0d98..3720720 100644
--- a/lars/preprocessing/__init__.py
+++ b/lars/preprocessing/__init__.py
@@ -1,2 +1,2 @@
 from .radar_preprocessing import preprocess_radar_data # noqa: F401
-from .labels import load_labels, save_labels, change_file_path, copy_labels, apply_criteria_to_labels, combine_labels # noqa: F401
\ No newline at end of file
+from .labels import load_labels, save_labels, change_file_path, copy_labels, apply_criteria_to_labels, combine_labels, standardize_labels # noqa: F401
\ No newline at end of file
diff --git a/lars/preprocessing/labels.py b/lars/preprocessing/labels.py
index 6955944..1b62f64 100644
--- a/lars/preprocessing/labels.py
+++ b/lars/preprocessing/labels.py
@@ -1,6 +1,44 @@
 import pandas as pd
 import os
 
+# Maps non-canonical label spellings (matched case- and whitespace-
+# insensitively) to their canonical codebook form.
+STANDARD_LABEL_MAP = {
+    "ambiguous": "Ambiguous / Uncertain",
+    "unknown": "Ambiguous / Uncertain",
+    "stratiform": "Stratiform Precipitation",
+    "ambiguous / uncertain": "Ambiguous / Uncertain",
+}
+
+
+def standardize_labels(df, label_column='label'):
+    """
+    Standardize inconsistent label spellings to their canonical codebook form.
+
+    Maps the ambiguous/unknown labels ('Ambiguous', 'UNKNOWN', and any casing
+    thereof) to the single canonical value 'Ambiguous / Uncertain', and maps
+    the bare 'Stratiform' label to the full codebook name 'Stratiform
+    Precipitation'. Matching is case- and whitespace-insensitive. Values that
+    don't match one of these variants (including labels already in their
+    canonical form, e.g. 'Stratiform Precipitation') are left unchanged.
+
+    Parameters
+    ----------
+    df (pd.DataFrame): DataFrame containing a label column to standardize.
+    label_column (str): Name of the column containing labels. Default 'label'.
+
+    Returns
+    -------
+    pd.DataFrame
+        A copy of ``df`` with standardized values in ``label_column``.
+    """
+    df = df.copy()
+    keys = df[label_column].astype(str).str.strip().str.lower()
+    mapped = keys.map(STANDARD_LABEL_MAP)
+    df[label_column] = mapped.where(mapped.notna(), df[label_column])
+    return df
+
+
 def change_file_path(radar_df, new_path):
     """
     Change the file paths in the radar DataFrame to a new path.
diff --git a/lars/util/__init__.py b/lars/util/__init__.py
index d7a1c04..70c5cda 100644
--- a/lars/util/__init__.py
+++ b/lars/util/__init__.py
@@ -1,3 +1,7 @@
 from .confusion_matrix import plot_confusion_matrix, calculate_cohen_kappa # noqa: F401
 from .image_grid import plot_label_images # noqa: F401
 from .label_timeseries import plot_label_timeseries # noqa: F401
+from .kappa_matrix import plot_kappa_matrix, calculate_kappa_matrix # noqa: F401
+from .label_rate_matrix import plot_label_rate_diff_matrix, calculate_label_rate_diff_matrix # noqa: F401
+from .label_disagreement_matrix import plot_label_disagreement_matrix, calculate_label_disagreement_matrix # noqa: F401
+from .dawid_skene import fit_dawid_skene, score_against_consensus, plot_dawid_skene_confusion # noqa: F401
diff --git a/lars/util/dawid_skene.py b/lars/util/dawid_skene.py
new file mode 100644
index 0000000..5f10398
--- /dev/null
+++ b/lars/util/dawid_skene.py
@@ -0,0 +1,257 @@
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from sklearn.metrics import accuracy_score, f1_score, cohen_kappa_score
+
+
+def _log_sum_exp(log_values, axis):
+    max_val = np.max(log_values, axis=axis, keepdims=True)
+    max_val = np.where(np.isfinite(max_val), max_val, 0.0)
+    summed = np.sum(np.exp(log_values - max_val), axis=axis, keepdims=True)
+    return (max_val + np.log(summed)).squeeze(axis)
+
+
+def fit_dawid_skene(df, columns=None, max_iter=100, tol=1e-4, smoothing=0.1):
+    """
+    Fit a Dawid-Skene model to estimate a consensus label per item and a
+    confusion matrix per rater, without requiring any rater's labels to be
+    treated as ground truth.
+
+    Each rater's label is modeled as a noisy observation of an unknown true
+    class. The EM algorithm jointly estimates (a) each rater's confusion
+    matrix ``P(observed | true)`` and (b) a posterior distribution over the
+    true class for every item. The entropy of that posterior is a per-item
+    uncertainty estimate: near zero when raters agree, higher when they
+    don't.
+
+    Parameters
+    ----------
+    df (pd.DataFrame): DataFrame containing one column per rater/source.
+        Missing values (NaN) are treated as "this rater did not label this
+        item".
+    columns (list of str or None): Columns (raters) to fit on. Defaults to
+        all columns in ``df``.
+    max_iter (int): Maximum number of EM iterations.
+    tol (float): Convergence tolerance on the maximum absolute change in the
+        item posterior between iterations.
+    smoothing (float): Additive (Laplace) smoothing applied to each rater's
+        confusion matrix during the M-step, to avoid zero probabilities.
+
+    Returns
+    -------
+    dict with keys:
+        consensus (pd.Series): MAP consensus label per item, indexed like
+            ``df``. NaN for items with no non-missing rater.
+        consensus_proba (pd.DataFrame): Item x class posterior probabilities.
+        item_entropy (pd.Series): Shannon entropy (bits) of each item's
+            posterior -- the per-item uncertainty estimate.
+        confusion_matrices (dict[str, pd.DataFrame]): Per-rater confusion
+            matrix, indexed by true class and labeled by observed class.
+            Each row sums to 1.
+        class_prior (pd.Series): Estimated prevalence of each true class.
+        columns (list of str): Columns used to fit the model.
+        n_iter (int): Number of EM iterations actually run.
+        log_likelihood (list of float): Observed-data log-likelihood at each
+            iteration (should increase monotonically).
+    """
+    if columns is None:
+        columns = list(df.columns)
+
+    labels = df[columns].apply(lambda s: s.astype(str).str.lower())
+    labels = labels.where(df[columns].notna())
+
+    classes = sorted(set(labels.values.flatten()) - {None, np.nan})
+    classes = [c for c in classes if isinstance(c, str)]
+    n_classes = len(classes)
+    class_index = {c: k for k, c in enumerate(classes)}
+
+    n_items = len(df)
+    n_raters = len(columns)
+
+    observed = np.full((n_items, n_raters), -1, dtype=int)
+    for j, col in enumerate(columns):
+        for i, value in enumerate(labels[col].values):
+            if isinstance(value, str):
+                observed[i, j] = class_index[value]
+
+    has_any_label = (observed >= 0).any(axis=1)
+
+    q = np.full((n_items, n_classes), 1.0 / n_classes)
+    for i in range(n_items):
+        if not has_any_label[i]:
+            continue
+        votes = np.zeros(n_classes)
+        for j in range(n_raters):
+            if observed[i, j] >= 0:
+                votes[observed[i, j]] += 1
+        q[i] = votes / votes.sum()
+
+    log_likelihood_history = []
+    n_iter = 0
+
+    for n_iter in range(1, max_iter + 1):
+        # M-step
+        class_prior = q[has_any_label].mean(axis=0)
+        confusion = np.zeros((n_raters, n_classes, n_classes))
+        for j in range(n_raters):
+            mask = observed[:, j] >= 0
+            numer = np.full((n_classes, n_classes), smoothing)
+            denom = np.full(n_classes, smoothing * n_classes)
+            for k in range(n_classes):
+                weight = q[mask, k]
+                denom[k] += weight.sum()
+                for l in range(n_classes):
+                    numer[k, l] += weight[observed[mask, j] == l].sum()
+            confusion[j] = numer / denom[:, None]
+
+        # E-step
+        log_confusion = np.log(confusion)
+        log_prior = np.log(class_prior)
+        log_q_unnorm = np.tile(log_prior, (n_items, 1))
+        for j in range(n_raters):
+            mask = observed[:, j] >= 0
+            log_q_unnorm[mask] += log_confusion[j, :, observed[mask, j]]
+
+        log_norm = _log_sum_exp(log_q_unnorm, axis=1)
+        log_likelihood_history.append(log_norm[has_any_label].sum())
+
+        new_q = np.exp(log_q_unnorm - log_norm[:, None])
+        new_q[~has_any_label] = 1.0 / n_classes
+
+        delta = np.max(np.abs(new_q - q))
+        q = new_q
+        if delta < tol:
+            break
+
+    index = df.index
+    consensus_proba = pd.DataFrame(q, index=index, columns=classes)
+    consensus_proba.loc[~has_any_label, :] = np.nan
+
+    consensus = pd.Series(
+        [classes[k] for k in np.argmax(q, axis=1)], index=index
+    )
+    consensus[~has_any_label] = np.nan
+
+    with np.errstate(divide="ignore", invalid="ignore"):
+        entropy_terms = np.where(q > 0, q * np.log2(q), 0.0)
+    item_entropy = pd.Series(-entropy_terms.sum(axis=1), index=index)
+    item_entropy[~has_any_label] = np.nan
+
+    confusion_matrices = {
+        col: pd.DataFrame(confusion[j], index=classes, columns=classes)
+        for j, col in enumerate(columns)
+    }
+
+    return {
+        "consensus": consensus,
+        "consensus_proba": consensus_proba,
+        "item_entropy": item_entropy,
+        "confusion_matrices": confusion_matrices,
+        "class_prior": pd.Series(class_prior, index=classes),
+        "columns": columns,
+        "n_iter": n_iter,
+        "log_likelihood": log_likelihood_history,
+    }
+
+
+def score_against_consensus(df, result, columns=None):
+    """
+    Score one or more raters/experiments against a Dawid-Skene consensus
+    label, e.g. to rank several LLM labelling experiments without treating
+    any single human rater as ground truth.
