SWOT HR elevation comparison to ICESat-2

%matplotlib widget

from pathlib import Path
import io, zipfile, tempfile, os
import warnings
import requests
import fiona
import geopandas as gpd
import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.colors import to_rgba
from matplotlib.patches import Patch
import cartopy.crs as ccrs

# Quiet noisy driver warnings (optional)
warnings.filterwarnings("ignore", message=".*ERROR parsing kml Style: No id.*", category=RuntimeWarning)

# --- Paths ---

# A shared folder on CryoCloud / local machine
DATADIR = Path("/home/jovyan/shared/SWOT/data")  # adjust if needed
DATADIR.mkdir(parents=True, exist_ok=True)

# CryoCloud file targets
GL_SHP = DATADIR / "Antarctica_masks" / "scripps_antarctica_polygons_v1.shp"
HR_KML = DATADIR / "hr_Mar2025_below60S.kml"   # culled HR KML we’ll create if missing
LR_KMZ = DATADIR / "lr_Sept2015.kmz"           # LR orbit

# Source URLs (try HR 2025, fall back to 2024 seasonal if needed)
HR_URLS = [
    "https://podaac.jpl.nasa.gov/SWOT-events/swot_science_hr_Mar2025-v10_perPass.kml",
    "https://podaac.jpl.nasa.gov/SWOT-events/swot_science_hr_Nov2024-v09-seasonal_perPass.kml",
]
LR_URL = "https://www.aviso.altimetry.fr/fileadmin/documents/missions/Swot/swot_science_orbit_sept2015-v2_10s.kmz"

# Scripps polygons zip (contains multiple shapefile components)
GL_ZIP_URL = "https://doi.pangaea.de/10013/epic.42133.d001"

# Bach Ice Shelf bounding box location in South Polar Stereographic [ESPG: 3031]
bach = [-1920000, 480000, -1740000, 680000]

# --- Small helper functions ---

def download_bytes(url: str) -> bytes:
    r = requests.get(url, timeout=180)
    r.raise_for_status()
    return r.content

def ensure_scripps_polygons(gl_shp_path: Path, source_zip_url: str) -> None:
    """Download & extract the Scripps Antarctica polygons if missing."""
    if gl_shp_path.exists():
        return
    gl_shp_path.parent.mkdir(parents=True, exist_ok=True)
    print("Downloading Scripps Antarctica polygons…")
    content = download_bytes(source_zip_url)
    with zipfile.ZipFile(io.BytesIO(content)) as zf:
        zf.extractall(gl_shp_path.parent)
    print(f"Extracted to: {gl_shp_path.parent}")

def read_all_kml_layers(path: Path, driver="KML") -> gpd.GeoDataFrame:
    """Load all non-empty KML/KMZ layers into a single GeoDataFrame."""
    layers = fiona.listlayers(str(path))
    frames = []
    for lyr in layers:
        try:
            gdf = gpd.read_file(path, driver=driver, layer=lyr)
            if not gdf.empty:
                frames.append(gdf)
        except Exception as exc:
            print(f"Skipped layer '{lyr}': {exc}")
    if not frames:
        raise RuntimeError(f"No valid layers in {path}")
    return gpd.GeoDataFrame(pd.concat(frames, ignore_index=True))

def save_kml(gdf: gpd.GeoDataFrame, out_path: Path) -> None:
    """Save a GeoDataFrame as KML (for future reuse)."""
    out_path.parent.mkdir(parents=True, exist_ok=True)
    gdf.to_file(out_path, driver="KML")

# --- Ensure basemap (Scripps polygons) is available ---
ensure_scripps_polygons(GL_SHP, GL_ZIP_URL)

# Load and prep Scripps polygons
gl_gdf = gpd.read_file(GL_SHP)
# Reproject to EPSG:3031 (Antarctic Polar Stereographic)
gl_gdf_3031 = gl_gdf.to_crs(epsg=3031)

# Split out grounded ice vs ice shelves for nicer styling
name_field = "Id_text" if "Id_text" in gl_gdf_3031.columns else None
if name_field is None:
    # Fall back to plotting all polygons uniformly if classification is missing
    grounded_3031 = gl_gdf_3031
    shelves_3031 = gpd.GeoDataFrame(geometry=[])
else:
    grounded_classes = {"Grounded ice or land", "Isolated island", "Ice rise or connected island"}
    grounded_3031 = gl_gdf_3031[gl_gdf_3031[name_field].isin(grounded_classes)]
    shelves_3031  = gl_gdf_3031[gl_gdf_3031[name_field] == "Ice shelf"]

