Skip to content
VISTA — VIIRS Surface Water Temporal Atlas

VISTA Water Atlas User Guide

VISTA provides global flood and surface-water products derived from NOAA JPSS/VIIRS flood mapping data at approximately 375 m resolution. The products are intended for GIS and remote sensing users who need either time-specific archive observations or multi-year statistics describing typical water and flood dynamics.

VISTA products are used operationally by FAO, and they are also useful for inspection, analysis, and method review.

Access

FAO STAC catalogue

The freq_flood_any baseline is also published through FAO STAC as 36 dekadal and 36 rolling monthly layers. Use the STAC access guide to choose a period, open a remote raster in QGIS or download a global GeoTIFF without an account. Other bands and dated observations remain available through the routes below.

Use the VISTA bucket browser to browse and download published GeoTIFF products:

The bucket browser provides GeoTIFF downloads for the dekadal and monthly archives, plus GeoTIFF downloads for dekadal and monthly statistical products: count, frequency, and dominant class.

VISTA Explorer

For interactive viewing, open the public VISTA Explorer. It runs in the browser and does not require an Earth Engine account. It provides two modes: Explore reference and Compare a dekad. A dekad covers days 1–10, 11–20 or 21 to month end. Rolling monthly references combine the previous, selected and following dekads; they are not calendar-month summaries.

In Explore reference, choose a calendar dekad and display historical flood frequency, observation counts or dominant class. Switch between dekadal and rolling monthly references, or compare them using the swipe divider. Zero values are hidden by default and can be displayed using the checkbox.

Compare a dekad starts on the latest available date and follows three steps: Observe, Reference and Compare. Choose an available year and dekad to inspect the observed classes. Then choose the reference window and flood definition: low, medium and high flood classes, or medium and high classes only. In Compare, adjust the reference-frequency threshold, initially 25%. Observed flooded pixels are blue where reference frequency is at or above the threshold, orange where it is below the threshold, and grey where valid reference data are insufficient.

In either mode, click a pixel to inspect flood frequency for both flood definitions, the selected reference window and definition, flood and valid-observation counts, per-class counts, and an annual strip showing the class for that calendar dekad in each archive year. Compare mode also reports the observed class and seasonal or exceptional classification where reference support is available.

These classifications describe historical frequency under the selected threshold. They do not measure flood severity or establish a seasonal cause. The reference may include the event year and is not a leave-one-year-out comparison. Cloud indicates obscured observations, not dry conditions. Flood frequency uses valid observations as its denominator.

Earth Engine assets

Use Google Earth Engine to inspect or load the asset collections:

The Earth Engine links open the folder asset page in the Code Editor. Depending on the Code Editor view, you may see only the folder ID and not a complete list of contained images. To use an image, copy the folder ID and append the image name in an ee.Image(...) call, or browse product files through the VISTA bucket browser first and then use the matching Earth Engine asset name.

Choosing A Product

VISTA can be used in two different ways.

Archive products:

  • Use them when you need the observed flood or water situation at a specific moment in time.
  • Dekadal archive products provide 10-day aggregated observations from the third dekad of January 2012 to the third dekad of April 2026.
  • Monthly archive products provide rolling monthly aggregated observations for the same type of time-specific analysis.
  • Quality-control exclusion: 202407d1 is omitted because the provider’s observation-derived classes are displaced. Monthly observations centered on 202406d3, 202407d1, and 202407d2 use their two remaining valid dekads; the exclusion is recorded in their metadata.
  • They answer questions such as: “What was the maximum flood extent and intensity in this pixel during this dekad or monthly window?”

Statistical products:

  • Use them when you need the typical flood or water behavior for a recurring part of the year.
  • They group all available years for the same recurring dekad or monthly center-dekad.
  • For example, a statistic for the first dekad of January combines the first 10 days of January across the archive years.
  • They answer questions such as: “How often is this pixel flooded in the first dekad of January?” or “What is the dominant water, land, or flood state for this approximately 375 m pixel in a typical monthly window?”

