Himalayan Glacier Retreat Rate and Melt Volume Tracking

Track ice boundary retreat and seasonal melt volume across high-altitude glaciers. Klarety processes Sentinel-2 multispectral time-series to delineate glacier extents, compute annual retreat rates, and estimate seasonal meltwater contribution to downstream basins. Water managers use multi-year trend reports for adaptation planning.

Monitor glaciers

Ice boundary and melt tracking for water managers

Klarety AI

Annual glacier retreat reports from satellite data

Klarety agents delineate glacier extents from Sentinel-2 and return retreat rate and meltwater contribution estimates per basin.

NDSI - Sentinel-2

Snow-ice classification and boundary delineation

Agents apply NDSI thresholding and supervised classification to map annual glacier extents and compute boundary retreat vectors.

GIS Output

Glacier extent layers for hydrological GIS tools

Export annual boundary polygons and retreat vectors for hydrological model and water resource GIS platform integration.

NDSI glacier delineation and annual retreat rate computation

Klarety agents apply NDSI thresholding to late-summer Sentinel-2 composites (minimum snow coverage period) to map glacier ice extents annually. Boundary shift vectors are computed between consecutive years to measure retreat rate in metres per year. Volume change is estimated using an area-volume scaling relationship (V = c * A^gamma). Output is an annual extent polygon time-series and retreat rate table per glacier.

Klarety AITrack glacier retreat in your watershed

analysis/himalayan_glacier_retreat.pyAgent code

python

import eeimport numpy as npimport pandas as pd
ee.Initialize()
# Gangotri Glacier, Uttarakhand Indiaglacier_aoi = ee.Geometry.Rectangle([79.0, 30.8, 79.6, 31.1])
def get_glacier_extent_km2(year):    """NDSI-based glacier area for late summer (min snow)."""    s2 = (ee.ImageCollection('COPERNICUS/S2_SR_HARMONIZED')          .filterBounds(glacier_aoi)          .filterDate(f'{year}-09-01', f'{year}-10-15')          .filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE', 30))          .median())
    green = s2.select('B3')    swir1 = s2.select('B11')
    # NDSI = (Green - SWIR1) / (Green + SWIR1)    ndsi = green.subtract(swir1).divide(green.add(swir1)).rename('ndsi')
    # Glacier ice: NDSI > 0.4, exclude shadow (NIR > 500)    b8 = s2.select('B8')    glacier_mask = ndsi.gt(0.4).And(b8.gt(500))
    area = glacier_mask.multiply(ee.Image.pixelArea()).reduceRegion(        ee.Reducer.sum(), glacier_aoi, 10, maxPixels=1e9    ).get('ndsi').getInfo()    return round((area or 0) / 1e6, 2)
years = [2019, 2020, 2021, 2022, 2023, 2024, 2025]areas = {y: get_glacier_extent_km2(y) for y in years}
# Area-volume scaling: V (km3) = 0.0365 * A^1.375 (Bahr et al.)def area_to_volume(area_km2):    return round(0.0365 * area_km2**1.375, 4)
print(f"{'Year':<6} {'Area (km2)':<14} {'Volume (km3)':<16} {'Annual Loss (km2)'}")prev_area = Nonefor y in years:    a = areas[y]    v = area_to_volume(a)    delta = round(a - prev_area, 2) if prev_area else 0.0    print(f"{y:<6} {a:<14.2f} {v:<16.4f} {delta:<.2f}")    prev_area = a

output/glacier_retreat_report.mdOutput report