ctao-calibpipe 0.3.0rc2__py3-none-any.whl

calibpipe/tools/telescope_cross_calibration_calculator.py

@@ -0,0 +1,721 @@
+"""Calculate the relative throughput of the telescopes."""
+
+from collections import defaultdict
+from itertools import combinations  # noqa: D100
+
+import astropy.units as u
+import numpy as np
+from astropy.table import Table, join
+from ctapipe.core import (
+    TelescopeComponent,
+    Tool,
+)
+from ctapipe.core.traits import (
+    AstroQuantity,
+    Bool,
+    Dict,
+    Float,
+    FloatTelescopeParameter,
+    IntTelescopeParameter,
+    Path,
+    Unicode,
+)
+from ctapipe.instrument import SubarrayDescription
+from ctapipe.io import read_table, write_table
+from iminuit import Minuit
+from scipy.stats import norm
+from tables.exceptions import NoSuchNodeError
+
+
+class RelativeThroughputFitter(TelescopeComponent):
+    """Perform relative throughput fitting for telescopes.
+
+    This component is used to fit the relative throughput of telescopes based on the
+    energy asymmetry of telescope pairs.
+    """
+
+    reference_telescopes = IntTelescopeParameter(
+        default_value=None,
+        help="ID of the telescopes whose throughput kept fixed during the intercalibration minimization",
+        allow_none=True,
+    ).tag(config=True)
+    throughput_normalization = FloatTelescopeParameter(
+        default_value=1.0,
+        help="Setting the overall telescope throughput normalization. "
+        "Depending the use case, it could reflect the "
+        "absolute optical throughput measured by the muon rings / illuminator, "
+        "if we want to compare or complement these methods. "
+        "Alternatively it could get an arbitrary number, e.g. one that sets the average throughput to 1, "
+        "if the user wants to 'flat-field' the array. "
+        "Finally it could be set to 1, if we want to identify outlier telescopes or study the aging.",
+        allow_none=False,
+    ).tag(config=True)
+
+    def fit(self, subarray_name, subarray_data):
+        """
+        Perform minimization for each telescope subsystem to compute optical throughput factors.
+
+        Parameters
+        ----------
+        subarray_data : dict
+            A dictionary where keys are telescope pairs and values are dictionaries containing
+            "mean_asymmetry": the average energy asymmetry and "mean_uncertainty": the average
+            uncertainty of the asymmetry for each telescope pair.
+        subarray_name : str
+            The name of the intercalibration subarray, used to identify the telescopes.
+            Usually corresponds to the telescope type (e.g., "LST", "MST", "SST").
+
+        Returns
+        -------
+        dict
+            Dictionary of the form:
+            {
+                'subarray_name': {tel_id: (value, error), ...}
+            }
+        """
+        results = {}
+
+        tel_ids = self.subarray.tel_ids
+
+        initial_values = {
+            str(tel_id): self.throughput_normalization.tel[tel_id] for tel_id in tel_ids
+        }
+
+        def _chi2(*values):
+            params = dict(zip(tel_ids, values))
+            chi2 = 0.0
+            for (i, j), entry in subarray_data.items():
+                a_ij = entry["mean_asymmetry"]
+                s_ij = entry["mean_uncertainty"]
+                ci = params[i]
+                cj = params[j]
+                model = (ci - cj) / (ci + cj)
+                chi2 += ((a_ij - model) ** 2) / (s_ij**2)
+            return chi2
+
+        fit = Minuit(_chi2, name=initial_values.keys(), **initial_values)
+
+        # Create a list of unique reference telescope IDs
+        unique_reference_telescopes = set(
+            self.reference_telescopes.tel[tel_id]
+            for tel_id in tel_ids
+            if self.reference_telescopes.tel[tel_id] is not None
+        )
+
+        # Pass each unique reference telescope ID to fit.fixed
+        for ref_tel_id in unique_reference_telescopes:
+            fit.fixed[str(ref_tel_id)] = True
+
+        fit.errordef = 1
+        fit.migrad()
+
+        results[subarray_name] = {
+            tel_id: (fit.values[tel_id], fit.errors[tel_id])
+            for tel_id in initial_values.keys()
+        }
+        return results
+
+
+class PairFinder(TelescopeComponent):
+    """Find pairs of telescopes based on their types and distances between them."""