+
+    Parameters
+    ----------
+    df (pd.DataFrame): DataFrame containing the columns to score. Must share
+        an index with the DataFrame ``result`` was fit on.
+    result (dict): Return value of ``fit_dawid_skene``.
+    columns (list of str or None): Columns to score. Defaults to the columns
+        the model was fit on (``result['columns']``).
+
+    Returns
+    -------
+    pd.DataFrame
+        Indexed by column name, with columns ``accuracy``, ``macro_f1``, and
+        ``kappa`` computed against the consensus label (rows missing either
+        the consensus or the rater's label are excluded per column). Sorted
+        by ``macro_f1`` descending, so the top row is the best-agreeing
+        experiment.
+    """
+    if columns is None:
+        columns = result["columns"]
+
+    consensus = result["consensus"]
+    rows = []
+    for col in columns:
+        pair = pd.DataFrame({"consensus": consensus, "rater": df[col]}).dropna()
+        true_values = pair["consensus"]
+        pred_values = pair["rater"].astype(str).str.lower()
+        rows.append(
+            {
+                "column": col,
+                "accuracy": accuracy_score(true_values, pred_values),
+                "macro_f1": f1_score(
+                    true_values, pred_values, average="macro", zero_division=0
+                ),
+                "kappa": cohen_kappa_score(true_values, pred_values),
+            }
+        )
+
+    return (
+        pd.DataFrame(rows)
+        .set_index("column")
+        .sort_values("macro_f1", ascending=False)
+    )
+
+
+def plot_dawid_skene_confusion(result, column, ax=None, cmap=None, annot=True):
+    """
+    Plot one rater's estimated Dawid-Skene confusion matrix as a heatmap.
+
+    Parameters
+    ----------
+    result (dict): Return value of ``fit_dawid_skene``.
+    column (str): Which rater's confusion matrix to plot (a key of
+        ``result['confusion_matrices']``).
+    ax (matplotlib axis handle): The axis handle to plot on. Set to None to
+        use the current axis.
+    cmap (matplotlib colormap or None): Colormap for the heatmap. Defaults
+        to ``plt.cm.Blues``.
+    annot (bool): Whether to annotate each cell with its value.
+
+    Returns
+    -------
+    matplotlib.axes.Axes
+        The axis the confusion matrix was drawn on.
+    """
+    matrix = result["confusion_matrices"][column]
+    classes = list(matrix.columns)
+
+    if ax is None:
+        ax = plt.gca()
+    if cmap is None:
+        cmap = plt.cm.Blues
+
+    im = ax.imshow(matrix.values, cmap=cmap, vmin=0, vmax=1)
+
+    n = len(classes)
+    ax.set_xticks(range(n))
+    ax.set_yticks(range(n))
+    ax.set_xticklabels(classes, rotation=45, ha="right")
+    ax.set_yticklabels(classes)
+
+    if annot:
+        for i in range(n):
+            for j in range(n):
+                value = matrix.values[i, j]
+                ax.text(
+                    j, i, f"{value:.2f}", ha="center", va="center",
+                    color="white" if value > 0.5 else "black",
+                )
+
+    ax.figure.colorbar(im, ax=ax, label="P(observed | true)")
+    ax.set_xlabel("Observed label")
+    ax.set_ylabel("True label (estimated)")
+    ax.set_title(f"Dawid-Skene Confusion Matrix: {column}")
+
+    return ax
diff --git a/lars/util/kappa_matrix.py b/lars/util/kappa_matrix.py
new file mode 100644
index 0000000..ac5b45d
--- /dev/null
+++ b/lars/util/kappa_matrix.py
@@ -0,0 +1,96 @@
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from sklearn.metrics import cohen_kappa_score
+
+
+def calculate_kappa_matrix(df, columns=None):
+    """
+    Compute pairwise Cohen's kappa scores between multiple label columns.
+
+    Parameters
+    ----------
+    df (pd.DataFrame): DataFrame containing one column per labeler/source.
+    columns (list of str or None): Columns to compare. Defaults to all
+        columns in ``df``.
+
+    Returns
+    -------
+    pd.DataFrame
+        Square DataFrame indexed and labeled by ``columns``, where entry
+        (i, j) is the Cohen's kappa score between columns i and j (1.0 on
+        the diagonal). Rows with a missing value in either of the two
+        compared columns are excluded from that pair's calculation.
+    """
+    if columns is None:
+        columns = list(df.columns)
+
+    matrix = pd.DataFrame(np.eye(len(columns)), index=columns, columns=columns)
+
+    for i, col_i in enumerate(columns):
+        for col_j in columns[i + 1:]:
+            pair = df[[col_i, col_j]].dropna()
+            a = pair[col_i].astype(str).str.lower()
+            b = pair[col_j].astype(str).str.lower()
+            kappa = cohen_kappa_score(a, b) if len(pair) > 0 else np.nan
+            matrix.loc[col_i, col_j] = kappa
+            matrix.loc[col_j, col_i] = kappa
+
+    return matrix
+
+
+def plot_kappa_matrix(df, columns=None, labels=None, ax=None, cmap=None, annot=True, vmin=-1, vmax=1,
+                      matrix=None):
+    """
+    Plot a matrix of pairwise Cohen's kappa scores between multiple label
+    columns as a heatmap.
+
+    Parameters
+    ----------
+    df (pd.DataFrame): DataFrame containing one column per labeler/source.
+    columns (list of str or None): Columns to compare. Defaults to all
+        columns in ``df``.
+    labels (list of str or None): Display names for the tick labels, in the
+        same order as ``columns``. Defaults to the column names.
+    ax (matplotlib axis handle): The axis handle to plot on. Set to None to
+        use the current axis.
+    cmap (matplotlib colormap or None): Colormap for the heatmap. Defaults
+        to ``plt.cm.RdYlGn``.
+    annot (bool): Whether to annotate each cell with its kappa value.
+    vmin (float): Minimum value for the colormap. Defaults to -1.
+    vmax (float): Maximum value for the colormap. Defaults to 1.
+
+    Returns
+    -------
+    matplotlib.axes.Axes
+        The axis the kappa matrix was drawn on.
