Harmonize and interpolate wavelengths for optimization

.github/workflows/main.yml

@@ -0,0 +1,20 @@
+name: CI
+on: [push]
+
+jobs:
+  miniconda:
+    name: Setup and test
+    runs-on: ${{ matrix.os }}
+    strategy:
+        matrix:
+            os: ["ubuntu-latest"]
+    steps:
+      - uses: actions/checkout@v4
+      - name: Python Linter
+        uses: sunnysid3up/python-linter@master
+        with:
+          source: "snowlaps"
+          mypy-options: "--ignore-missing-imports --show-error-codes"
+          pylint-options: "--rcfile=setup.cfg"
+          isort-options: "-w 100"
+          django: true

snowlaps/snowlaps.py

@@ -6,17 +6,17 @@
 
 """
 
+from datetime import datetime
+import joblib
 import numpy as np
 import pandas as pd
-from typing import Union
-import joblib
-import tensorflow as tf
-import time
 import pysolar
 import pytz
-from timezonefinder import TimezoneFinder
-from datetime import datetime
 import sklearn.preprocessing
+import tensorflow as tf
+import time
+from timezonefinder import TimezoneFinder
+from typing import Union
 
 
 class SnowlapsEmulator:
@@ -150,6 +150,7 @@ def optimize(
         nb_optimization_repeats: int = 20,
         optimizer: tf.keras.optimizers = tf.keras.optimizers.Adagrad(learning_rate=1.0),
         optimization_init: Union[list, None] = None,
+        optimization_wavelengths: np.ndarray = np.arange(355, 1365, 10),
         gradient_mask: list = [0.0, 1.0, 1.0, 1.0, 1.0, 1.0],
         sza_list: Union[list, None] = None,
         spectra_metadata_path: Union[str, None] = None,
@@ -185,6 +186,11 @@ def optimize(
                                   Default to None and random initialisations.
         :type optimization_init: list, optional
 
+        :param optimization_wavelengths: An array containing the wavelengths over which to optimize.
+                                         Default to the range used in Chevrollier et al 2024, i.e.
+                                         np.arange(355, 1365, 10).
+        :type optimization_wavelengths: np.ndarray, optional
+
         :param gradient_mask: Boolean mask to enable (1) or freeze (0) the optimization of each
                               input variable of the emulator. Default to SZA frozen and all other
                               variables enabled, as SZA is either given or computed.
@@ -223,7 +229,8 @@ def optimize(
         """
 
         def get_minimum_MAE_index(sub_df: pd.Series) -> int:
-            """Get the global index of the minimum MAE of each batch
+            """
+            Get the global index of the minimum MAE of each batch
 
             :return: the global index.
             :rtype: int
@@ -238,6 +245,14 @@ def get_minimum_MAE_index(sub_df: pd.Series) -> int:
             return global_index
 
         albedo_spectra = self.read_data(albedo_spectra_path)
+        albedo_spectra_wavelengths = albedo_spectra.index.values
+
+        # harmonize emulator, input albedo and optimization wavelengths
+        if set(optimization_wavelengths).issubset(set(albedo_spectra_wavelengths)):
+            albedo_spectra_mask = np.isin(
+                optimization_wavelengths, albedo_spectra_wavelengths
+            )
+            emulator_mask = np.isin(self.emulator_wavelengths, optimization_wavelengths)
 
         if sza_list is None and spectra_metadata_path is not None:
             self.spectra_metadata = self.read_data(spectra_metadata_path)