+
+    max_impact_distance = FloatTelescopeParameter(
+        default_value=[
+            ("type", "LST*", 125.0),
+            ("type", "MST*", 125.0),
+            ("type", "SST*", 225.0),
+        ],
+        help="Maximum distance between the telescopes and a shower core in meters. "
+        "The maximum distance between the telescopes in pair "
+        "should not exceed the sum of these values for the telescopes in pair.",
+        allow_none=False,
+    ).tag(config=True)
+
+    def find_pairs(self, by_tel_type=True, cross_type_only=False):
+        """
+        Find pairs of telescope IDs.
+
+        Parameters
+        ----------
+        by_tel_type : bool
+            If True, find pairs of telescopes of the same type.
+            If False, find pairs of all telescopes in the array.
+        cross_type_only : bool
+            If True, find pairs of telescopes of different types only.
+
+        Returns
+        -------
+        dict
+            A dictionary where keys are telescope types (str) and values are sets of
+            pairs of telescope IDs (int, int). If by_tel_type is False, the key will be "ALL".
+            If cross_type_only is True, the key will be "XTEL".
+        """
+        result = defaultdict(set)
+        telescope_positions = self.subarray.positions
+
+        def check_distance(tel1, tel2):
+            """Check if the pair of telescopes is within the maximum distance."""
+            pos1 = telescope_positions[tel1]
+            pos2 = telescope_positions[tel2]
+            distance = np.linalg.norm((pos1 - pos2).to(u.m).value)
+            return (
+                distance
+                <= self.max_impact_distance.tel[tel1]
+                + self.max_impact_distance.tel[tel2]
+            )
+
+        def check_type(tel1, tel2):
+            """Check if the pair of telescopes are of the same type."""
+            return self.subarray.tel[tel1].type == self.subarray.tel[tel2].type
+
+        telescope_types = defaultdict(list)
+        if by_tel_type:
+            for tel_id, desc in self.subarray.tel.items():
+                telescope_types[desc.type].append(tel_id)
+        else:
+            if cross_type_only:
+                telescope_types["XTEL"] = list(self.subarray.tel_ids)
+            else:
+                telescope_types["ALL"] = list(self.subarray.tel_ids)
+
+        for tel_type, tel_ids in telescope_types.items():
+            for tel1, tel2 in combinations(tel_ids, 2):
+                if cross_type_only and check_type(tel1, tel2):
+                    continue
+                if check_distance(tel1, tel2):
+                    result[tel_type].add((tel1, tel2))
+
+        return result
+
+
+class CalculateCrossCalibration(Tool):
+    """Calibrator that performs cross calibration of telescopes."""
+
+    input_url = Path(
+        default_value="dl2.h5",
+        help="Path to the input file with the DL2 (merged) data.",
+    ).tag(config=True)
+    output_url = Path(
+        default_value="monitoring_cross_calibration_dl2.h5",
+        help="Path to the output file where the produced calibration"
+        "products will be stored",
+    ).tag(config=True)
+    overwrite = Bool(default_value=False, help="Overwrite output file.").tag(
+        config=True
+    )
+    reconstruction_algorithm = Unicode(
+        default_value="RandomForest",
+        help="Name of the reconstruction algorithm",
+        allow_none=False,
+    ).tag(config=True)
+    event_filters = Dict(
+        per_key_traits={
+            "min_gammaness": Float(allow_none=True),
+            "min_energy": AstroQuantity(
+                physical_type=u.physical.energy, allow_none=True
+            ),
+            "max_distance_asymmetry": Float(allow_none=True),
+        },
+        default_value={
+            "min_gammaness": None,
+            "min_energy": None,
+            "max_distance_asymmetry": None,
+        },
+        help=(
+            "Dictionary of event filters:\n"
+            "  - min_gammaness is used to select the gamma-like events\n"
+            "  - min_energy defines the energy range suitable for different types of telescopes\n"
+            "  - max_distance_asymmetry sets the equidistance asymmetry limit\n"
+            "If a filter parameter is set to None (the default value), "
+            "it will not be applied during event selection."
+        ),
+    ).tag(config=True)
+
+    aliases = {
+        ("i", "input_url"): "CalculateCrossCalibration.input_url",
+        ("o", "output_url"): "CalculateCrossCalibration.output_url",
+    }
+
+    classes = [RelativeThroughputFitter, PairFinder]
+
+    def merge_tables(self, telescope_pairs):
+        """
+        Merge telescope energy tables for specified telescope pairs.
+
+        Parameters
+        ----------
+        telescope_pairs : dict
+            A dictionary where keys are telescope types (str), and values are lists of
+            tuples. Each tuple contains two telescope IDs (int) representing a pair.