# --- Ensure SWOT HR coverage (culled to south of 60S) exists in CryoCloud ---
# Note: This process takes 23 min if not already completed
if not HR_KML.exists():
    for url in HR_URLS:
        try:
            # Download HR orbital file into temp
            print(f"Trying HR KML: {url}")
            with tempfile.TemporaryDirectory() as tdir:
                kml_path = Path(tdir) / "hr.kml"
                kml_path.write_bytes(download_bytes(url))
            
                hr_all = read_all_kml_layers(kml_path, driver="KML") # with some KMZ/KML may need to use driver="LIBKML"
                hr_all = hr_all.assign(miny=hr_all.geometry.bounds.miny)
                # Cull below 60 S
                hr_south = hr_all[hr_all["miny"] < -60].drop(columns=["miny"])
                if "Name" in hr_south.columns:
                    # Remove nadir coverage to keep only KaRIn's
                    hr_south = hr_south[~hr_south["Name"].str.contains("Nadir", na=False)]
                save_kml(hr_south, HR_KML)
            print(f"Saved culled HR KML: {HR_KML}")
            break
        except Exception as exc:
            print(f"Failed HR download/parse from {url}: {exc}")
    else:
        raise RuntimeError("Could not obtain a usable SWOT HR KML.")

# --- Ensure SWOT LR orbit exists in CryoCloud ---
if not LR_KMZ.exists():
    print("Downloading SWOT LR KMZ…")
    LR_KMZ.write_bytes(download_bytes(LR_URL))
    print(f"Saved: {LR_KMZ}")

# --- Load HR/LR coverages and reproject to EPSG:3031 ---

# HR (already culled), stored as KML
hr_gdf = read_all_kml_layers(HR_KML, driver="KML")
hr_3031 = hr_gdf.to_crs(epsg=3031)

# LR (KMZ) — many GDAL/Fiona builds can read KMZ directly with driver="KML"   
lr_gdf = read_all_kml_layers(LR_KMZ, driver="KML")
if "Name" in lr_gdf.columns:
    lr_gdf = lr_gdf[~lr_gdf["Name"].str.contains("Nadir", na=False)]
lr_3031 = lr_gdf.to_crs(epsg=3031)

len(hr_3031), len(lr_3031)

# --- Plot availability over Antarctica (EPSG:3031) ---
ps71 = ccrs.Stereographic(central_latitude=-90, central_longitude=0, true_scale_latitude=-71)

fig = plt.figure(figsize=(8, 8), dpi=110)
ax = plt.axes(projection=ps71)

# Antarctica basemap: grounded vs shelves
grounded_3031.plot(ax=ax, transform=ps71, color="lightgray", edgecolor="darkgray", linewidth=0.3, zorder=1)
shelves_3031.plot(ax=ax,   transform=ps71, color="gainsboro", edgecolor="silver", linewidth=0.25, zorder=2)

# SWOT LR (red), SWOT HR (blue)
# Build colors with different transparency for facecolor and edgecolor
swot_hr_fill  = to_rgba("blue", alpha=0.20)
swot_hr_edge  = to_rgba("blue", alpha=0.30)
swot_lr_fill  = to_rgba("red", alpha=0.06)
swot_lr_edge  = to_rgba("red", alpha=0.10)

# Note: Many HR/LR geometries are polygons; we use both facecolor and edgecolor lightly
lr_3031.plot(ax=ax, transform=ps71, color=swot_lr_fill, edgecolor=swot_lr_edge, linewidth=0.2, zorder=3)
hr_3031.plot(ax=ax, transform=ps71, color=swot_hr_fill, edgecolor=swot_hr_edge, linewidth=0.2, zorder=4)

bach_rect = plt.Rectangle((bach[0], bach[1]), bach[2] - bach[0], bach[3] - bach[1], 
                          zorder=5, linewidth=2, edgecolor="white", facecolor="none")
ax.add_patch(bach_rect)

# Crop axes to Antarctica
ax.set_xlim(-3000000, 3000000) # 3000 km2
ax.set_ylim(-3000000, 3000000)

legend_handles = [
    Patch(facecolor="gainsboro", edgecolor="silver", label="Ice shelves"),
    Patch(facecolor="lightgray", edgecolor="darkgray", label="Grounded ice / land"),
    Patch(facecolor=swot_lr_fill, edgecolor=swot_lr_edge, label="SWOT LR"),
    Patch(facecolor=swot_hr_fill, edgecolor=swot_hr_edge, label="SWOT HR"),
]
leg = ax.legend(handles=legend_handles, loc="lower left", frameon=True, framealpha=0.9)
for text in leg.get_texts():
    text.set_fontsize(10)

ax.set_title("SWOT Availability over Antarctica (N. Hemisphere Summer)", fontsize=14, pad=10)

plt.tight_layout()
plt.show()

# --- Pip Install a few packages that are not in the default CryoCloud environment: --- 
%pip install -q pyTMD # For downloading and running the Circum-Antarctic Tidal Simulation (CATS) for tide corrections
%pip install -q icesat2_toolkit # For converting ICESat-2 time to GMT
%pip install -q cmap # For perceptually uniform colormaps
%pip install -q PyAstronomy # For date conversions
%pip install -q earthaccess==0.14