+    """
+    if matrix is None:
+        matrix = calculate_kappa_matrix(df, columns=columns)
+    display_labels = labels if labels is not None else list(matrix.columns)
+
+    if ax is None:
+        ax = plt.gca()
+    if cmap is None:
+        cmap = plt.cm.RdYlGn
+
+    im = ax.imshow(matrix.values, cmap=cmap, vmin=vmin, vmax=vmax)
+
+    n = len(display_labels)
+    ax.set_xticks(range(n))
+    ax.set_yticks(range(n))
+    ax.set_xticklabels(display_labels, rotation=45, ha="right")
+    ax.set_yticklabels(display_labels)
+
+    if annot:
+        for i in range(n):
+            for j in range(n):
+                value = matrix.values[i, j]
+                text = "" if np.isnan(value) else f"{value:.2f}"
+                color = "white" if (not np.isnan(value)) and value < 0.5 else "black"
+                ax.text(j, i, text, ha="center", va="center", color=color)
+
+    ax.figure.colorbar(im, ax=ax)
+    ax.set_title("Cohen's Kappa Matrix")
+
+    return ax
diff --git a/lars/util/label_disagreement_matrix.py b/lars/util/label_disagreement_matrix.py
new file mode 100644
index 0000000..e6bb0d7
--- /dev/null
+++ b/lars/util/label_disagreement_matrix.py
@@ -0,0 +1,115 @@
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+
+
+def calculate_label_disagreement_matrix(df, label, columns=None):
+    """
+    Compute a matrix of pairwise item-level disagreement rates for a single
+    label, treated as a one-vs-rest binary classification.
+
+    For each pair of labellers, only rows where both columns have a
+    non-missing value are considered. Each value is binarized to whether it
+    equals ``label`` (case-insensitively), and entry (i, j) of the returned
+    matrix is the percentage of those rows where the two labellers'
+    binarized values differ.
+
+    Unlike ``calculate_label_rate_diff_matrix`` (see ``label_rate_matrix.py``),
+    which only compares how often each labeller uses ``label`` overall, this
+    compares labellers item-by-item, so it can be high even when two
+    labellers use ``label`` at the same overall rate but disagree on which
+    items it applies to.
+
+    Parameters
+    ----------
+    df (pd.DataFrame): DataFrame containing one column per labeller/source.
+    label (str): The label value to treat as the positive class. Compared
+        case-insensitively.
+    columns (list of str or None): Columns (labellers) to compare. Defaults
+        to all columns in ``df``.
+
+    Returns
+    -------
+    pd.DataFrame
+        Square DataFrame indexed and labeled by ``columns``, where entry
+        (i, j) is the item-level disagreement rate (0-100) between columns
+        i and j (0.0 on the diagonal). NaN if a pair shares no non-missing
+        rows.
+    """
+    if columns is None:
+        columns = list(df.columns)
+
+    label_lower = str(label).lower()
+    matrix = pd.DataFrame(0.0, index=columns, columns=columns)
+
+    for i, col_i in enumerate(columns):
+        for col_j in columns[i + 1:]:
+            pair = df[[col_i, col_j]].dropna()
+            if len(pair) == 0:
+                diff = np.nan
+            else:
+                a = pair[col_i].astype(str).str.lower() == label_lower
+                b = pair[col_j].astype(str).str.lower() == label_lower
+                diff = 100.0 * (a != b).mean()
+            matrix.loc[col_i, col_j] = diff
+            matrix.loc[col_j, col_i] = diff
+
+    return matrix
+
+
+def plot_label_disagreement_matrix(df, label, columns=None, labels=None, ax=None,
+                                   cmap=None, annot=True):
+    """
+    Plot a matrix of pairwise item-level disagreement rates for a single
+    label as a heatmap.
+
+    Parameters
+    ----------
+    df (pd.DataFrame): DataFrame containing one column per labeller/source.
+    label (str): The label value to treat as the positive class.
+    columns (list of str or None): Columns to compare. Defaults to all
+        columns in ``df``.
+    labels (list of str or None): Display names for the tick labels, in the
+        same order as ``columns``. Defaults to the column names.
+    ax (matplotlib axis handle): The axis handle to plot on. Set to None to
+        use the current axis.
+    cmap (matplotlib colormap or None): Colormap for the heatmap. Defaults
+        to ``plt.cm.Reds``.
+    annot (bool): Whether to annotate each cell with its value.
+
+    Returns
+    -------
+    matplotlib.axes.Axes
+        The axis the matrix was drawn on.
+    """
+    matrix = calculate_label_disagreement_matrix(df, label, columns=columns)
+    display_labels = labels if labels is not None else list(matrix.columns)
+
+    if ax is None:
+        ax = plt.gca()
+    if cmap is None:
+        cmap = plt.cm.Reds
+
+    finite_values = matrix.values[~np.isnan(matrix.values)]
+    vmax = finite_values.max() if finite_values.size > 0 and finite_values.max() > 0 else 100.0
+
+    im = ax.imshow(matrix.values, cmap=cmap, vmin=0, vmax=vmax)
+
+    n = len(display_labels)
+    ax.set_xticks(range(n))
+    ax.set_yticks(range(n))
+    ax.set_xticklabels(display_labels, rotation=45, ha="right")
+    ax.set_yticklabels(display_labels)
+
+    if annot:
+        for i in range(n):
+            for j in range(n):
+                value = matrix.values[i, j]
+                text = "" if np.isnan(value) else f"{value:.1f}"
+                color = "white" if (not np.isnan(value)) and value > vmax / 2 else "black"
+                ax.text(j, i, text, ha="center", va="center", color=color)
+
+    ax.figure.colorbar(im, ax=ax, label="Disagreement rate (%)")
+    ax.set_title(f"Item-Level Disagreement: '{label}'")
+
+    return ax
diff --git a/lars/util/label_rate_matrix.py b/lars/util/label_rate_matrix.py
new file mode 100644
index 0000000..828b013
--- /dev/null
+++ b/lars/util/label_rate_matrix.py
@@ -0,0 +1,111 @@
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+
+
+def calculate_label_rate_diff_matrix(df, label, columns=None):
+    """
+    Compute a matrix of pairwise usage-rate differences for a single label
+    across multiple labellers.
+
+    For each column (labeller), the usage rate is the percentage of its
+    non-missing values that equal ``label``. Entry (i, j) of the returned
+    matrix is the absolute difference, in percentage points, between the
+    usage rates of labellers i and j.
+
+    Unlike Cohen's kappa (see ``kappa_matrix.py``), this does not require
+    labellers to agree item-by-item -- it only compares how often each
+    labeller applies ``label`` overall, so it can flag labellers that are
+    systematically biased toward or away from a class even when they are
+    scored on different items.
+
+    Parameters
+    ----------
+    df (pd.DataFrame): DataFrame containing one column per labeller/source.
+    label (str): The label value to compute usage rates for. Compared
+        case-insensitively.
+    columns (list of str or None): Columns (labellers) to compare. Defaults
+        to all columns in ``df``.
+
+    Returns
+    -------
+    pd.DataFrame
+        Square DataFrame indexed and labeled by ``columns``, where entry
+        (i, j) is the absolute usage-rate difference (0-100) between
+        columns i and j (0.0 on the diagonal).
+    """
+    if columns is None:
+        columns = list(df.columns)
+
+    label_lower = str(label).lower()
+    rates = {}
+    for col in columns:
+        values = df[col].dropna().astype(str).str.lower()
+        rates[col] = 100.0 * (values == label_lower).mean() if len(values) > 0 else np.nan
+
+    matrix = pd.DataFrame(0.0, index=columns, columns=columns)
+    for i, col_i in enumerate(columns):
+        for col_j in columns[i + 1:]:
+            diff = abs(rates[col_i] - rates[col_j])
+            matrix.loc[col_i, col_j] = diff
+            matrix.loc[col_j, col_i] = diff
+
+    return matrix
+
+
+def plot_label_rate_diff_matrix(df, label, columns=None, labels=None, ax=None,
+                                cmap=None, annot=True):
+    """
+    Plot a matrix of pairwise usage-rate differences for a single label
+    across multiple labellers as a heatmap.
+
+    Parameters
+    ----------
+    df (pd.DataFrame): DataFrame containing one column per labeller/source.
+    label (str): The label value to compute usage rates for.
+    columns (list of str or None): Columns to compare. Defaults to all
+        columns in ``df``.
+    labels (list of str or None): Display names for the tick labels, in the
+        same order as ``columns``. Defaults to the column names.
+    ax (matplotlib axis handle): The axis handle to plot on. Set to None to
+        use the current axis.
+    cmap (matplotlib colormap or None): Colormap for the heatmap. Defaults
+        to ``plt.cm.Reds``.
+    annot (bool): Whether to annotate each cell with its value.