+
+        Returns
+        -------
+        dict
+            A dictionary where keys are tuples of telescope IDs (int, int),
+            representing the telescope pairs. The values are lists of merged tables
+            for the specified telescope pairs. If no tables are successfully merged
+            for a telescope pair, the list will be empty.
+        """
+        merged_tables = {}
+        unique_telescope_ids = {
+            tel_id
+            for pairs in telescope_pairs.values()
+            for pair in pairs
+            for tel_id in pair
+        }
+
+        telescope_tables = {}
+        for tel_id in unique_telescope_ids:
+            try:
+                path = f"dl2/event/telescope/energy/{self.reconstruction_algorithm}Regressor/tel_{tel_id:03d}"
+                telescope_tables[tel_id] = read_table(
+                    self.input_url,
+                    path,
+                    condition=f"{self.reconstruction_algorithm}Regressor_tel_is_valid",
+                )
+            except NoSuchNodeError:
+                pass  # data for one telescope is missing, will be reported in application to a particular pair later.
+        for telescope_type, pairs in telescope_pairs.items():
+            if telescope_type not in merged_tables:
+                merged_tables[telescope_type] = {}
+            for tel1, tel2 in pairs:
+                table_1 = telescope_tables.get(tel1)
+                table_2 = telescope_tables.get(tel2)
+                if table_1 is None or table_2 is None:
+                    self.log.warning(
+                        "Missing telescope data for tel %s in pair (%s, %s)",
+                        tel1 if table_1 is None else tel2,
+                        tel1,
+                        tel2,
+                    )
+                    continue
+
+                try:
+                    merged_table = join(
+                        table_1[
+                            [
+                                "obs_id",
+                                "event_id",
+                                f"{self.reconstruction_algorithm}Regressor_tel_energy",
+                                f"{self.reconstruction_algorithm}Regressor_tel_energy_uncert",
+                            ]
+                        ],
+                        table_2[
+                            [
+                                "obs_id",
+                                "event_id",
+                                f"{self.reconstruction_algorithm}Regressor_tel_energy",
+                                f"{self.reconstruction_algorithm}Regressor_tel_energy_uncert",
+                            ]
+                        ],
+                        keys=["obs_id", "event_id"],
+                    )
+
+                    if len(merged_table) > 0:
+                        pair = (tel1, tel2)
+                        if pair not in merged_tables[telescope_type]:
+                            merged_tables[telescope_type][pair] = []
+
+                        merged_tables[telescope_type][pair].append(merged_table)
+                except MemoryError as e:
+                    self.log.error(
+                        "MemoryError: The data is too large to process: %s", e
+                    )
+                    raise
+
+        return merged_tables
+
+    def _apply_min_gammaness(self, tel_id1, tel_id2, threshold):
+        """Select gamma-like events based on the gammaness threshold."""
+        events_tel1 = self._get_gamma_like_events(tel_id1, threshold)
+        events_tel2 = self._get_gamma_like_events(tel_id2, threshold)
+        return events_tel1 & events_tel2
+
+    def _apply_min_energy(self, tel_id1, tel_id2, threshold):
+        """Select showers over an energy threshold."""
+        events_tel1 = self._set_energy_threshold(tel_id1, threshold)
+        events_tel2 = self._set_energy_threshold(tel_id2, threshold)
+        return events_tel1 & events_tel2
+
+    def _apply_max_distance_asymmetry(self, tel_id1, tel_id2, threshold):
+        """Select showers with similar impact distances."""
+
+        def load_distance_table(tel_id):
+            path = f"dl2/event/telescope/impact/HillasReconstructor/tel_{tel_id:03d}"
+            return read_table(self.input_url, path)[
+                ["obs_id", "event_id", "HillasReconstructor_tel_impact_distance"]
+            ]
+
+        table1 = load_distance_table(tel_id1)
+        table2 = load_distance_table(tel_id2)
+
+        merged = join(table1, table2, keys=["obs_id", "event_id"])
+        dist1 = merged["HillasReconstructor_tel_impact_distance_1"]
+        dist2 = merged["HillasReconstructor_tel_impact_distance_2"]
+
+        asym = ((dist2 - dist1) / (dist1 + dist2)).value
+        mask = np.abs(asym) < threshold
+        return set(zip(merged["obs_id"][mask], merged["event_id"][mask]))
+
+    def _get_gamma_like_events(self, tel_id, threshold):
+        """Select showers that are gamma like."""
