diff --git a/modules/deeplearning/icing_fcn.py b/modules/deeplearning/icing_fcn.py
index ccceb7deb99fde685f5edc775ef1f7371b1482bb..0da9aa42ec492ccc0a74871485fd613e7183bfe0 100644
--- a/modules/deeplearning/icing_fcn.py
+++ b/modules/deeplearning/icing_fcn.py
@@ -543,8 +543,22 @@ class IcingIntensityFCN:
             h5f = self.h5f_l1b_tst
         else:
             h5f = self.h5f_l2_tst
+
         time = h5f['time']
+        flt_alt = h5f['flight_altitude'][:]
         tst_idxs = np.arange(time.shape[0])
+
+        if self.flight_level == 0:
+            tst_idxs = tst_idxs[np.logical_and(flt_alt >= 0, flt_alt < 2000)]
+        elif self.flight_level == 1:
+            tst_idxs = tst_idxs[np.logical_and(flt_alt >= 2000, flt_alt < 4000)]
+        elif self.flight_level == 2:
+            tst_idxs = tst_idxs[np.logical_and(flt_alt >= 4000, flt_alt < 6000)]
+        elif self.flight_level == 3:
+            tst_idxs = tst_idxs[np.logical_and(flt_alt >= 6000, flt_alt < 8000)]
+        elif self.flight_level == 4:
+            tst_idxs = tst_idxs[np.logical_and(flt_alt >= 8000, flt_alt < 15000)]
+
         self.num_data_samples = len(tst_idxs)
 
         self.get_test_dataset(tst_idxs)
@@ -1081,7 +1095,8 @@ class IcingIntensityFCN:
         self.do_evaluate(ckpt_dir)
 
 
-def run_restore_static(filename_l1b, filename_l2, ckpt_dir_s_path, day_night='DAY', l1b_or_l2='both', use_flight_altitude=False):
+def run_restore_static(filename_l1b, filename_l2, ckpt_dir_s_path, day_night='DAY', l1b_or_l2='both',
+                       use_flight_altitude=False, flight_level=0):
     ckpt_dir_s = os.listdir(ckpt_dir_s_path)
     cm_s = []
     prob_s = []
@@ -1092,6 +1107,7 @@ def run_restore_static(filename_l1b, filename_l2, ckpt_dir_s_path, day_night='DA
         if not os.path.isdir(ckpt_dir):
             continue
         nn = IcingIntensityFCN(day_night=day_night, l1b_or_l2=l1b_or_l2, use_flight_altitude=use_flight_altitude)
+        nn.flight_level = flight_level
         nn.run_restore(filename_l1b, filename_l2, ckpt_dir)
         cm_s.append(tf.math.confusion_matrix(nn.test_labels.flatten(), nn.test_preds.flatten()))
         prob_s.append(nn.test_probs.flatten())