+
+    Returns
+    -------
+    matplotlib.axes.Axes
+        The axis the matrix was drawn on.
+    """
+    matrix = calculate_label_rate_diff_matrix(df, label, columns=columns)
+    display_labels = labels if labels is not None else list(matrix.columns)
+
+    if ax is None:
+        ax = plt.gca()
+    if cmap is None:
+        cmap = plt.cm.Reds
+
+    finite_values = matrix.values[~np.isnan(matrix.values)]
+    vmax = finite_values.max() if finite_values.size > 0 and finite_values.max() > 0 else 100.0
+
+    im = ax.imshow(matrix.values, cmap=cmap, vmin=0, vmax=vmax)
+
+    n = len(display_labels)
+    ax.set_xticks(range(n))
+    ax.set_yticks(range(n))
+    ax.set_xticklabels(display_labels, rotation=45, ha="right")
+    ax.set_yticklabels(display_labels)
+
+    if annot:
+        for i in range(n):
+            for j in range(n):
+                value = matrix.values[i, j]
+                text = "" if np.isnan(value) else f"{value:.1f}"
+                color = "white" if (not np.isnan(value)) and value > vmax / 2 else "black"
+                ax.text(j, i, text, ha="center", va="center", color=color)
+
+    ax.figure.colorbar(im, ax=ax, label="Usage rate difference (pp)")
+    ax.set_title(f"Usage Rate Difference: '{label}'")
+
+    return ax
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.000231.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.000231.png
new file mode 100644
index 0000000..d7a2c5d
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.000231.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.001232.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.001232.png
new file mode 100644
index 0000000..bc93a09
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.001232.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.002234.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.002234.png
new file mode 100644
index 0000000..df586d0
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.002234.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.003237.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.003237.png
new file mode 100644
index 0000000..1362111
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.003237.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.004239.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.004239.png
new file mode 100644
index 0000000..6cb3246
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.004239.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.005240.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.005240.png
new file mode 100644
index 0000000..18e671f
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.005240.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.010243.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.010243.png
new file mode 100644
index 0000000..cddb026
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.010243.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.011001.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.011001.png
new file mode 100644
index 0000000..d9a0784
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.011001.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.012002.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.012002.png
new file mode 100644
index 0000000..6588580
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.012002.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.013005.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.013005.png
new file mode 100644
index 0000000..623a3d2
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.013005.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.014022.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.014022.png
new file mode 100644
index 0000000..313c223
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.014022.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.015024.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.015024.png
new file mode 100644
index 0000000..bf2371f
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.015024.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.020026.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.020026.png
new file mode 100644
index 0000000..cb374cb
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.020026.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.022000.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.022000.png
new file mode 100644
index 0000000..7cacd30
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.022000.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.023004.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.023004.png
new file mode 100644
index 0000000..26aaae7
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.023004.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.024005.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.024005.png
new file mode 100644
index 0000000..49fa451
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.024005.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.025006.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.025006.png
new file mode 100644
index 0000000..14da98f
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.025006.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.030007.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.030007.png
new file mode 100644
index 0000000..288f927
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.030007.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.031008.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.031008.png
new file mode 100644
index 0000000..8070932
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.031008.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.032009.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.032009.png
new file mode 100644
index 0000000..c6df1f6
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.032009.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.033010.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.033010.png
new file mode 100644
index 0000000..f9ecc3a
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.033010.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.034011.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.034011.png
new file mode 100644
index 0000000..1ba551e
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.034011.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.035012.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.035012.png
new file mode 100644
index 0000000..15245b5
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.035012.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.040012.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.040012.png
new file mode 100644
index 0000000..f9906df
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.040012.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.041013.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.041013.png
new file mode 100644
index 0000000..05c8355
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.041013.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.042014.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.042014.png
new file mode 100644
index 0000000..63d2ebb
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.042014.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.043004.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.043004.png
new file mode 100644
index 0000000..a3a0f01
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.043004.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.044016.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.044016.png
new file mode 100644
index 0000000..4bd84b2
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.044016.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.045052.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.045052.png
new file mode 100644
index 0000000..f409f68
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.045052.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.050053.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.050053.png
new file mode 100644
index 0000000..e811151
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.050053.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.051054.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.051054.png
new file mode 100644
index 0000000..384223d
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.051054.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.052055.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.052055.png
new file mode 100644
index 0000000..656ace1
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.052055.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.053002.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.053002.png
new file mode 100644
index 0000000..72010b1
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.053002.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.054001.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.054001.png
new file mode 100644
index 0000000..eee4be4
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.054001.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.055004.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.055004.png
new file mode 100644
index 0000000..4632a12
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.055004.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.060005.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.060005.png
new file mode 100644
index 0000000..6442f19
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.060005.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.061007.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.061007.png
new file mode 100644
index 0000000..dfeb81b
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.061007.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.062009.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.062009.png
new file mode 100644
index 0000000..2ff14a3
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.062009.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.063013.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.063013.png
new file mode 100644
index 0000000..4562f6e
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.063013.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.064016.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.064016.png
new file mode 100644
index 0000000..e3774da
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.064016.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.065019.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.065019.png
new file mode 100644
index 0000000..96eaa40
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.065019.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.070021.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.070021.png
new file mode 100644
index 0000000..11888da
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.070021.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.071025.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.071025.png
new file mode 100644
index 0000000..5af8362
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.071025.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.072027.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.072027.png
new file mode 100644
index 0000000..9d0b404
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.072027.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.073030.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.073030.png
new file mode 100644
index 0000000..b7270ab
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.073030.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.074018.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.074018.png
new file mode 100644
index 0000000..250d6ac
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.074018.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.075002.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.075002.png
new file mode 100644
index 0000000..1275951
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.075002.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.080020.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.080020.png
new file mode 100644
index 0000000..ccfe116
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.080020.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.081025.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.081025.png
new file mode 100644
index 0000000..f2018fa
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.081025.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.082027.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.082027.png
new file mode 100644
index 0000000..bb8a499
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.082027.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.083029.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.083029.png
new file mode 100644
index 0000000..b477843
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.083029.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.084033.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.084033.png
new file mode 100644
index 0000000..c803508
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.084033.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.085035.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.085035.png
new file mode 100644
index 0000000..b700a4c
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.085035.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.090105.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.090105.png
new file mode 100644
index 0000000..2083e18
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.090105.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.091107.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.091107.png
new file mode 100644
index 0000000..4172737
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.091107.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.092110.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.092110.png
new file mode 100644
index 0000000..fe382a8
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.092110.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.093112.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.093112.png
new file mode 100644
index 0000000..1dc33a9
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.093112.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.094114.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.094114.png
new file mode 100644
index 0000000..406c258
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.094114.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.095117.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.095117.png
new file mode 100644
index 0000000..079a15f
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.095117.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.100119.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.100119.png
new file mode 100644
index 0000000..0651a5f
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.100119.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.101141.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.101141.png
new file mode 100644
index 0000000..214ebe6
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.101141.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.102143.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.102143.png
new file mode 100644
index 0000000..e308b03
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.102143.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.103147.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.103147.png
new file mode 100644
index 0000000..17f5293
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.103147.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.104149.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.104149.png
new file mode 100644
index 0000000..aaf518f
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.104149.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.105153.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.105153.png
new file mode 100644
index 0000000..e74f167
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.105153.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.110155.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.110155.png
new file mode 100644
index 0000000..2d2e42e
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.110155.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.111158.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.111158.png
new file mode 100644
index 0000000..e08528b
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.111158.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.112201.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.112201.png
new file mode 100644
index 0000000..982a292
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.112201.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.113205.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.113205.png
new file mode 100644
index 0000000..2587fab
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.113205.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.114207.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.114207.png
new file mode 100644
index 0000000..ccb373b