+        path = f"dl2/event/telescope/classification/{self.reconstruction_algorithm}Classifier/tel_{tel_id:03d}"
+        class_table = read_table(
+            self.input_url,
+            path,
+            condition=f"{self.reconstruction_algorithm}Classifier_tel_is_valid",
+        )
+        mask = (
+            class_table[f"{self.reconstruction_algorithm}Classifier_tel_prediction"]
+            > threshold
+        )
+        return set(zip(class_table["obs_id"][mask], class_table["event_id"][mask]))
+
+    def _set_energy_threshold(self, tel_id, threshold):
+        """Select showers over an energy threshold."""
+        path = f"dl2/event/telescope/energy/{self.reconstruction_algorithm}Regressor/tel_{tel_id:03d}"
+        energy_table = read_table(
+            self.input_url,
+            path,
+            condition=f"{self.reconstruction_algorithm}Regressor_tel_is_valid",
+        )
+        mask = (
+            energy_table[f"{self.reconstruction_algorithm}Regressor_tel_energy"]
+            > threshold
+        )
+        return set(zip(energy_table["obs_id"][mask], energy_table["event_id"][mask]))
+
+    def event_selection(self, merged_table, tel1, tel2):
+        """
+        Filter merged table based on energy, classification, and optionally equidistant impact criteria.
+
+        Parameters
+        ----------
+        merged_table : Table
+            Merged table from two telescopes.
+        tel1, tel2 : int
+            Telescope IDs.
+
+        Returns
+        -------
+        Table
+            Filtered table.
+        """
+        selected_events = self._get_valid_events(tel1, tel2)
+
+        event_ids = set(zip(merged_table["obs_id"], merged_table["event_id"]))
+        valid_mask = np.array(
+            [(obs_id, event_id) in selected_events for obs_id, event_id in event_ids]
+        )
+        return merged_table[valid_mask]
+
+    def _get_valid_events(self, tel_id1, tel_id2):
+        """Apply all event filters for a given telescope."""
+        event_sets = []
+        for filter_name, threshold in self.event_filters.items():
+            if threshold is None:
+                self.log.debug("Skipping filter %s", filter_name)
+                continue
+            try:
+                filter_func = getattr(self, f"_apply_{filter_name}")
+            except AttributeError:
+                self.log.error(
+                    "Filter function _apply_%s not found for filter %s",
+                    filter_name,
+                    filter_name,
+                )
+                raise ValueError(
+                    f"Filter {filter_name} is not implemented or not recognized."
+                )
+            event_sets.append(filter_func(tel_id1, tel_id2, threshold))
+
+        return set.intersection(*event_sets) if event_sets else set()
+
+    def calculate_energy_asymmetry(self, merged_tables):
+        """
+        Calculate the mean energy asymmetry and its uncertainty for each telescope pair from the merged tables.
+
+        Parameters
+        ----------
+        merged_tables : dict
+            A dictionary where keys are telescope types. Each value is another dictionary where keys are tuples
+            of telescope IDs (int, int), representing telescope pairs, and the values are lists of merged tables
+            containing energy data.
+
+        Returns
+        -------
+        dict
+            A nested dictionary with telescope types as the first level of keys.
+            Each value is a dictionary with telescope pairs as keys and a dictionary as value containing:
+            - "mean_asymmetry": the average energy asymmetry
+            - "mean_uncertainty": the average uncertainty of the asymmetry
+        """
+        energy_asymmetry_results = {}
+
+        for telescope_type, pairs in merged_tables.items():
+            energy_asymmetry_results[telescope_type] = {}
+            for pair, merged_tables_list in pairs.items():
+                asymmetry_values = []
+                uncertainty_values = []
+                for merged_table_unfiltered in merged_tables_list:
+                    merged_table = self.event_selection(
+                        merged_table_unfiltered,
+                        pair[0],
+                        pair[1],
+                    )
+                    energy_tel1 = merged_table[
+                        f"{self.reconstruction_algorithm}Regressor_tel_energy_1"
+                    ]
+                    energy_tel2 = merged_table[
+                        f"{self.reconstruction_algorithm}Regressor_tel_energy_2"
+                    ]
+                    energy_uncertainty_tel1 = merged_table[
+                        f"{self.reconstruction_algorithm}Regressor_tel_energy_uncert_1"
+                    ]
+                    energy_uncertainty_tel2 = merged_table[
+                        f"{self.reconstruction_algorithm}Regressor_tel_energy_uncert_2"
+                    ]
+
+                    energy_sum = energy_tel1 + energy_tel2
+                    energy_asymmetry = ((energy_tel1 - energy_tel2) / energy_sum).value
+                    mean_asymmetry = energy_asymmetry.mean()
+
+                    energy_asymmetry_uncertainty = (2.0 / energy_sum**2) * np.sqrt(
+                        (energy_tel1 * energy_uncertainty_tel2) ** 2
+                        + (energy_tel2 * energy_uncertainty_tel1) ** 2
+                    )
+                    asymmetry_uncertainty = energy_asymmetry_uncertainty.mean()
+                    mean_asymmetry_uncertainty = asymmetry_uncertainty / len(
+                        energy_asymmetry
+                    )
+
+                    asymmetry_values.append(mean_asymmetry)
+                    uncertainty_values.append(mean_asymmetry_uncertainty)
+
+                if asymmetry_values:
+                    energy_asymmetry_results[telescope_type][pair] = {
+                        "mean_asymmetry": sum(asymmetry_values) / len(asymmetry_values),
+                        "mean_uncertainty": sum(uncertainty_values)
+                        / len(uncertainty_values),
+                    }
+
+        return energy_asymmetry_results
+
+    def _load_telescope_energy_tables(self, pairs):
+        """Load energy tables for unique telescopes in the given pairs."""
