ROMS-AI-Hybrid-model-for-float-trajectory

#####################################################################

Latitude and longitude will be trained separately.

Train latitude AI model first.

Using "float_trajectory_lat_ai_model_train.py"

You need to set "data_folder" and "case_fn" based on your directory. (The cases periods are in "case_period.txt") The input data for training are under "AI-input-data". (including lat, lon, time, wind in x component, wind in y component, current in x component, current in y component, and water depth.)

After you train latitude AI model, you will get AI model parameters and saved as "float_model_lat.h5" and "float_model_weights_lat.h5".

Then, the test results (will be used for analysis) of latitudes will be calculated and saved as "test_ai_results_lat_new.mat". You can find the AI test results in "AI-Model-results". #################################################################### After you run latitude AI model. you will also get "data_start_info.mat" and "training_data_use_new.mat", which include the float information and input data for training.

We use "training_data_use_new.mat" to train longitude AI model, which means you must run latitude AI model first.

Train longitude AI model secondly.

Using "float_trajectory_lon_ai_model_train.py"

After you train longitude AI model, you will get AI model parameters and saved as "float_model_lon.h5" and "float_model_weights_lon.h5".

The test results of longitude will be calculated and saved as "test_ai_results_lon_new.mat" after you run this python script. ###################################################################### If you want to load AI model without training to get the test results, you can use the following python scripts: "float_trajectory_lat_ai_model_load_model.py" and "float_trajectory_lon_ai_model_load_model.py"

You need to set "aimodel_path" and "aimodel_weight_path" based on your directory. #######################################################################

Post processing

###################################################################

For Figure 1

Using "test_figure1a.m" and "test_figure1b.m" This script will call M_Map mapping package. You can download M_Map from https://www.eoas.ubc.ca/~rich/map.html A global coastline data will be used and called as "global_coast.mat"

You need to set "root_dir" based on your directory. The input data are from directory under "AI-Model". ###################################################################

For Figure 4

Using "figure4.m"

You need to set "root_dir" based on your directory. The input data are under "/AI-Model". ###################################################################

For Fogure 5

Using "test_u_figure5a.m" and "test_v_figure5b.m"

You need to test "data_folder" based on your directory. The input data are under "Post-processing/data_rmse_hf_roms" ###################################################################

For Figure 6

Using "test_ai_results_pdf_12_figure6a.m" and "test_ai_results_pdf_24_figure6b.m"

You need to set "data_folder" based on your directory. The input data are under "AI-Model-results".

The total mean separation error of AI and ROMS models will be printed on the screen after you run these scripts. ###################################################################

For Figure 7

Using "test_time_series_overall_figure7.m"

You need to set "data_folder" based on your directory. The input data are under "AI-Model-results". ###################################################################

For Figure 8

Using "figure8.m"

You need to set "root_dir" based on your directory. The input data are under "AI-Model" ###################################################################

For Figure 9

Using "test_ai_results_pdf_12_figure9a.m" and "test_ai_results_pdf_24_figure9b.m"

You need to set "data_folder" based on your directory. The input data are under "AI-Model-results".

The AI and ROMS models' performances are based on the following variables: "bias_model_good, bias_model_bad, bias_ai_good, and bias_ai_bad" ###################################################################

For Figure 10

Using "test_distance_with_error_bad_figure10a.m" and "test_distance_with_error_goof_figure10b.m"

You need to set "root_dir" based on your directory. The input data are under "AI-Model". ###################################################################

For Figure 11

Using "test_mon_with_error_figure11a.m" and "test_west_east_figure11b.m"

You need to set "root_dir" based on your directory. The input data are under "AI-Model". ###################################################################

For Figure 12

Using "test_ai_results_figure12.m" This script will call M_Map mapping package as Figure 1.

You need to set "data_folder" and "out_dir" based on your directory. The input data are under "AI-Model". Meanwhile, you need to set the "roms_root" based on your directory. These data are under "/figure12" ###################################################################

For calculating relative RMS error (REE)

Using "test_hf_roms_ree.m"

You need to set "data_folder" based on your directory. The input data are under "/Post-processing/data_rmse_hf_roms"

The results will be printed on the screen after you run this script. ###################################################################

The performances under hurricanes

You can use "test_ai_results_pdf_hurricanes.m" to get the performances under hurricanes. ###################################################################

Figures 2 and 3 are drawn by Mircosoft PowerPoint.