diff --git a/modules/icing/util.py b/modules/icing/util.py index 3395d0b802eb014c14f62d7bf31ee41d9a246db7..aa3d1f2b714cc045cd65f8001709029fcb9aac3a 100644 --- a/modules/icing/util.py +++ b/modules/icing/util.py @@ -1483,19 +1483,11 @@ def run_icing_predict_image_1x1(clvrx_dir='/Users/tomrink/data/clavrx/RadC/', ou pirep_file='/Users/tomrink/data/pirep/pireps_202109200000_202109232359.csv', obs_lons=None, obs_lats=None, obs_times=None, obs_alt=None, flight_level=None, obs_intensity=None): - # import deeplearning.icing_fcn as icing_fcn - # model_module = icing_fcn - # - # if day_model_path is not None: - # day_model = model_module.load_model(day_model_path, day_night='DAY', l1b_andor_l2=l1b_andor_l2, - # use_flight_altitude=use_flight_altitude) - # if night_model_path is not None: - # night_model = model_module.load_model(night_model_path, day_night='NIGHT', l1b_andor_l2=l1b_andor_l2, - # use_flight_altitude=use_flight_altitude) # load parameter stats and model from disk stdSclr_day = joblib.load('/home/rink/stdSclr_4_day.pkl') day_model = joblib.load('/home/rink/icing_gbm_day.pkl') + stdSclr_nght = joblib.load('/home/rink/stdSclr_4_nght.pkl') nght_model = joblib.load('/home/rink/icing_gbm_nght.pkl') @@ -1513,8 +1505,8 @@ def run_icing_predict_image_1x1(clvrx_dir='/Users/tomrink/data/clavrx/RadC/', ou if flight_level is not None: alt_lo, alt_hi = flt_level_ranges[flight_level] - # day_train_params, _, _ = get_training_parameters(day_night='DAY', l1b_andor_l2=l1b_andor_l2) - # nght_train_params, _, _ = get_training_parameters(day_night='NIGHT', l1b_andor_l2=l1b_andor_l2) + day_train_params = ['cld_temp_acha', 'supercooled_cloud_fraction', 'cld_reff_dcomp', 'cld_opd_dcomp'] + nght_train_params = ['cld_temp_acha', 'supercooled_cloud_fraction', 'cld_reff_acha', 'cld_opd_acha'] # # if day_night == 'AUTO': # train_params = list(set(day_train_params + nght_train_params)) @@ -1522,8 +1514,6 @@ def run_icing_predict_image_1x1(clvrx_dir='/Users/tomrink/data/clavrx/RadC/', ou # train_params = day_train_params # elif day_night == 'NIGHT': # train_params = nght_train_params - day_train_params = ['cld_temp_acha', 'supercooled_cloud_fraction', 'cld_reff_dcomp', 'cld_opd_dcomp'] - nght_train_params = ['cld_temp_acha', 'supercooled_cloud_fraction', 'cld_reff_acha', 'cld_opd_acha'] if satellite == 'H08': clvrx_ds = CLAVRx_H08(clvrx_dir) @@ -1571,9 +1561,9 @@ def run_icing_predict_image_1x1(clvrx_dir='/Users/tomrink/data/clavrx/RadC/', ou num_nght_tiles = np.sum(nght_idxs) print('num nght tiles: ', num_nght_tiles) - cldy_idxs = cldmsk >= 2 - num_cldy_tiles = np.sum(cldy_idxs) - print('num cloudy tiles: ', num_cldy_tiles) + clr_idxs = cldmsk < 2 + num_clr_tiles = np.sum(clr_idxs) + print('num clear tiles: ', num_clr_tiles) fd_preds = np.zeros(num_lines * num_elems, dtype=np.int8) fd_probs = np.zeros(num_lines * num_elems, dtype=np.float32) @@ -1581,18 +1571,17 @@ def run_icing_predict_image_1x1(clvrx_dir='/Users/tomrink/data/clavrx/RadC/', ou fd_probs[:] = -1.0 if (day_night == 'AUTO' or day_night == 'DAY') and num_day_tiles > 0: - varX_std = stdSclr_day.transform(varX_day) varX_std = np.where(np.isnan(varX_std), 0, varX_std) probs = day_model.predict_proba(varX_std)[:, 1] fd_probs[day_idxs] = probs[day_idxs] if (day_night == 'AUTO' or day_night == 'NIGHT') and num_nght_tiles > 0: - varX_std = stdSclr_nght.transform(varX_ngth) varX_std = np.where(np.isnan(varX_std), 0, varX_std) probs = nght_model.predict_proba(varX_std)[:, 1] - fd_probs[nght_idxs] = probs[nght_idxs] + fd_probs[np.invert(day_idxs)] = probs[np.invert(day_idxs)] + fd_probs[clr_idxs] = 0.0 max_prob = fd_probs.reshape((num_lines, num_elems))