Time Naming Convention

A dekadal statistic uses a key in MMdD format: MM is the month and D is the dekad number. For example:

  • 01d1: first dekad of January, days 1-10
  • 01d2: second dekad of January, days 11-20
  • 10d2: second dekad of October, days 11-20

Monthly archive and monthly statistical product names also use a dekad key, but the key identifies the center dekad of a rolling monthly window. When available, the monthly window combines the previous dekad, the center dekad, and the following dekad. For example, vista_count_monthly_01d2 is centered on January dekad 2 and summarizes the window formed by 01d1, 01d2, and 01d3.

Classes for Monthly and Dekadal Archive Rasters

The following single-band categorical values are used in the dekadal archive and monthly archive.

Value Meaning
0 No data, fill, bad data, shadow, or unclassified
1 Flood low
2 Flood medium
3 Flood high
4 Snow cover
5 River or lake ice cover
6 Supra-snow water, mixed ice-water, or melting ice
9 Cloud
10 Land
11 Permanent water

Flood classes are derived from NOAA VIIRS water-fraction classes. The source water fraction represents the percentage of the pixel covered by water:

  • 1: flood low, source water fraction 1-49%
  • 2: flood medium, source water fraction 50-79%
  • 3: flood high, source water fraction 80-100%

VISTA uses the following priority rule to convert NOAA source observations into dekadal and monthly archive data. The rule is applied pixel by pixel when combining observations inside a dekadal or rolling monthly time window:

flood high (3)
> flood medium (2)
> flood low (1)
> supra-snow / mixed / melting ice (6)
> river or lake ice cover (5)
> snow cover (4)
> land (10)
> permanent water (11)
> cloud (9)
> no data / unclassified (0)

This preserves maximum flood extent and intensity.

Archive Product: Dekadal

The dekadal archive contains one categorical raster per year-month-dekad, from vista_dekad_201201d3 (third dekad of January 2012) to vista_dekad_202604d3 (third dekad of April 2026). It is used to inspect the observed maximum flood extent and intensity during a specific 10-day window.

Example:

vista_dekad_202510d2

This is the maximum flood extent and intensity observed during the second dekad of October 2025, from 11 to 20 October 2025.

Earth Engine root:

projects/fao-oer/assets/vista_dekadal_archive

Bucket example:

  • Browser: open VISTA bucket browser, then browse to vista_dekadal_archive/
  • Direct example: https://storage.googleapis.com/vista_project/vista_dekadal_archive/vista_dekad_202510d2.tif

Content:

  • single uint8 band; select it positionally with .select(0) rather than by name
  • values use the archive class values 0, 1, 2, 3, 4, 5, 6, 9, 10, 11

Earth Engine script:

var img = ee.Image('projects/fao-oer/assets/vista_dekadal_archive/vista_dekad_202510d2');

var vistaArchiveVis = {
  min: 0,
  max: 11,
  palette: [
    '000000', // 0 unclassified / nodata
    '7fcdbb', // 1 flood low
    '41b6c4', // 2 flood medium
    '225ea8', // 3 flood high
    'ffffff', // 4 snow cover
    'b3e5fc', // 5 river/lake ice cover
    '00ffff', // 6 supra-snow/ice water, mixed ice-water, or ice in melting status
    'bdbdbd', // 7 unused
    'bdbdbd', // 8 unused
    'd9d9d9', // 9 cloud
    'c2b280', // 10 land
    '08306b'  // 11 permanent water
  ]
};

Map.addLayer(img, vistaArchiveVis, 'VISTA dekadal archive');
Map.centerObject(img.geometry(), 3);

Archive Product: Monthly

The monthly archive contains time-specific rolling monthly observations covering the same period as the dekadal archive, from the third dekad of January 2012 to the third dekad of April 2026. Each product is centered on a dekad and is built from the previous, current, and following dekads when available. At least two source dekads are required. It is not a strict calendar-month product.

Example:

vista_monthly_202401d2

This summarizes maximum flood extent and intensity around center dekad 202401d2.