# --- Imports ---
%matplotlib widget

from pathlib import Path
from itertools import chain
from concurrent.futures import ProcessPoolExecutor
import datetime as dt
import os, logging, warnings, tempfile, gzip, shutil

# Data and processing
import h5py
import s3fs
import fiona
import h5coro
import rasterio
import requests
import earthaccess
import numpy as np
import pandas as pd
import xarray as xr
import icesat2_toolkit
from io import BytesIO
import geopandas as gpd
import cartopy.crs as ccrs
from PyAstronomy import pyasl
from scipy.interpolate import RegularGridInterpolator

import pyTMD
import timescale  

# Plotting
import shapely
import shapefile
from shapely.plotting import plot_line
from shapely.geometry import LineString

from cmap import Colormap
import cartopy.crs as ccrs
from pyproj import CRS, Transformer

import matplotlib as mpl
import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec
import matplotlib.ticker as ticker
from matplotlib.lines import Line2D
from matplotlib.patches import Patch
from mpl_toolkits.axes_grid1.inset_locator import inset_axes
from mpl_toolkits.axes_grid1.anchored_artists import AnchoredSizeBar

# Quiet noisy logs
warnings.filterwarnings("ignore", message=".*ERROR parsing kml Style: No id.*", category=RuntimeWarning)
logging.getLogger("urllib3.connectionpool").setLevel(logging.ERROR)

# --- Study area --- 
DATADIR = Path("/home/jovyan/shared-public/SWOT/data") # Shared data files that cannot be streamed
GL_SHP   = DATADIR / "Antarctica_masks/scripps_antarctica_polygons_v1.shp" # Grounding line shapefile
BBOX = [-1_970_000, 490_000, -1_820_000, 670_000]  # PS71 [EPSG:3031] bounding box for Bach Ice Shelf
RIFT_BBOX = [-1_885_000, 570_000, -1_855_000, 590_000]  # PS71 [EPSG:3031] bounding box for rift of interest

PS71 = ccrs.Stereographic(central_latitude=-90, central_longitude=0, true_scale_latitude=-71)

tick_label_size = 16
legend_label_size = 18

# --- Earthdata login (interactive once; persists to ~/.netrc) --- 
import earthaccess
auth = earthaccess.login(strategy="interactive", persist=True)
assert auth.authenticated, "Earthdata Login failed."

# Coordinate transforms
def xy2ll(x, y):
    """Convert PS71 (EPSG:3031) coordinates to WGS84 (EPSG:4326)"""
    crs_ll = CRS("EPSG:4326")
    crs_xy = CRS("EPSG:3031")
    xy_to_ll = Transformer.from_crs(crs_xy, crs_ll, always_xy=True)
    lon, lat = xy_to_ll.transform(x, y)
    return lon, lat

def ll2xy(lon, lat):
    """Convert WGS84 (EPSG:4326) coordinates to PS71 (EPSG:3031)"""
    crs_ll = CRS("EPSG:4326")
    crs_xy = CRS("EPSG:3031")
    ll_to_xy = Transformer.from_crs(crs_ll, crs_xy, always_xy=True)
    x, y = ll_to_xy.transform(lon, lat)
    return x, y

def ps712utm(x, y, crs):
    """Convert UTM zone {crs} (SWOT native) to PS71 (EPSG:3031) coordinates"""
    crs_utm = CRS(f"EPSG:{crs}")
    crs_xy = CRS("EPSG:3031")
    ll_to_xy = Transformer.from_crs(crs_xy, crs_utm, always_xy=True)
    utmx, utmy = ll_to_xy.transform(x, y)
    return utmx, utmy

# Colorbar Logic
def add_inset_colorbar(fig, ax, mappable, label,
                       anchor=(0.999, 0.90),  # (x, y) in axes fraction
                       width=3.0, height=0.9,  # inches
                       orientation="horizontal", extend=None, fontsize=tick_label_size):
    # outer box (absolute size in inches, 2-tuple anchor is OK)
    cbbox = inset_axes(ax, width=width, height=height,
                       bbox_to_anchor=anchor, bbox_transform=ax.transAxes, loc="upper right")
    
    for sp in cbbox.spines.values():
        sp.set_visible(False)
    cbbox.set_facecolor([0, 0, 0, 0.7])
    cbbox.set_xticks([]); cbbox.set_yticks([])

    # actual bar inside
    inner = inset_axes(cbbox, "92%", "20%", loc="center")
    cbar = fig.colorbar(mappable, cax=inner, orientation=orientation, extend=extend)
    cbar.outline.set_edgecolor("white"); cbar.outline.set_linewidth(1)
    cbar.ax.tick_params(labelsize=fontsize, color="white", labelcolor="white")
    cbar.set_label(label, fontsize=fontsize+2, color="white")
    if orientation == "horizontal":
        cbar.ax.xaxis.set_label_position("top")
    cbar.ax.minorticks_on(); cbar.ax.tick_params(which="minor", length=4, color="white")
    return cbar