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.114207.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.115209.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.115209.png
new file mode 100644
index 0000000..170fb1b
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.115209.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.120212.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.120212.png
new file mode 100644
index 0000000..2e217c0
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.120212.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.121214.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.121214.png
new file mode 100644
index 0000000..f41c25f
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.121214.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.122216.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.122216.png
new file mode 100644
index 0000000..bf67fb4
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.122216.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.123002.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.123002.png
new file mode 100644
index 0000000..de64029
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.123002.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.124006.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.124006.png
new file mode 100644
index 0000000..68e2105
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.124006.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.125010.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.125010.png
new file mode 100644
index 0000000..6544090
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.125010.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.130013.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.130013.png
new file mode 100644
index 0000000..b1192e4
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.130013.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.131015.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.131015.png
new file mode 100644
index 0000000..875cd3d
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.131015.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.132017.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.132017.png
new file mode 100644
index 0000000..09fd9ec
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.132017.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.133021.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.133021.png
new file mode 100644
index 0000000..922b53b
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.133021.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.134023.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.134023.png
new file mode 100644
index 0000000..4b348dc
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.134023.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.135027.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.135027.png
new file mode 100644
index 0000000..09db53f
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.135027.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.140030.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.140030.png
new file mode 100644
index 0000000..b4d17f0
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.140030.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.141032.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.141032.png
new file mode 100644
index 0000000..55a7315
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.141032.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.142036.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.142036.png
new file mode 100644
index 0000000..f9ab597
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.142036.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.143002.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.143002.png
new file mode 100644
index 0000000..12bf3fa
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.143002.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.144002.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.144002.png
new file mode 100644
index 0000000..664d9b1
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.144002.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.145003.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.145003.png
new file mode 100644
index 0000000..06f72a1
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.145003.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.150003.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.150003.png
new file mode 100644
index 0000000..cccb49b
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.150003.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.151003.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.151003.png
new file mode 100644
index 0000000..d5873c5
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.151003.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.152015.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.152015.png
new file mode 100644
index 0000000..854db41
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.152015.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.153016.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.153016.png
new file mode 100644
index 0000000..981d853
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.153016.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.154017.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.154017.png
new file mode 100644
index 0000000..df2eaea
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.154017.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.155024.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.155024.png
new file mode 100644
index 0000000..dba4892
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.155024.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.160025.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.160025.png
new file mode 100644
index 0000000..bb0caf1
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.160025.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.161028.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.161028.png
new file mode 100644
index 0000000..9a81b19
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.161028.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.162028.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.162028.png
new file mode 100644
index 0000000..0924211
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.162028.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.163030.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.163030.png
new file mode 100644
index 0000000..37b5fc2
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.163030.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.164032.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.164032.png
new file mode 100644
index 0000000..994f572
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.164032.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.165034.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.165034.png
new file mode 100644
index 0000000..b645098
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.165034.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.170035.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.170035.png
new file mode 100644
index 0000000..0b79008
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.170035.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.171036.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.171036.png
new file mode 100644
index 0000000..52e223e
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.171036.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.172038.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.172038.png
new file mode 100644
index 0000000..d2fd062
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.172038.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.173040.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.173040.png
new file mode 100644
index 0000000..238b0bb
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.173040.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.174042.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.174042.png
new file mode 100644
index 0000000..bb59727
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.174042.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.175043.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.175043.png
new file mode 100644
index 0000000..a806ed3
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.175043.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.180045.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.180045.png
new file mode 100644
index 0000000..c1d6d37
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.180045.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.181047.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.181047.png
new file mode 100644
index 0000000..1b21b43
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.181047.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.182048.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.182048.png
new file mode 100644
index 0000000..a9f9fbc
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.182048.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.183049.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.183049.png
new file mode 100644
index 0000000..cecdd66
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.183049.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.184050.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.184050.png
new file mode 100644
index 0000000..2a5288f
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.184050.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.185050.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.185050.png
new file mode 100644
index 0000000..8f9a0fe
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.185050.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.190050.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.190050.png
new file mode 100644
index 0000000..43e2362
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.190050.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.191051.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.191051.png
new file mode 100644
index 0000000..bc1c1e2
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.191051.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.192115.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.192115.png
new file mode 100644
index 0000000..6c50714
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.192115.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.193116.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.193116.png
new file mode 100644
index 0000000..2d5fbe0
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.193116.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.194116.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.194116.png
new file mode 100644
index 0000000..b8c1aa9
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.194116.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.200000.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.200000.png
new file mode 100644
index 0000000..932a205
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.200000.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.201002.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.201002.png
new file mode 100644
index 0000000..f3a25a8
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.201002.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.202004.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.202004.png
new file mode 100644
index 0000000..0a698be
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.202004.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.203006.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.203006.png
new file mode 100644
index 0000000..3cbdf0c
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.203006.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.204009.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.204009.png
new file mode 100644
index 0000000..82a638f
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.204009.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.205011.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.205011.png
new file mode 100644
index 0000000..3e3b0ca
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.205011.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.210012.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.210012.png
new file mode 100644
index 0000000..858fb91
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.210012.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.211014.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.211014.png
new file mode 100644
index 0000000..1968358
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.211014.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.212016.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.212016.png
new file mode 100644
index 0000000..95117e9
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.212016.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.213018.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.213018.png
new file mode 100644
index 0000000..4e23571
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.213018.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.214020.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.214020.png
new file mode 100644
index 0000000..1d093be
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.214020.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.215048.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.215048.png
new file mode 100644
index 0000000..3c5500c
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.215048.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.220050.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.220050.png
new file mode 100644
index 0000000..c252638
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.220050.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.221053.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.221053.png
new file mode 100644
index 0000000..bb26409
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.221053.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.222055.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.222055.png
new file mode 100644
index 0000000..f7c5f60
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.222055.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.223057.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.223057.png
new file mode 100644
index 0000000..9e857d7
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.223057.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.224030.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.224030.png
new file mode 100644
index 0000000..9891e22
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.224030.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.225032.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.225032.png
new file mode 100644
index 0000000..462b7c7
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.225032.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.230034.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.230034.png
new file mode 100644
index 0000000..9336fcc
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.230034.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.231036.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.231036.png
new file mode 100644
index 0000000..2f5d7af
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.231036.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.232038.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.232038.png
new file mode 100644
index 0000000..54fd341
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.232038.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.233040.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.233040.png
new file mode 100644
index 0000000..605c325
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.233040.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.234042.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.234042.png
new file mode 100644
index 0000000..dca0386
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.234042.png differ
diff --git a/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.235045.png b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.235045.png
new file mode 100644
index 0000000..65ae550
Binary files /dev/null and b/tests/data/test_images/bnfcsapr2cmacS3.c1.20250527.235045.png differ
diff --git a/tests/test_dawid_skene.py b/tests/test_dawid_skene.py
new file mode 100644
index 0000000..561e903
--- /dev/null
+++ b/tests/test_dawid_skene.py
@@ -0,0 +1,123 @@
+import pytest
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+
+
+@pytest.fixture
+def sample_df():
+    n_cycles = 10
+    true_pattern = ["convective", "stratiform", "anvil"] * n_cycles
+
+    rater_a = list(true_pattern)
+    rater_b = list(true_pattern)
+    rater_b[0], rater_b[1] = "stratiform", "convective"
+    rater_c = ["convective" if label == "anvil" else label for label in true_pattern]
+
+    # An extra row where all three raters pick a different class: genuinely
+    # ambiguous, unlike the systematic rater_c bias above (which the model
+    # learns to discount and remains confident about).