Earth Engine root:

projects/fao-oer/assets/vista_monthly_archive

Bucket access:

Content:

  • single uint8 band named monthly_class; .select(0) also works
  • values use the archive class values 0, 1, 2, 3, 4, 5, 6, 9, 10, 11

Earth Engine script:

var img = ee.Image('projects/fao-oer/assets/vista_monthly_archive/vista_monthly_202401d2');

var vistaArchiveVis = {
  min: 0,
  max: 11,
  palette: [
    '000000', // 0 unclassified / nodata
    '7fcdbb', // 1 flood low
    '41b6c4', // 2 flood medium
    '225ea8', // 3 flood high
    'ffffff', // 4 snow cover
    'b3e5fc', // 5 river/lake ice cover
    '00ffff', // 6 supra-snow/ice water, mixed ice-water, or ice in melting status
    'bdbdbd', // 7 unused
    'bdbdbd', // 8 unused
    'd9d9d9', // 9 cloud
    'c2b280', // 10 land
    '08306b'  // 11 permanent water
  ]
};

Map.addLayer(img, vistaArchiveVis, 'VISTA monthly archive');
Map.centerObject(img.geometry(), 3);

Statistical Product: Count

Count products summarize how many yearly observations were assigned to each class for a recurring dekad or monthly center-dekad. They answer questions such as: “How many times, or years, between 2012 and 2026 was this pixel covered by flood, cloud, land, snow, ice, or permanent water during the first dekad of January?” For monthly statistics, the same question applies to the rolling monthly window centered on the selected dekad key.

The number of contributing years varies by dekad, because the archive does not cover every dekad in every year. Each count asset records its own year_start, year_end, n_years, and years properties; read those before comparing one dekad with another.

Examples:

vista_count_01d1
vista_count_monthly_01d1

Earth Engine roots:

projects/fao-oer/assets/vista_products/dekadal/count
projects/fao-oer/assets/vista_products/monthly/count

Bucket examples:

  • Dekadal direct example: https://storage.googleapis.com/vista_project/vista_products/dekadal/count/vista_count_01d1.tif
  • Use the VISTA bucket browser for monthly count files and additional GeoTIFF examples.

Content:

Band Meaning
class_0_nodata Count of no-data or unclassified observations
class_1_flood_low Count of low-flood observations
class_2_flood_medium Count of medium-flood observations
class_3_flood_high Count of high-flood observations
class_4_snow Count of snow observations
class_5_river_lake_ice Count of river/lake ice observations
class_6_supra_snow_melting_ice Count of supra-snow/mixed/melting ice observations
class_9_cloud Count of cloud observations
class_10_land Count of land observations
class_11_permanent_water Count of permanent-water observations
N_tot Total observations counted
N_valid Total observations excluding cloud and no-data
N_flood_any class_1_flood_low + class_2_flood_medium + class_3_flood_high
N_flood_medium_high class_2_flood_medium + class_3_flood_high

Dekadal and monthly count products share this same 14-band schema, so the same band names work for vista_count_<dekad> and vista_count_monthly_<center-dekad>. Note that these bands hold counts; the frequency products use freq_* names for the corresponding proportions.

Earth Engine script:

var count = ee.Image('projects/fao-oer/assets/vista_products/dekadal/count/vista_count_01d1');
var floodAnyCount = count.select('N_flood_any');

Map.addLayer(
  floodAnyCount,
  {min: 0, max: 14, palette: ['ffffff', 'ccece6', '66c2a4', '238b45', '00441b']},
  'Count of flood-any observations, 01d1'
);

Statistical Product: Frequency

The freq_flood_any band is also mirrored in FAO STAC; that single-band copy does not include the observation-support layers.

Frequency products summarize how often each class occurred for a recurring dekad or monthly center-dekad. Values are float32 proportions in the 0.0-1.0 range. A value of 0.30 means that the condition occurred in 30 percent of the valid observations used by that product.

Examples:

vista_frequency_01d1
vista_frequency_monthly_01d1

Earth Engine roots:

projects/fao-oer/assets/vista_products/dekadal/frequency
projects/fao-oer/assets/vista_products/monthly/frequency

Bucket examples:

  • Dekadal direct example: https://storage.googleapis.com/vista_project/vista_products/dekadal/frequency/vista_frequency_01d1.tif
  • Use the VISTA bucket browser for monthly frequency files and additional GeoTIFF examples.