# --- MOA 2009 grayscale basemap (gzipped GeoTIFF) ---

# URL of the protected gzip file
moa_url = "https://daacdata.apps.nsidc.org/pub/DATASETS/nsidc0593_moa2009_v02/geotiff/moa750_2009_hp1_v02.0.tif.gz"

# Use Earthdata login stored in ~/.netrc (standard for requests)
session = requests.Session()
session.auth = requests.utils.get_netrc_auth(moa_url)

# Add user-agent to mimic a real browser
headers = {"User-Agent": "NASA Earthdata Python client"}

# Download and check for success
response = session.get(moa_url, headers=headers)
response.raise_for_status()

# Confirm content-type is not HTML (debug print)
if b"<html" in response.content[:100].lower():
    raise RuntimeError("Download returned HTML (login failed or wrong credentials?)")

# Decompress the GZip in memory
with gzip.GzipFile(fileobj=BytesIO(response.content)) as gz:
    tif_bytes = BytesIO(gz.read())

# Read with Rasterio from memory
with rasterio.MemoryFile(tif_bytes) as memfile:
    with memfile.open() as moa:
        bounds = moa.bounds
        left, bottom, right, top = bounds.left, bounds.bottom, bounds.right, bounds.top
        moa_dat = moa.read(1)

ext = (left, right, bottom, top)

# --- SWOT HR granule (search via granule name) ---
search_days = 9

swot_results = earthaccess.search_data(
    short_name="SWOT_L2_HR_Raster_D",
    temporal=("2025-05-23", "2025-05-23"),
    granule_name="SWOT_L2_HR_Raster_100m_UTM18D_N_x_x_x_033_143_012F_20250523T232533_20250523T232551_PID0_01.nc",
)
swot_file = earthaccess.open(swot_results)[0]
ds_swot = xr.open_dataset(swot_file, engine="h5netcdf")

# Build LAT/LON polygon for ATL06 search
poly_xy = np.array([[RIFT_BBOX[0], RIFT_BBOX[1]],[RIFT_BBOX[2], RIFT_BBOX[1]],
                    [RIFT_BBOX[2], RIFT_BBOX[3]],[RIFT_BBOX[0], RIFT_BBOX[3]],[RIFT_BBOX[0], RIFT_BBOX[1]]])
poly_lon, poly_lat = xy2ll(poly_xy[:,0], poly_xy[:,1])
polygon_ll = np.c_[poly_lon, poly_lat].tolist()

# Time window: SWOT start/end ± 9 days
t0 = dt.datetime.strptime(ds_swot.time_granule_start, "%Y-%m-%dT%H:%M:%S.%fZ")
t1 = dt.datetime.strptime(ds_swot.time_granule_end,   "%Y-%m-%dT%H:%M:%S.%fZ")
atl06_window = (t0 - dt.timedelta(days=search_days), t1 + dt.timedelta(days=search_days))


atl06_results = earthaccess.search_data(short_name="ATL06",
                                        temporal=atl06_window,
                                        polygon=polygon_ll)
print(f'{len(atl06_results)} files found in time window in RIFT_BBOX')

# --- Read a subset of ATL06 beams and variables - 30 s ---
beam_roots = ["gt1l","gt1r","gt2l","gt2r","gt3l","gt3r"]

s3_creds = auth.get_s3_credentials(daac="NSIDC")
fs_nsidc = s3fs.S3FileSystem(
    key=s3_creds["accessKeyId"],
    secret=s3_creds["secretAccessKey"],
    token=s3_creds["sessionToken"],
)

def read_atl06_points(result):
    """Return a list of dicts (one per beam) with coords, heights, corrections."""
    out = []
    s3_url = result.data_links(access="direct")[0].replace("s3://","")
    for beam in beam_roots:
        try:
            with fs_nsidc.open(s3_url, mode="rb", cache_type="blockcache", block_size=4*1024*1024) as s3obj:
                ds_main = xr.open_dataset(s3obj, engine="h5netcdf",
                                          group=f"{beam}/land_ice_segments",
                                          driver_kwds={"rdcc_nbytes": 1024*1024},
                                          decode_times=False, phony_dims="sort")
                ds_dem  = xr.open_dataset(s3obj, engine="h5netcdf",
                                          group=f"{beam}/land_ice_segments/dem",
                                          driver_kwds={"rdcc_nbytes": 1024*1024},
                                          decode_times=False, phony_dims="sort")
                ds_geo  = xr.open_dataset(s3obj, engine="h5netcdf",
                                          group=f"{beam}/land_ice_segments/geophysical",
                                          driver_kwds={"rdcc_nbytes": 1024*1024},
                                          decode_times=False, phony_dims="sort")
                data = {
                    "lon": ds_main["longitude"].values,
                    "lat": ds_main["latitude"].values,
                    "h_li": ds_main["h_li"].values.astype("float32"),
                    "t_dt": ds_main["delta_time"].values,
                    "q":    ds_main["atl06_quality_summary"].values,
                    "geoid_h": ds_dem["geoid_h"].values.astype("float32"),
                    "geoid_free2mean": ds_dem["geoid_free2mean"].values.astype("float32"),
                    "tide_ocean": ds_geo["tide_ocean"].values.astype("float32"),
                    "dac": ds_geo["dac"].values.astype("float32"),
                    "tide_earth": ds_geo["tide_earth"].values.astype("float32"),
                    "tide_earth_free2mean": ds_geo["tide_earth_free2mean"].values.astype("float32"),
                    "tide_pole": ds_geo["tide_pole"].values.astype("float32"),
                    "tide_load": ds_geo["tide_load"].values.astype("float32"),
                    "beam": beam, "name": s3_url,
                }
                # Mask poor quality data
                bad = (data["q"] == 1) | (data["h_li"] > 3.0e38)
                data["h_li"][bad] = np.nan
                # PS71 coords
                data["x"], data["y"] = ll2xy(data["lon"], data["lat"])
                out.append(data)
        except Exception as exc:
            # Skip missing beams
            continue
    return out