+    rater_a.append("convective")
+    rater_b.append("stratiform")
+    rater_c.append("anvil")
+    exp_good = true_pattern + ["convective"]
+    exp_bad = ["anvil"] * len(true_pattern) + ["anvil"]
+
+    return pd.DataFrame({
+        "rater_a": rater_a,
+        "rater_b": rater_b,
+        "rater_c": rater_c,
+        "exp_good": exp_good,
+        "exp_bad": exp_bad,
+    })
+
+
+@pytest.fixture(autouse=True)
+def close_figures():
+    yield
+    plt.close("all")
+
+
+@pytest.fixture
+def fitted(sample_df):
+    from lars.util.dawid_skene import fit_dawid_skene
+
+    return fit_dawid_skene(sample_df, columns=["rater_a", "rater_b", "rater_c"])
+
+
+def test_confusion_matrices_are_row_stochastic(fitted):
+    for matrix in fitted["confusion_matrices"].values():
+        np.testing.assert_array_almost_equal(matrix.sum(axis=1).values, np.ones(len(matrix)))
+
+
+def test_confusion_matrices_indexed_by_columns(fitted, sample_df):
+    assert set(fitted["confusion_matrices"].keys()) == {"rater_a", "rater_b", "rater_c"}
+    for matrix in fitted["confusion_matrices"].values():
+        assert set(matrix.index) == {"convective", "stratiform", "anvil"}
+        assert set(matrix.columns) == {"convective", "stratiform", "anvil"}
+
+
+def test_noisy_rater_has_lower_diagonal_mass_for_biased_class(fitted):
+    clean = fitted["confusion_matrices"]["rater_a"].loc["anvil", "anvil"]
+    noisy = fitted["confusion_matrices"]["rater_c"].loc["anvil", "anvil"]
+    assert clean > noisy
+
+
+def test_consensus_recovers_true_pattern(fitted, sample_df):
+    cyclic_rows = slice(0, -1)
+    assert (fitted["consensus"].values[cyclic_rows] == sample_df["exp_good"].values[cyclic_rows]).all()
+
+
+def test_unanimous_row_has_near_zero_entropy(fitted):
+    assert fitted["item_entropy"].iloc[3] < 0.01
+
+
+def test_three_way_disagreement_has_higher_entropy_than_unanimous(fitted):
+    assert fitted["item_entropy"].iloc[-1] > fitted["item_entropy"].iloc[3]
+
+
+def test_consensus_proba_rows_sum_to_one(fitted):
+    np.testing.assert_array_almost_equal(
+        fitted["consensus_proba"].sum(axis=1).values, np.ones(len(fitted["consensus_proba"]))
+    )
+
+
+def test_subset_of_columns(sample_df):
+    from lars.util.dawid_skene import fit_dawid_skene
+
+    result = fit_dawid_skene(sample_df, columns=["rater_a", "rater_c"])
+    assert result["columns"] == ["rater_a", "rater_c"]
+    assert set(result["confusion_matrices"].keys()) == {"rater_a", "rater_c"}
+
+
+def test_score_against_consensus_ranks_good_above_bad(fitted, sample_df):
+    from lars.util.dawid_skene import score_against_consensus
+
+    scores = score_against_consensus(sample_df, fitted, columns=["exp_good", "exp_bad"])
+    assert list(scores.index)[0] == "exp_good"
+    assert scores.loc["exp_good", "accuracy"] == pytest.approx(1.0)
+    assert scores.loc["exp_good", "macro_f1"] > scores.loc["exp_bad", "macro_f1"]
+
+
+def test_score_against_consensus_sorted_by_macro_f1(fitted, sample_df):
+    from lars.util.dawid_skene import score_against_consensus
+
+    scores = score_against_consensus(sample_df, fitted, columns=["exp_bad", "exp_good"])
+    assert scores["macro_f1"].is_monotonic_decreasing
+
+
+def test_plot_returns_axes(fitted):
+    from lars.util.dawid_skene import plot_dawid_skene_confusion
+
+    _, ax = plt.subplots()
+    result = plot_dawid_skene_confusion(fitted, "rater_a", ax=ax)
+    assert result is ax
+
+
+def test_plot_tick_labels_match_classes(fitted):
+    from lars.util.dawid_skene import plot_dawid_skene_confusion
+
+    _, ax = plt.subplots()
+    plot_dawid_skene_confusion(fitted, "rater_a", ax=ax)
+    labels = set(t.get_text() for t in ax.get_xticklabels())
+    assert labels == {"convective", "stratiform", "anvil"}
diff --git a/tests/test_kappa_matrix.py b/tests/test_kappa_matrix.py
new file mode 100644
index 0000000..91eb533
--- /dev/null
+++ b/tests/test_kappa_matrix.py
@@ -0,0 +1,114 @@
+import io
+import pytest
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+
+CSV_DATA = """\
+rater_a,rater_b,rater_c
+stratiform,stratiform,stratiform
+convective,convective,stratiform
+stratiform,convective,convective
+convective,stratiform,convective
+anvil,anvil,anvil
+stratiform,stratiform,convective
+"""
+
+
+@pytest.fixture
+def sample_df():
+    return pd.read_csv(io.StringIO(CSV_DATA))
+
+
+@pytest.fixture(autouse=True)
+def close_figures():
+    yield
+    plt.close("all")
+
+
+def test_matrix_is_square_and_symmetric(sample_df):
+    from lars.util.kappa_matrix import calculate_kappa_matrix
+
+    matrix = calculate_kappa_matrix(sample_df)
+    assert list(matrix.columns) == ["rater_a", "rater_b", "rater_c"]
+    assert list(matrix.index) == ["rater_a", "rater_b", "rater_c"]
+    np.testing.assert_array_almost_equal(matrix.values, matrix.values.T)
+
+
+def test_diagonal_is_one(sample_df):
+    from lars.util.kappa_matrix import calculate_kappa_matrix
+
+    matrix = calculate_kappa_matrix(sample_df)
+    np.testing.assert_array_almost_equal(np.diag(matrix.values), np.ones(3))
+
+
+def test_matrix_matches_pairwise_kappa(sample_df):
+    from lars.util.kappa_matrix import calculate_kappa_matrix
+    from lars.util.confusion_matrix import calculate_cohen_kappa
+
+    matrix = calculate_kappa_matrix(sample_df)
+    expected = calculate_cohen_kappa(
+        sample_df.rename(columns={"rater_a": "label", "rater_b": "llm_label"})
+    )
+    assert matrix.loc["rater_a", "rater_b"] == pytest.approx(expected)
+
+
+def test_subset_of_columns(sample_df):
+    from lars.util.kappa_matrix import calculate_kappa_matrix
+
+    matrix = calculate_kappa_matrix(sample_df, columns=["rater_a", "rater_c"])
+    assert list(matrix.columns) == ["rater_a", "rater_c"]
+
+
+def test_excludes_rows_with_missing_values():
+    from lars.util.kappa_matrix import calculate_kappa_matrix
+
+    df = pd.DataFrame({
+        "a": ["x", "y", "x", None],
+        "b": ["x", "y", "y", "x"],
+    })
+    matrix = calculate_kappa_matrix(df)
+    assert not np.isnan(matrix.loc["a", "b"])
+
+
+def test_plot_returns_axes(sample_df):
+    from lars.util.kappa_matrix import plot_kappa_matrix
+
+    _, ax = plt.subplots()
+    result = plot_kappa_matrix(sample_df, ax=ax)
+    assert result is ax
+
+
+def test_plot_title_is_set(sample_df):
+    from lars.util.kappa_matrix import plot_kappa_matrix
+
+    _, ax = plt.subplots()
+    plot_kappa_matrix(sample_df, ax=ax)
+    assert ax.get_title() == "Cohen's Kappa Matrix"
+
+
+def test_plot_tick_labels_match_columns(sample_df):
+    from lars.util.kappa_matrix import plot_kappa_matrix
+
+    _, ax = plt.subplots()
+    plot_kappa_matrix(sample_df, ax=ax)
+    labels = [t.get_text() for t in ax.get_xticklabels()]
+    assert labels == ["rater_a", "rater_b", "rater_c"]
+
+
+def test_plot_custom_labels(sample_df):
+    from lars.util.kappa_matrix import plot_kappa_matrix
+
+    _, ax = plt.subplots()
+    plot_kappa_matrix(sample_df, ax=ax, labels=["Alice", "Bob", "LLM"])
+    labels = [t.get_text() for t in ax.get_xticklabels()]
