diff --git a/modules/icing/util.py b/modules/icing/util.py
index b946ff4828e0e34c7aacdadce0859615c7c417dd..90991b9784e8dcedc4139ec2d0e8ffadb49857cd 100644
--- a/modules/icing/util.py
+++ b/modules/icing/util.py
@@ -1,6 +1,7 @@
 import numpy as np
 import deeplearning.icing_fcn as icing_fcn
 import deeplearning.icing_cnn as icing_cnn
+from deeplearning.icing_fcn import IcingIntensityFCN
 from icing.pirep_goes import setup, time_filter_3
 from icing.moon_phase import moon_phase
 from util.util import get_time_tuple_utc, is_day, check_oblique, get_median, homedir, write_icing_file_nc4,\
@@ -12,6 +13,8 @@ from util.setup import model_path_day, model_path_night
 from aeolus.datasource import CLAVRx, CLAVRx_VIIRS, CLAVRx_H08, CLAVRx_H09
 import h5py
 import datetime
+import tensorflow as tf
+import os
 # from scipy.signal import medfilt2d
 
 
@@ -353,6 +356,34 @@ def run_icing_predict_image(clvrx_dir='/Users/tomrink/data/clavrx/RadC/', output
     elif model_type == 'FCN':
         model_module = icing_fcn
 
+    if day_model_path is not None:
+        ckpt_dir_s = os.listdir(day_model_path)
+        ckpt_dir = day_model_path + ckpt_dir_s[0]
+
+        day_model = IcingIntensityFCN(day_night=day_night, l1b_or_l2=l1b_andor_l2, satellite=satellite, use_flight_altitude=use_flight_altitude)
+        day_model.num_data_samples = 10000
+        day_model.build_model()
+        day_model.build_training()
+        day_model.build_evaluation()
+
+        ckpt = tf.train.Checkpoint(step=tf.Variable(1), model=day_model.model)
+        ckpt_manager = tf.train.CheckpointManager(ckpt, ckpt_dir, max_to_keep=3)
+        ckpt.restore(ckpt_manager.latest_checkpoint)
+
+    if night_model_path is not None:
+        ckpt_dir_s = os.listdir(night_model_path)
+        ckpt_dir = night_model_path + ckpt_dir_s[0]
+
+        night_model = IcingIntensityFCN(day_night=day_night, l1b_or_l2=l1b_andor_l2, satellite=satellite, use_flight_altitude=use_flight_altitude)
+        night_model.num_data_samples = 10000
+        night_model.build_model()
+        night_model.build_training()
+        night_model.build_evaluation()
+
+        ckpt = tf.train.Checkpoint(step=tf.Variable(1), model=night_model.model)
+        ckpt_manager = tf.train.CheckpointManager(ckpt, ckpt_dir, max_to_keep=3)
+        ckpt.restore(ckpt_manager.latest_checkpoint)
+
     alt_lo, alt_hi = 0.0, 15000.0
     if use_flight_altitude is True:
         flight_levels = flight_levels
@@ -440,7 +471,7 @@ def run_icing_predict_image(clvrx_dir='/Users/tomrink/data/clavrx/RadC/', output
             for ds_name in day_train_params:
                 day_grd_dct[ds_name] = np.stack(day_data_dct[ds_name])
 
-            preds_day_dct, probs_day_dct = model_module.run_evaluate_static(day_grd_dct, num_day_tiles, day_model_path,
+            preds_day_dct, probs_day_dct = model_module.run_evaluate_static_2(day_model, day_grd_dct, num_day_tiles,
                                                                             day_night='DAY', l1b_or_l2=l1b_andor_l2,
                                                                             prob_thresh=prob_thresh,
                                                                             use_flight_altitude=use_flight_altitude,
@@ -465,7 +496,7 @@ def run_icing_predict_image(clvrx_dir='/Users/tomrink/data/clavrx/RadC/', output
             for ds_name in nght_train_params:
                 nght_grd_dct[ds_name] = np.stack(nght_data_dct[ds_name])
 
-            preds_nght_dct, probs_nght_dct = model_module.run_evaluate_static(nght_grd_dct, num_nght_tiles, night_model_path,
+            preds_nght_dct, probs_nght_dct = model_module.run_evaluate_static_2(night_model, nght_grd_dct, num_nght_tiles,
                                                                               day_night='NIGHT', l1b_or_l2=l1b_andor_l2,
                                                                               prob_thresh=prob_thresh,
                                                                               use_flight_altitude=use_flight_altitude,