atl06 = []
for r in atl06_results:
    atl06.extend(read_atl06_points(r))

# Convert delta_time -> decimal years using icesat2_toolkit
for d in atl06:
    dec = icesat2_toolkit.convert_delta_time(d["t_dt"])
    d["time"] = dec["decimal"]

print(f'{len(atl06)} beams loaded')

# --- Choose a slice of ATL06 candidates to preview (adjust as needed) ---
is2_start, is2_end = 0, 18  # tweak these indices

subset = atl06[is2_start:is2_end]
print(f"Previewing ATL06 candidates [{is2_start}:{is2_end}] out of total {len(atl06)}")
for d in subset:
    print(f"{d['beam']:>4}  {d['name'].split('/')[-1]}")

# --- Figure: MOA (gray) + SWOT swath (viridis) + ATL06 points ---
fig = plt.figure(figsize=(8, 8))
ax = plt.axes(projection=PS71)

# MOA basemap
ax.imshow(moa_dat, extent=ext, cmap="gray", vmin=15000, vmax=17000)

# SWOT swath (water surface elevation)
wse = ds_swot["wse"]
utm = ccrs.UTM(zone=ds_swot.utm_zone_num, southern_hemisphere=True)
X, Y = np.meshgrid(wse["x"].values, wse["y"].values)
mesh = ax.pcolormesh(
    X, Y, wse.values,
    transform=utm, cmap="viridis", shading="auto",
    vmin=0, vmax=100, zorder=2
)

# ATL06 candidate tracks (colored by index within the slice)
cmap = plt.get_cmap("plasma")
n = max(1, len(subset))
for i, d in enumerate(subset):
    color = cmap(i / (n - 1) if n > 1 else 0.5)
    ax.scatter(d["x"], d["y"], s=2, color=color, transform=PS71, zorder=3)
# Convert cmap into ScalarMappable for colorbar
norm = plt.Normalize(vmin=0, vmax=n - 1)
sm = plt.cm.ScalarMappable(cmap=cmap, norm=norm)
sm.set_array([]) 

bach_rect = plt.Rectangle((RIFT_BBOX[0], RIFT_BBOX[1]), RIFT_BBOX[2] - RIFT_BBOX[0], RIFT_BBOX[3] - RIFT_BBOX[1], 
                          zorder=5, linewidth=2, edgecolor="white", facecolor="none")
ax.add_patch(bach_rect)

# View & labels
ax.set_xlim(BBOX[0], BBOX[2])
ax.set_ylim(BBOX[1], BBOX[3])
ax.set_title("ATL06 candidate tracks over SWOT swath (Bach Ice Shelf)", fontsize=14)

# Colorbar for SWOT Swath
add_inset_colorbar(fig, ax, mesh, "SWOT WSE [m]", anchor=(0.999, 0.87), extend="max")
# ICESat-2 tracks colorbar
add_inset_colorbar(fig, ax, sm, "ICESat-2 Index", anchor=(0.999, 0.999))

plt.tight_layout()
plt.show()

# --- Find granule matching ATL06_20241213201922_13522510_006_01.h5 --- 
gran = 'ATL06_20250515130244_09102710_007_01.h5'
beam = 'gt2r'
is2_tracks_to_use = []
for i, data in enumerate(atl06):
    if gran in data['name'] and beam in data['beam']:
        is2_tracks_to_use.append(atl06[i])

assert is2_tracks_to_use, 'No tracks found' # Check at least one track chosen

# --- Stream MeASUREs ice velocity data via earthaccess (NSIDC-0484), 450 m grid ---
v_results = earthaccess.search_data(short_name="NSIDC-0484", 
                                    cloud_hosted=True,
                                    temporal=("1996-01-01","1996-01-02"), 
                                    count=1)

v_file = earthaccess.open(v_results)[0]
vel = xr.open_dataset(v_file)