+        telescope_tables = {}
+        unique_tel_ids = {tel for pair in pairs for tel in pair}
+        for tel_id in unique_tel_ids:
+            try:
+                path = f"dl2/event/telescope/energy/{self.reconstruction_algorithm}Regressor/tel_{tel_id:03d}"
+                telescope_tables[tel_id] = read_table(self.input_url, path)
+            except NoSuchNodeError:
+                continue
+        return telescope_tables
+
+    def _merge_valid_energy_tables(self, tel1, table1, tel2, table2):
+        """Merge valid energy tables for two telescopes based on common events."""
+        valid1 = table1[
+            table1[f"{self.reconstruction_algorithm}Regressor_tel_is_valid"]
+        ]
+        valid2 = table2[
+            table2[f"{self.reconstruction_algorithm}Regressor_tel_is_valid"]
+        ]
+
+        merged = join(
+            valid1[
+                [
+                    "obs_id",
+                    "event_id",
+                    f"{self.reconstruction_algorithm}Regressor_tel_energy",
+                ]
+            ],
+            valid2[
+                [
+                    "obs_id",
+                    "event_id",
+                    f"{self.reconstruction_algorithm}Regressor_tel_energy",
+                ]
+            ],
+            keys=["obs_id", "event_id"],
+            table_names=["1", "2"],
+            uniq_col_name="{col_name}_{table_name}",
+        )
+        return merged
+
+    def compute_cross_type_energy_ratios(self, cross_type_pairs, fit_results):
+        """
+        Compute energy ratios grouped by telescope type pairs using throughput-corrected energies.
+
+        Parameters
+        ----------
+        telescope_pairs : dict
+            A dictionary where keys are telescope types and values are lists of
+            tuples. Each tuple contains two telescope IDs (int) representing a pair.
+        fit_results : dict
+            Dictionary from `fit()` with format {subarray_name: {tel_id: (throughput, error)}}
+
+        Returns
+        -------
+        dict
+            Dictionary of format {(type1, type2): [corrected energy ratios]}
+        """
+        energy_ratios_by_type = {}
+        mean_ratios = {}
+        telescope_tables = self._load_telescope_energy_tables(cross_type_pairs)
+
+        # Build lookup tables
+        throughput_map = {
+            tel_id: value[0]
+            for subarray in fit_results.values()
+            for tel_id, value in subarray.items()
+        }
+        type_map = {
+            tel_id: self.subarray.tel[tel_id].type for tel_id in self.subarray.tel_ids
+        }
+
+        self.relative_throughput_fitter.subarray = self.subarray
+        for tel1, tel2 in cross_type_pairs:
+            table1 = telescope_tables.get(tel1)
+            table2 = telescope_tables.get(tel2)
+            if table1 is None or table2 is None:
+                continue
+            merged = self._merge_valid_energy_tables(tel1, table1, tel2, table2)
+            if (
+                len(merged) > 0
+                and str(tel1) in throughput_map
+                and str(tel2) in throughput_map
+            ):
+                throughput1 = throughput_map[str(tel1)]
+                throughput2 = throughput_map[str(tel2)]
+                normalisation1 = (
+                    self.relative_throughput_fitter.throughput_normalization.tel[tel1]
+                )
+                normalisation2 = (
+                    self.relative_throughput_fitter.throughput_normalization.tel[tel2]
+                )
+
+                corrected_energy1 = merged[
+                    f"{self.reconstruction_algorithm}Regressor_tel_energy_1"
+                ] * (throughput1 / normalisation1)
+                corrected_energy2 = merged[
+                    f"{self.reconstruction_algorithm}Regressor_tel_energy_2"
+                ] * (throughput2 / normalisation2)
+                ratios = corrected_energy1 / corrected_energy2
+
+                type1 = type_map[tel1]
+                type2 = type_map[tel2]
+                type_pair = tuple(sorted((type1, type2)))
+
+                if type_pair not in energy_ratios_by_type:
+                    energy_ratios_by_type[type_pair] = []
+
+                energy_ratios_by_type[type_pair].extend(ratios.tolist())
+
+        for type_pair, ratios in energy_ratios_by_type.items():
+            if ratios:
+                mu, std = norm.fit(ratios)
+                err = std / np.sqrt(len(ratios))
+                mean_ratios[type_pair] = (mu, err)
+
+        return mean_ratios
+
+    def save_monitoring_data(self, intercalibration_results, cross_calibration_results):
+        """
+        Save the calibration products in DL2 monitoring data.