+    assert labels == ["Alice", "Bob", "LLM"]
+
+
+def test_plot_uses_gca_when_no_ax(sample_df):
+    from lars.util.kappa_matrix import plot_kappa_matrix
+
+    fig, ax = plt.subplots()
+    plt.sca(ax)
+    plot_kappa_matrix(sample_df)
+    assert ax.get_title() == "Cohen's Kappa Matrix"
diff --git a/tests/test_label_disagreement_matrix.py b/tests/test_label_disagreement_matrix.py
new file mode 100644
index 0000000..c6f3f7b
--- /dev/null
+++ b/tests/test_label_disagreement_matrix.py
@@ -0,0 +1,141 @@
+import io
+import pytest
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+
+CSV_DATA = """\
+rater_a,rater_b,rater_c
+stratiform,stratiform,stratiform
+convective,convective,stratiform
+stratiform,convective,convective
+convective,stratiform,convective
+anvil,anvil,anvil
+stratiform,stratiform,convective
+"""
+
+
+@pytest.fixture
+def sample_df():
+    return pd.read_csv(io.StringIO(CSV_DATA))
+
+
+@pytest.fixture(autouse=True)
+def close_figures():
+    yield
+    plt.close("all")
+
+
+def test_matrix_is_square_and_symmetric(sample_df):
+    from lars.util.label_disagreement_matrix import calculate_label_disagreement_matrix
+
+    matrix = calculate_label_disagreement_matrix(sample_df, "stratiform")
+    assert list(matrix.columns) == ["rater_a", "rater_b", "rater_c"]
+    assert list(matrix.index) == ["rater_a", "rater_b", "rater_c"]
+    np.testing.assert_array_almost_equal(matrix.values, matrix.values.T)
+
+
+def test_diagonal_is_zero(sample_df):
+    from lars.util.label_disagreement_matrix import calculate_label_disagreement_matrix
+
+    matrix = calculate_label_disagreement_matrix(sample_df, "stratiform")
+    np.testing.assert_array_almost_equal(np.diag(matrix.values), np.zeros(3))
+
+
+def test_disagreement_values(sample_df):
+    """Binarized 'stratiform' (T/F) per row:
+    a: T F T F F T
+    b: T F F T F T
+    c: T T F F F F
+    a vs b disagree at rows 2,3 -> 2/6
+    a vs c disagree at rows 1,2,5 -> 3/6
+    b vs c disagree at rows 1,3,5 -> 3/6
+    """
+    from lars.util.label_disagreement_matrix import calculate_label_disagreement_matrix
+
+    matrix = calculate_label_disagreement_matrix(sample_df, "stratiform")
+    assert matrix.loc["rater_a", "rater_b"] == pytest.approx(100 / 3)
+    assert matrix.loc["rater_a", "rater_c"] == pytest.approx(50.0)
+    assert matrix.loc["rater_b", "rater_c"] == pytest.approx(50.0)
+
+
+def test_label_is_case_insensitive(sample_df):
+    from lars.util.label_disagreement_matrix import calculate_label_disagreement_matrix
+
+    lower = calculate_label_disagreement_matrix(sample_df, "stratiform")
+    upper = calculate_label_disagreement_matrix(sample_df, "STRATIFORM")
+    np.testing.assert_array_almost_equal(lower.values, upper.values)
+
+
+def test_subset_of_columns(sample_df):
+    from lars.util.label_disagreement_matrix import calculate_label_disagreement_matrix
+
+    matrix = calculate_label_disagreement_matrix(sample_df, "stratiform", columns=["rater_a", "rater_c"])
+    assert list(matrix.columns) == ["rater_a", "rater_c"]
+
+
+def test_excludes_rows_with_missing_values():
+    from lars.util.label_disagreement_matrix import calculate_label_disagreement_matrix
+
+    df = pd.DataFrame({
+        "a": ["x", "y", "x", None],
+        "b": ["x", "y", "y", "x"],
+    })
+    matrix = calculate_label_disagreement_matrix(df, "x")
+    # Row 4 dropped (a missing). Remaining rows: (x,x) agree, (y,y) agree
+    # (both non-x -> agree), (x,y) disagree -> 1/3 disagreement.
+    assert matrix.loc["a", "b"] == pytest.approx(100 / 3)
+
+
+def test_nan_when_no_shared_rows():
+    from lars.util.label_disagreement_matrix import calculate_label_disagreement_matrix
+
+    df = pd.DataFrame({
+        "a": ["x", None],
+        "b": [None, "y"],
+    })
+    matrix = calculate_label_disagreement_matrix(df, "x")
+    assert np.isnan(matrix.loc["a", "b"])
+
+
+def test_plot_returns_axes(sample_df):
+    from lars.util.label_disagreement_matrix import plot_label_disagreement_matrix
+
+    _, ax = plt.subplots()
+    result = plot_label_disagreement_matrix(sample_df, "stratiform", ax=ax)
+    assert result is ax
+
+
+def test_plot_title_includes_label(sample_df):
+    from lars.util.label_disagreement_matrix import plot_label_disagreement_matrix
+
+    _, ax = plt.subplots()
+    plot_label_disagreement_matrix(sample_df, "stratiform", ax=ax)
+    assert ax.get_title() == "Item-Level Disagreement: 'stratiform'"
+
+
+def test_plot_tick_labels_match_columns(sample_df):
+    from lars.util.label_disagreement_matrix import plot_label_disagreement_matrix
+
+    _, ax = plt.subplots()
+    plot_label_disagreement_matrix(sample_df, "stratiform", ax=ax)
+    labels = [t.get_text() for t in ax.get_xticklabels()]
+    assert labels == ["rater_a", "rater_b", "rater_c"]
+
+
+def test_plot_custom_labels(sample_df):
+    from lars.util.label_disagreement_matrix import plot_label_disagreement_matrix
+
+    _, ax = plt.subplots()
+    plot_label_disagreement_matrix(sample_df, "stratiform", ax=ax, labels=["Alice", "Bob", "LLM"])
+    labels = [t.get_text() for t in ax.get_xticklabels()]
+    assert labels == ["Alice", "Bob", "LLM"]
+
+
+def test_plot_uses_gca_when_no_ax(sample_df):
+    from lars.util.label_disagreement_matrix import plot_label_disagreement_matrix
+
+    fig, ax = plt.subplots()
+    plt.sca(ax)
+    plot_label_disagreement_matrix(sample_df, "stratiform")
+    assert ax.get_title() == "Item-Level Disagreement: 'stratiform'"
diff --git a/tests/test_label_rate_matrix.py b/tests/test_label_rate_matrix.py
new file mode 100644
index 0000000..f825319
--- /dev/null
+++ b/tests/test_label_rate_matrix.py
@@ -0,0 +1,122 @@
+import io
+import pytest
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+
+CSV_DATA = """\
+rater_a,rater_b,rater_c
+stratiform,stratiform,stratiform
+convective,convective,stratiform
+stratiform,convective,convective
+convective,stratiform,convective
+anvil,anvil,anvil
+stratiform,stratiform,convective
+"""
+
+
+@pytest.fixture
+def sample_df():
+    return pd.read_csv(io.StringIO(CSV_DATA))
+
+
+@pytest.fixture(autouse=True)
+def close_figures():
+    yield
+    plt.close("all")
+
+
+def test_matrix_is_square_and_symmetric(sample_df):
+    from lars.util.label_rate_matrix import calculate_label_rate_diff_matrix
+
+    matrix = calculate_label_rate_diff_matrix(sample_df, "stratiform")
+    assert list(matrix.columns) == ["rater_a", "rater_b", "rater_c"]
+    assert list(matrix.index) == ["rater_a", "rater_b", "rater_c"]
+    np.testing.assert_array_almost_equal(matrix.values, matrix.values.T)
+
+
+def test_diagonal_is_zero(sample_df):
+    from lars.util.label_rate_matrix import calculate_label_rate_diff_matrix
+
+    matrix = calculate_label_rate_diff_matrix(sample_df, "stratiform")
+    np.testing.assert_array_almost_equal(np.diag(matrix.values), np.zeros(3))
+
+
+def test_usage_rate_diff_values(sample_df):
+    """rater_a: 3/6=50% stratiform, rater_b: 3/6=50%, rater_c: 2/6=33.33%."""