# Crop a margin around our box; compute magnitude for plotting
off = 10_000
v_crop = vel.sel(x=slice(BBOX[0]-off, BBOX[2]+off),
                 y=slice(BBOX[3]+off, BBOX[1]-off))
vel_mag = np.hypot(v_crop.VX, v_crop.VY)  # m/yr
oslo = Colormap("crameri:oslo").to_matplotlib() # Velocity colormap

# --- Advect ICESat-2 to location at time of SWOT data collection ---

# Create regular grid interpolators to speed up running the advection code (~10x speed up over using native grid)
def prep_velocity_interpolators(vel):
    vx_interp = RegularGridInterpolator((vel.y.values[::-1], vel.x.values),
                        vel["VX"].values[::-1, :],bounds_error=False,fill_value=np.nan)
    vy_interp = RegularGridInterpolator((vel.y.values[::-1], vel.x.values),
                        vel["VY"].values[::-1, :],bounds_error=False,fill_value=np.nan)
    return vx_interp, vy_interp

for track in is2_tracks_to_use:
    # Copy arrays so we don't mutate the originals
    start_x = track["x"].copy()
    start_y = track["y"].copy()
    start_time = track["time"].copy()      # decimal years (one per point)

    # Spatial mask to the study bounding box (BBOX: [xmin, ymin, xmax, ymax])
    mask = (
        (start_x > BBOX[0]) &
        (start_x < BBOX[2]) &
        (start_y > BBOX[1]) &
        (start_y < BBOX[3])
    )
    # Keep full array length; mark points outside the box as NaN in X
    # (Y is left unchanged to preserve indexing; NaNs in X will propagate during updates)
    start_x[~mask] = np.nan

    # Convert decimal year → Python datetime for each sample
    start_time = [pyasl.decimalYearGregorianDate(t) for t in start_time]

    # Integration setup
    target_time = t0 - dt.timedelta(days=0)   # SWOT target time (t0); minus 0 keeps semantics
    step_size = 1 / 365                       # years per step (1 day)

    # For each point, compute time difference to target, then number of steps
    steps = [(target_time - t) for t in start_time]  # list of timedeltas (can be ±)
    steps = [int( td.total_seconds() / (365 * 24 * 60 * 60) / step_size ) for td in steps]
    
    # Use the sign of the FIRST point's step count to set direction (forward or backward in time)
    step_size = np.sign(steps[0]) * step_size
    # Integrate a UNIFORM number of steps for all points (assumes similar times)
    steps = np.abs(steps)

    # Velocity interpolators on the cropped grid
    # v_crop has coordinates (y, x). Arrays are stored row-major with y decreasing,
    # so the helper flips y when building RegularGridInterpolator.
    vx_interp, vy_interp = prep_velocity_interpolators(v_crop)

    # Euler integration (meters per year * years per step = meters per step)
    x = start_x.copy()
    y = start_y.copy()
    for _ in range(steps[0]):                 # same number of steps for every point
        # Sample velocity at current positions; inputs are (y, x)
        vx = vx_interp((y, x))                # meters / year
        vy = vy_interp((y, x))
        # Update positions; NaNs remain NaN and will propagate
        x = x + vx * step_size
        y = y + vy * step_size

    # Store advected coordinates back on the track
    track["x_advected"] = x
    track["y_advected"] = y

# --- Ensure CATS2008-v2023 is available locally for PyTMD ---
fs = s3fs.S3FileSystem(anon=True)
pyTMD_path = pyTMD.utilities.get_data_path("data")
model = pyTMD.io.model(pyTMD_path, verify=False).elevation("CATS2008-v2023")
model.grid_file.parent.mkdir(parents=True, exist_ok=True)

if not model.grid_file.exists():
    with fs.open("pytmd/CATS2008_v2023/CATS2008_v2023.nc", "rb") as fin, \
         open(model.grid_file, "wb") as fout:
        shutil.copyfileobj(fin, fout)
model.grid_file.exists()

# Tides
def compute_tides(model, lons, lats, datetimes):
    """
    Computes the tides at given locations and times using PyTMD.

    Parameters
    model: PyTMD.model
         Tide model
    lons: list[float]
         Geodetic longitude in EPSG:4326
    lats: list[float]
         Geodetic latitude in EPSG:4326
    datetimes: list[datetime.datetime]
         Datetime to calculate tides

    Returns
    tides - list[float]
         Tide elevations in cm at locations and times sepcified
    """
    years  = np.array([d.year for d in datetimes])
    months = np.array([d.month for d in datetimes])
    days   = np.array([d.day for d in datetimes])
    hours  = np.array([d.hour for d in datetimes])
    minutes= np.array([d.minute for d in datetimes])

    tide_time = timescale.time.convert_calendar_dates(years, months, days, hours, minutes)