+
+        Parameters
+        ----------
+        intercalibration_results : dict
+            Dictionary of the form:
+            {
+                'subarray_name': {tel_id: (value, error), ...}
+            }
+            that stores the relative throughput coefficients for
+            each telescope.
+        cross_calibration_results : dict
+            Dictionary of format {(type1, type2): [corrected energy ratios]}
+        """
+        tel_rows = []
+        for size_type, tel_data in intercalibration_results.items():
+            size_type_str = str(size_type.value)
+            for tel_id, (value, error) in tel_data.items():
+                tel_rows.append((size_type_str, int(tel_id), value, error))
+
+        tel_table = Table(
+            rows=tel_rows, names=("size_type", "tel_id", "value", "error")
+        )
+
+        ratio_rows = []
+        for (type_a, type_b), (mean, error) in cross_calibration_results.items():
+            ratio_rows.append((str(type_a.value), str(type_b.value), mean, error))
+
+            ratio_table = Table(
+                rows=ratio_rows, names=("type_a", "type_b", "ratio", "error")
+            )
+
+        write_table(
+            tel_table,
+            self.output_url,
+            "/dl2/monitoring/inter_calibration",
+            overwrite=self.overwrite,
+        )
+        write_table(
+            ratio_table,
+            self.output_url,
+            "/dl2/monitoring/cross_calibration",
+            overwrite=self.overwrite,
+        )
+
+    def setup(self):
+        """Set up the logic."""
+        # Read subarray description
+        self.subarray = SubarrayDescription.read(self.input_url)
+        self.relative_throughput_fitter = RelativeThroughputFitter(
+            subarray=self.subarray,
+            parent=self,
+        )
+        self.pair_finder = PairFinder(
+            subarray=self.subarray,
+            parent=self,
+        )
+
+    def start(self):
+        """Perform cross-calibration per telescope subsystem."""
+        # Intercalibration
+        tel_pairs = self.pair_finder.find_pairs()
+        merged_tables = self.merge_tables(tel_pairs)
+        energy_asymmetry_results = self.calculate_energy_asymmetry(merged_tables)
+        intercalibration_results = {}
+
+        for subarray_name, subarray_data in energy_asymmetry_results.items():
+            measured_telescopes = set(
+                tel_id for (i, j), entry in subarray_data.items() for tel_id in (i, j)
+            )
+            self.relative_throughput_fitter.subarray = self.subarray.select_subarray(
+                tel_ids=measured_telescopes, name=subarray_name
+            )
+            intercalibration_results.update(
+                self.relative_throughput_fitter.fit(subarray_name, subarray_data)
+            )
+            self.log.debug("intercalibration minimization", intercalibration_results)
+
+        # Cross-calibration
+        cross_type_pairs = self.pair_finder.find_pairs(
+            by_tel_type=False, cross_type_only=True
+        )
+        cross_calibration_results = self.compute_cross_type_energy_ratios(
+            cross_type_pairs["XTEL"], intercalibration_results
+        )
+        # Save calibration products to dl2/monitoring
+        self.save_monitoring_data(intercalibration_results, cross_calibration_results)
+
+    def finish(self):
+        """Store the results."""
+        self.log.info("Shutting down.")
+
+
+def main():
+    """Run the app."""
+    tool = CalculateCrossCalibration()
+    tool.run()