+    from lars.util.label_rate_matrix import calculate_label_rate_diff_matrix
+
+    matrix = calculate_label_rate_diff_matrix(sample_df, "stratiform")
+    assert matrix.loc["rater_a", "rater_b"] == pytest.approx(0.0)
+    assert matrix.loc["rater_a", "rater_c"] == pytest.approx(100 / 6)
+    assert matrix.loc["rater_b", "rater_c"] == pytest.approx(100 / 6)
+
+
+def test_label_is_case_insensitive(sample_df):
+    from lars.util.label_rate_matrix import calculate_label_rate_diff_matrix
+
+    lower = calculate_label_rate_diff_matrix(sample_df, "stratiform")
+    upper = calculate_label_rate_diff_matrix(sample_df, "STRATIFORM")
+    np.testing.assert_array_almost_equal(lower.values, upper.values)
+
+
+def test_subset_of_columns(sample_df):
+    from lars.util.label_rate_matrix import calculate_label_rate_diff_matrix
+
+    matrix = calculate_label_rate_diff_matrix(sample_df, "stratiform", columns=["rater_a", "rater_c"])
+    assert list(matrix.columns) == ["rater_a", "rater_c"]
+
+
+def test_ignores_missing_values():
+    from lars.util.label_rate_matrix import calculate_label_rate_diff_matrix
+
+    df = pd.DataFrame({
+        "a": ["x", "x", None],
+        "b": ["x", "y", "y"],
+    })
+    matrix = calculate_label_rate_diff_matrix(df, "x")
+    # a: 2/2 non-missing are 'x' -> 100%; b: 1/3 are 'x' -> 33.33%
+    assert matrix.loc["a", "b"] == pytest.approx(100.0 - 100.0 / 3)
+
+
+def test_plot_returns_axes(sample_df):
+    from lars.util.label_rate_matrix import plot_label_rate_diff_matrix
+
+    _, ax = plt.subplots()
+    result = plot_label_rate_diff_matrix(sample_df, "stratiform", ax=ax)
+    assert result is ax
+
+
+def test_plot_title_includes_label(sample_df):
+    from lars.util.label_rate_matrix import plot_label_rate_diff_matrix
+
+    _, ax = plt.subplots()
+    plot_label_rate_diff_matrix(sample_df, "stratiform", ax=ax)
+    assert ax.get_title() == "Usage Rate Difference: 'stratiform'"
+
+
+def test_plot_tick_labels_match_columns(sample_df):
+    from lars.util.label_rate_matrix import plot_label_rate_diff_matrix
+
+    _, ax = plt.subplots()
+    plot_label_rate_diff_matrix(sample_df, "stratiform", ax=ax)
+    labels = [t.get_text() for t in ax.get_xticklabels()]
+    assert labels == ["rater_a", "rater_b", "rater_c"]
+
+
+def test_plot_custom_labels(sample_df):
+    from lars.util.label_rate_matrix import plot_label_rate_diff_matrix
+
+    _, ax = plt.subplots()
+    plot_label_rate_diff_matrix(sample_df, "stratiform", ax=ax, labels=["Alice", "Bob", "LLM"])
+    labels = [t.get_text() for t in ax.get_xticklabels()]
+    assert labels == ["Alice", "Bob", "LLM"]
+
+
+def test_plot_uses_gca_when_no_ax(sample_df):
+    from lars.util.label_rate_matrix import plot_label_rate_diff_matrix
+
+    fig, ax = plt.subplots()
+    plt.sca(ax)
+    plot_label_rate_diff_matrix(sample_df, "stratiform")
+    assert ax.get_title() == "Usage Rate Difference: 'stratiform'"
diff --git a/tests/test_labels.py b/tests/test_labels.py
new file mode 100644
index 0000000..888edb7
--- /dev/null
+++ b/tests/test_labels.py
@@ -0,0 +1,168 @@
+import os
+import pytest
+import pandas as pd
+
+
+def _write_csv(path, contents):
+    with open(path, "w") as f:
+        f.write(contents)
+    return str(path)
+
+
+def test_combine_labels_matches_on_file_path(tmp_path):
+    from lars.preprocessing.labels import combine_labels
+
+    human_csv = _write_csv(
+        tmp_path / "human.csv",
+        "file_path,label\n"
+        "/a/img1.png,No Precipitation\n"
+        "/a/img2.png,Stratiform Precipitation\n",
+    )
+    llm_csv = _write_csv(
+        tmp_path / "llm.csv",
+        "file_path,label\n"
+        "/b/img1.png,No Precipitation\n"
+        "/b/img2.png,Isolated Convection\n",
+    )
+
+    combined = combine_labels(
+        [human_csv, llm_csv], ["human_alice", "llm_l4scout"]
+    )
+
+    assert list(combined.columns) == ["file_path", "label", "source"]
+    assert len(combined) == 4
+    assert set(combined["source"]) == {"human_alice", "llm_l4scout"}
+    assert set(combined["file_path"]) == {"img1.png", "img2.png"}
+
+
+def test_combine_labels_matches_on_time(tmp_path):
+    from lars.preprocessing.labels import combine_labels
+
+    human_csv = _write_csv(
+        tmp_path / "human.csv",
+        "time,label\n3/4/25 0:00,No Precipitation\n3/4/25 0:12,Isolated Convection\n",
+    )
+    llm_csv = _write_csv(
+        tmp_path / "llm.csv",
+        "time,label\n3/4/25 0:00,No Precipitation\n3/4/25 0:12,Stratiform Precipitation\n",
+    )
+
+    combined = combine_labels(
+        [human_csv, llm_csv], ["human_bob", "llm_l4scout"], match_on="time"
+    )
+
+    assert list(combined.columns) == ["time", "label", "source"]
+    assert len(combined) == 4
+    assert set(combined["time"]) == {"3/4/25 0:00", "3/4/25 0:12"}
+
+
+def test_combine_labels_custom_label_column(tmp_path):
+    from lars.preprocessing.labels import combine_labels
+
+    csv1 = _write_csv(
+        tmp_path / "a.csv",
+        "file_path,llm_label\n/a/img1.png,No Precipitation\n",
+    )
+    csv2 = _write_csv(
+        tmp_path / "b.csv",
+        "file_path,llm_label\n/b/img1.png,Isolated Convection\n",
+    )
+
+    combined = combine_labels(
+        [csv1, csv2], ["run1", "run2"], label_column="llm_label"
+    )
+
+    assert list(combined.columns) == ["file_path", "llm_label", "source"]
+    assert combined["llm_label"].tolist() == [
+        "No Precipitation", "Isolated Convection"
+    ]
+
+
+def test_combine_labels_mismatched_lengths_raises(tmp_path):
+    from lars.preprocessing.labels import combine_labels
+
+    csv1 = _write_csv(tmp_path / "a.csv", "file_path,label\n/a/img1.png,x\n")
+
+    with pytest.raises(ValueError):
+        combine_labels([csv1], ["only_one", "too_many"])
+
+
+def test_combine_labels_invalid_match_on_raises(tmp_path):
+    from lars.preprocessing.labels import combine_labels
+
+    csv1 = _write_csv(tmp_path / "a.csv", "file_path,label\n/a/img1.png,x\n")
+
+    with pytest.raises(ValueError):
+        combine_labels([csv1], ["source1"], match_on="bogus")
+
+
+def test_standardize_labels_maps_ambiguous_and_unknown_variants():
+    from lars.preprocessing.labels import standardize_labels
+
+    df = pd.DataFrame({
+        "label": ["Ambiguous", "UNKNOWN", "unknown", " Unknown ", "ambiguous"],
+    })
+
+    result = standardize_labels(df)
+
+    assert result["label"].tolist() == ["Ambiguous / Uncertain"] * 5
+
+
+def test_standardize_labels_maps_bare_stratiform():
+    from lars.preprocessing.labels import standardize_labels
+
+    df = pd.DataFrame({"label": ["Stratiform", "stratiform", "STRATIFORM"]})
+
+    result = standardize_labels(df)
+
+    assert result["label"].tolist() == ["Stratiform Precipitation"] * 3
+
+
+def test_standardize_labels_leaves_canonical_and_other_labels_unchanged():
+    from lars.preprocessing.labels import standardize_labels
+
+    df = pd.DataFrame({
+        "label": [
+            "No Precipitation",
+            "Stratiform Precipitation",
+            "Isolated Convection",
+            "Mesoscale Convective System",
+        ],
+    })
+
+    result = standardize_labels(df)
+
+    assert result["label"].tolist() == df["label"].tolist()
+
+
+def test_standardize_labels_preserves_missing_values():
+    from lars.preprocessing.labels import standardize_labels
+
+    df = pd.DataFrame({"label": ["Ambiguous", None]})
+
+    result = standardize_labels(df)
+
+    assert result["label"].iloc[0] == "Ambiguous / Uncertain"
+    assert pd.isna(result["label"].iloc[1])
+
+
+def test_standardize_labels_custom_label_column():
+    from lars.preprocessing.labels import standardize_labels
+
+    df = pd.DataFrame({"llm_label": ["UNKNOWN", "Stratiform"]})
+
+    result = standardize_labels(df, label_column="llm_label")
+
+    assert result["llm_label"].tolist() == [
+        "Ambiguous / Uncertain", "Stratiform Precipitation"
+    ]
+
+
+def test_standardize_labels_does_not_mutate_input():
+    from lars.preprocessing.labels import standardize_labels
+
+    df = pd.DataFrame({"label": ["UNKNOWN"]})
+
+    standardize_labels(df)
+
+    assert df["label"].tolist() == ["UNKNOWN"]