    # Setup model
    constituents = pyTMD.io.OTIS.read_constants(model.grid_file, model.model_file,
                                         model.projection, type=model.type,
                                         grid=model.format, apply_flexure=False)
    c = constituents.fields
    amp, ph, _ = pyTMD.io.OTIS.interpolate_constants(np.atleast_1d(lons),
                                                     np.atleast_1d(lats), constituents,
                                                     type=model.type, method="spline",
                                                     extrapolate=True)
    cph = -1j * ph * np.pi / 180.0
    hc = amp * np.exp(cph)

    # Calculate tides and infer minor constitutents at each location and time
    tides = []
    for i in range(len(datetimes)):
        TIDE = pyTMD.predict.map(tide_time[i], hc, c, deltat=0, corrections=model.format)
        MINOR = pyTMD.predict.infer_minor(tide_time[i], hc, c, deltat=0, corrections=model.format)
        t_cm = (TIDE.data + MINOR.data) * 100.0
        tides.append(t_cm.astype("float32"))
    return np.array(tides) 

# --- Find tide corrections using CATS at times of interest - 1m 29s --- 

# Swot and ICESat-2 times
tide_times = [
    t0,
    pyasl.decimalYearGregorianDate(is2_tracks_to_use[0]["time"][0]),
]
sats = ["SWOT", "IS2"]
data = is2_tracks_to_use[0]
data["lon_advected"], data["lat_advected"] = xy2ll(data["x_advected"], data["y_advected"])

for tide_time, satellite in zip(tide_times, sats):
    tide_results = np.array(
        compute_tides(model, data["lon_advected"], data["lat_advected"], [tide_time])).T
    data[f"tide_{satellite}"] = tide_results

data["tide_IS2"] = data["tide_IS2"].astype("float32").squeeze()
data["tide_SWOT"] = data["tide_SWOT"].astype("float32").squeeze()

# --- Apply geophysical corrections to ICESat-2 heights for fair comparison to SWOT --- 
data["h_li_corrected"] = (
    data["h_li"]
    - data["geoid_h"]
    + data["geoid_free2mean"]
    - data["tide_earth"]
    - data["tide_pole"]
    - data["tide_load"]
)

# --- Interpolate SWOT swath along icesat-2 track --- 

# Create arrays to hold interpolated values
overall_wse = np.zeros_like(data["x_advected"])
overall_wse[:] = np.nan
overall_wse_sigma = np.zeros_like(data["x_advected"])
overall_wse_sigma[:] = np.nan

wse = ds_swot["wse"]
wse_sigma = ds_swot["wse_uncert"]

# Convert icesat2 to utm, and interpolate SWOT advected data
x_utm = wse["x"].values
y_utm = wse["y"].values
advected_x_utm, advected_y_utm = ps712utm(data["x_advected"], data["y_advected"], utm.to_epsg())

advected_x_utm = xr.DataArray(advected_x_utm, dims=["points"])
advected_y_utm = xr.DataArray(advected_y_utm, dims=["points"])

x_utm = wse["x"].values
y_utm = wse["y"].values

interpolated_wse = wse.interp(
    x=advected_x_utm,
    y=advected_y_utm,
    method="linear",
    kwargs={"fill_value": np.nan},
).compute()
interpolated_wse_sigma = wse_sigma.interp(
    x=advected_x_utm,
    y=advected_y_utm,
    method="linear",
    kwargs={"fill_value": np.nan},
).compute()

#Put all non-nan values into overall_wse
# If there are no SWOT data values where we have IS2 data, add a nan 
mask = np.isnan(overall_wse) & ~np.isnan(interpolated_wse)
# Only update at those places
overall_wse[mask] = interpolated_wse[mask]
overall_wse_sigma[mask] = interpolated_wse_sigma[mask]

# Compute +- 2 std
lower = overall_wse - 2 * overall_wse_sigma
upper = overall_wse + 2 * overall_wse_sigma

# Cropping data to the rift plot area
min_mask = -1872000
max_mask = -1865000
x_mask = (data["x_advected"] > min_mask) & (
    data["x_advected"] < max_mask
)

# --- Apply tide corrections to both datasets and uncertainty bounds ---
data["is2_tidecorrected"] = data["h_li_corrected"] - data["tide_IS2"] / 100
data["swot_tidecorrected"] = overall_wse - data["tide_SWOT"] / 100
data["swot_lower"] = lower - data["tide_SWOT"] / 100
data["swot_upper"] = upper - data["tide_SWOT"] / 100

# Plot comparison
label = "Elevation [m]"
vmin = 0
vmax = 50

fig = plt.figure(figsize=(8, 12))

utm = ccrs.UTM(zone=ds_swot.utm_zone_num, southern_hemisphere=True)
ps71_projection = ccrs.Stereographic(central_latitude=-90, central_longitude=0, true_scale_latitude=-71)
gs = gridspec.GridSpec(2, 1, height_ratios=[4, 1])
ax = fig.add_subplot(gs[0], projection=ps71_projection)