Content:

Dekadal and monthly frequency products share the same 14 float32 bands, all prefixed freq_, matching the count schema one-for-one.

Band Meaning Denominator
freq_0_nodata Frequency of no-data N_tot
freq_1_flood_low Frequency of low flood N_valid
freq_2_flood_medium Frequency of medium flood N_valid
freq_3_flood_high Frequency of high flood N_valid
freq_4_snow Frequency of snow N_valid
freq_5_river_lake_ice Frequency of river/lake ice N_valid
freq_6_supra_snow_melting_ice Frequency of supra-snow/mixed/melting ice N_valid
freq_9_cloud Frequency of cloud N_tot
freq_10_land Frequency of land N_valid
freq_11_permanent_water Frequency of permanent water N_valid
freq_tot N_tot / N_tot; always 1 where data exist N_tot
freq_valid Share of observations that were valid N_tot
freq_flood_any Frequency of any flood class (1 + 2 + 3) N_valid
freq_flood_medium_high Frequency of medium or high flood (2 + 3) N_valid

Each band is masked where its own denominator is 0. Bands with an N_tot denominator therefore survive on pixels that never had a valid observation, where the N_valid bands are masked.

Use freq_flood_any to map the frequency of any flood class, and freq_flood_medium_high for medium or high flood only. Use freq_valid to screen out pixels whose statistics rest on too few valid years before interpreting any flood frequency.

Earth Engine script:

var floodFrequency = ee.Image(
  'projects/fao-oer/assets/vista_products/dekadal/frequency/vista_frequency_01d1'
).select('freq_flood_any');

// The monthly product uses the same band names:
// var floodFrequency = ee.Image(
//   'projects/fao-oer/assets/vista_products/monthly/frequency/vista_frequency_monthly_01d1'
// ).select('freq_flood_any');

// Convert 0.0-1.0 frequency to 0.1 classes for display.
var floodFrequencyBins = floodFrequency
  .multiply(10)
  .floor()
  .clamp(0, 10)
  .toByte();

var frequencyPalette = [
  'ffffff', // 0.0-0.1
  'e5f5f9', // 0.1-0.2
  'ccece6', // 0.2-0.3
  '99d8c9', // 0.3-0.4
  '66c2a4', // 0.4-0.5
  '41ae76', // 0.5-0.6
  '238b45', // 0.6-0.7
  '006d2c', // 0.7-0.8
  '00441b', // 0.8-0.9
  '08306b', // 0.9-1.0
  '000000'  // 1.0
];

Map.addLayer(
  floodFrequencyBins,
  {min: 0, max: 10, palette: frequencyPalette},
  'Flood-any frequency, 01d1, 0.1 bins'
);

var legendLabels = [
  '0.0-0.1', '0.1-0.2', '0.2-0.3', '0.3-0.4', '0.4-0.5',
  '0.5-0.6', '0.6-0.7', '0.7-0.8', '0.8-0.9', '0.9-1.0', '1.0'
];

var legend = ui.Panel({style: {position: 'bottom-left', padding: '8px 12px'}});
legend.add(ui.Label({value: 'Flood-any frequency', style: {fontWeight: 'bold'}}));

function makeRow(color, name) {
  return ui.Panel({
    widgets: [
      ui.Label({style: {backgroundColor: '#' + color, padding: '8px', margin: '0 0 4px 0'}}),
      ui.Label({value: name, style: {margin: '0 0 4px 6px'}})
    ],
    layout: ui.Panel.Layout.Flow('horizontal')
  });
}

for (var i = 0; i < frequencyPalette.length; i++) {
  legend.add(makeRow(frequencyPalette[i], legendLabels[i]));
}

Map.add(legend);

Statistical Product: Dominant Class

Dominant-class products summarize the most frequent class for a recurring dekad or monthly center-dekad. They are useful for quick cartographic summaries of typical conditions, but count and frequency products are better for quantitative interpretation.

