Contact Grant Macdonald grantmacdonald@uvic.ca for questions

/Files/

FYI_UMD_gridDATA.csv - Contains all processed first-year ice data used in the analysis (sections 3 and 4), including HH and HV backscatter, incidence angle (IA), height (Field_4_me) and roughness (field_4_st). The following beam was excluded from this analysis and used as an independent beam in the section 5 model analysis: UMDRDA_ATL03_20220311103519_12091403_005_01_gt1l_Elevation_McClintockChannel_JoinedAtts

MYI_NASA_UMD_gridDATA.csv - Contains all processed multi-year ice data used in the analysis (sections 3 and 4).

BeaufortMaryland_ZonalStats.csv - Contains processed backscatter and ICESat-2 topography data for the Beaufort Sea site in section 5.
	
ZonalStats_Baffin_17032022.csv -  Contains processed backscatter and ICESat-2 topography data for the Baffin Bay site in section 5.

	
McClintockIndyData_20220324094442.csv - Contains processed backscatter and ICESat-2 topography data for one of the independent set of CAA tracks in section 5.
	
McClintockIndyData_20220324222936.csv - Contains processed backscatter and ICESat-2 topography data for one of the independent set of CAA tracks in section 5.

	
HV_HH_model_NewIA.pkl - model file. Predicts roughness in metres from HV and HH backscatter in dB. Co



## Example of model applied in Python

import numpy as np
import pandas as pd
import joblib
import matplotlib.pyplot as plt

#Paths to data and model
data_path = '/path_to_data/FYI_UMD_gridDATA.csv' #(Data with HV and HH backscatter and incidence angle in degrees)
model_path = 'path_to_model/HV_HH_model_NewIA.pkl' 

full_data = pd.read_csv(data_path)

# Filter data to subset for this example
layer_filter = (full_data['layer'] == 'UMDRDA_ATL03_20220311103519_12091403_005_01_gt1l_Elevation_McClintockChannel_JoinedAtts')
filename_filter = (full_data['Filename'] == 'Maryland26Feb.csv')
Zero_filter= (full_data['field_4_st'] != 0)


new_data = full_data.loc[layer_filter & filename_filter&Zero_filter].copy()

# Convert columns to numeric and drop NaN values
for column in ['HV_mean', 'HH_mean', 'IA_mean']:
    new_data.loc[:, column] = pd.to_numeric(new_data[column], errors='coerce')

new_data.dropna(subset=['HV_mean', 'HH_mean', 'IA_mean'], inplace=True)

# 'Slopes' based on relationship between incidence angle and backscatter in training data analysis
hv_slope = -1.7390836140798094e-05
hh_slope = -0.0010037221810645228



new_data['HV_mean_adj'] = new_data['HV_mean'] - (hv_slope * new_data['IA_mean'])
new_data['HH_mean_adj'] = new_data['HH_mean'] - (hh_slope * new_data['IA_mean'])
new_data['HV_mean_adj_dB'] = 10 * np.log10(new_data['HV_mean_adj'])
new_data['HH_mean_adj_dB'] = 10 * np.log10(new_data['HH_mean_adj'])

# Load model
adjusted_model = joblib.load(model_path)

# Apply model to HV and HH backscatter in dB
new_data['predicted_field_4_st_adj'] = adjusted_model.predict(new_data[['HV_mean_adj_dB', 'HH_mean_adj_dB']])

# Convert 'field_4_st' to numeric and drop NaN values
new_data['field_4_st'] = pd.to_numeric(new_data['field_4_st'], errors='coerce')
new_data.dropna(subset=['field_4_st'], inplace=True)