# Map View
ax.imshow(moa_dat, extent=ext, cmap="gray", vmin=15000, vmax=17000)
cb = ax.imshow(vel_mag, cmap=oslo, 
               extent=[vel_mag["x"].min(), vel_mag["x"].max(), vel_mag["y"].min(), vel_mag["y"].max()],
               origin="upper", alpha=0.6, vmin=0, vmax=550)

grounded_3031.plot(ax=ax, transform=ps71, color="none", edgecolor="white", linewidth=1, zorder=4)

wse = ds_swot['wse']
x = wse["x"].values
y = wse["y"].values

X, Y = np.meshgrid(x, y)
mesh = ax.pcolormesh(X, Y, wse.values, transform=utm, cmap="viridis", shading="auto", vmin=vmin, vmax=vmax)

ax.scatter(data["x_advected"], data["y_advected"], color="white", s=5, transform=ps71_projection)

x_plotting = data["x_advected"][x_mask]
y_plotting = data["y_advected"][x_mask]
ax.scatter(x_plotting, y_plotting, color="orangered", s=5, transform=ps71_projection)

# Cross section
ax2 = fig.add_subplot(gs[1])

ax2.plot(data["x_advected"], data["is2_tidecorrected"], 
         color="orangered", linewidth=2, label="ICESat-2")
ax2.plot(data["x_advected"], data["swot_tidecorrected"],
         color="navy", linewidth=2, label="SWOT")
ax2.fill_between(data["x_advected"], data["swot_lower"], 
         data["swot_upper"], color="navy", alpha=0.3)

# SWOT Swath colorbar
add_inset_colorbar(fig, ax, mesh, label="SWOT WSE [m]", anchor=(0.999, 0.87), extend="max")

# Ice Velocity colorbar
add_inset_colorbar(fig, ax, cb, label="Ice Velocity [m/a]", anchor=(0.999, 0.999))

# Prettify panel 1
ax.set_xlim(BBOX[0] + 1000, BBOX[2] - 1000)
ax.set_ylim(BBOX[1] + 1000, BBOX[3] - 1000)

# Scalebar
scalebar = AnchoredSizeBar(ax.transData, 20000, "20 km", "lower right", pad=0.1, 
                           sep=1, color="white", frameon=False, size_vertical=2000,
                           fontproperties=mpl.font_manager.FontProperties(size=20, weight="bold"), label_top=False)

# Antarctica Inset
inset = fig.add_axes([0.15, 0.73, 0.3, 0.3], projection=ps71_projection)  # [left, bottom, width, height]
inset.patch.set_facecolor("none")
for spine in inset.spines.values():
    spine.set_visible(False)
inset.set_xticks([])
inset.set_yticks([])

grounded_3031.plot(ax=inset, transform=ps71, color="lightgray", edgecolor="lightgray", linewidth=0.3, zorder=1)
shelves_3031.plot(ax=inset,   transform=ps71, color="white", edgecolor="white", linewidth=0.3, zorder=2)

rect = plt.Rectangle((BBOX[0], BBOX[1]), BBOX[2] - BBOX[0], BBOX[3] - BBOX[1],
                     zorder=3, linewidth=2, edgecolor="red", facecolor="none")
inset.add_patch(rect)
ax.add_artist(scalebar)
scalebar.set_bbox_to_anchor((0.99, 0.67), transform=ax.transAxes)

# Time Labels
ax.text(0.01, 0.01, "SWOT: 2025-05-23\nICESat-2: 2025-05-15", transform=ax.transAxes, fontsize=legend_label_size, color="white")

# Polish panel 2
ax2.set_xlabel("PS71 X [km]", fontsize=legend_label_size)
ax2.set_ylabel("Elevation [m]", fontsize=legend_label_size)
KM_SCALE = 1e3
ticks_x = ticker.FuncFormatter(lambda x, pos: "{0:g}".format(x / KM_SCALE))
ax2.xaxis.set_major_formatter(ticks_x)
ax2.tick_params(axis="both", which="major", labelsize=tick_label_size, size=10)
ax2.spines["top"].set_visible(False)
ax2.spines["right"].set_visible(False)
ax2.set_xlim(np.nanmin(data["x_advected"]), np.nanmax(data["x_advected"]))
ax2.set_xlim(min_mask, max_mask)  
ax2.set_ylim(-2, 32)

# Legend
handles = [
    Line2D([0], [0], color="orangered", linewidth=2, label="ICESat-2"),
    Line2D([0], [0], color="navy", linewidth=2, label="SWOT"),
    Patch(facecolor="navy", alpha=0.3, label="SWOT Uncertainty"),
]

ax2.legend(fontsize=legend_label_size, loc="lower left", handles=handles[:1] + [handles[1]])
leg = ax2.get_legend()
for line in leg.get_lines():
    line.set_linewidth(3)
leg.get_frame().set_edgecolor("white")
leg.get_frame().set_linewidth(1)

fig.tight_layout()