Examples:

vista_dominant_class_01d1
vista_dominant_class_monthly_01d1

Earth Engine roots:

projects/fao-oer/assets/vista_products/dekadal/dominant_class
projects/fao-oer/assets/vista_products/monthly/dominant_class

Bucket examples:

  • Dekadal direct example: https://storage.googleapis.com/vista_project/vista_products/dekadal/dominant_class/vista_dominant_class_01d1.tif
  • Use the VISTA bucket browser for monthly dominant-class files and additional GeoTIFF examples.

Content:

Dominant-class products contain one uint8 band named dominant_class.

Value Meaning
0 No data
4 Snow cover
5 River or lake ice cover
6 Supra-snow water, mixed ice-water, or melting ice
9 Cloud
10 Land
11 Permanent water
12 Flood any

Flood low, medium, and high are merged into flood_any class 12 before the dominant class is computed. Ties are resolved in this priority order:

permanent water (11)
> flood any (12)
> supra-snow / mixed / melting ice (6)
> river or lake ice cover (5)
> snow cover (4)
> land (10)
> cloud (9)
> no data (0)

If cloud is initially dominant, VISTA promotes the best real class only when that fallback class accounts for at least 30 percent of N_tot. Otherwise the pixel remains cloud-dominant. “Real” classes are permanent water, flood any, supra-snow/mixed/melting ice, river or lake ice, snow, and land; cloud and no-data are never used as the fallback.

Pixels with N_tot equal to 0 are forced to 0 (no data). Without this rule the tie priority above would assign permanent water to pixels that were never observed at all.

Earth Engine script:

var dominant = ee.Image(
  'projects/fao-oer/assets/vista_products/dekadal/dominant_class/vista_dominant_class_01d1'
).select('dominant_class');

// Remap sparse class values to consecutive display values.
var display = dominant.remap(
  [0, 4, 5, 6, 9, 10, 11, 12],
  [0, 1, 2, 3, 4, 5, 6, 7]
);

var dominantPalette = [
  '000000', // 0 no data
  'ffffff', // 4 snow cover
  'b3e5fc', // 5 river/lake ice cover
  '00ffff', // 6 supra-snow/ice water, mixed ice-water, or melting ice
  'd9d9d9', // 9 cloud
  'c2b280', // 10 land
  '08306b', // 11 permanent water
  '225ea8'  // 12 flood any
];

Map.addLayer(
  display,
  {min: 0, max: 7, palette: dominantPalette},
  'Dominant class, 01d1'
);

Grid and dimensions

All VISTA products are EPSG:4326 with a pixel size of 0.003372302158273381 degrees, approximately 375 m at the equator, and share the same grid origin. Pixel centres therefore align exactly across every product family, and products can be compared directly without resampling.

Raster dimensions differ slightly between the dekadal and monthly families:

Family Dimensions
Dekadal archive and dekadal statistical products 106753 x 44482 px
Monthly archive and monthly statistical products 106754 x 44482 px

The monthly products carry one additional column at the eastern edge, covering 180.000000 to 180.003372 degrees, just past the antimeridian. Because the origin and pixel size are identical, this does not shift or misalign anything: it only means the monthly rasters extend one pixel further east.

Software that reads these rasters through their geotransform, including Earth Engine, GDAL, QGIS, and rasterio, handles this correctly. Code that compares rasters as plain arrays by row and column index must account for the differing width, otherwise a dekadal and a monthly raster will either fail to broadcast or be silently offset by one column across the whole image.

Band naming summary

One table for every VISTA product family. Check it before writing any .select() call.

Product Asset name pattern Bands Naming scheme
Dekadal archive vista_dekad_<YYYYMMdD> 1 single uint8 band; use .select(0)
Monthly archive vista_monthly_<YYYYMMdD> 1 monthly_class
Dekadal count vista_count_<dekad> 14 class_*, N_tot, N_valid, N_flood_any, N_flood_medium_high
Monthly count vista_count_monthly_<center-dekad> 14 same as dekadal count
Dekadal frequency vista_frequency_<dekad> 14 freq_*
Monthly frequency vista_frequency_monthly_<center-dekad> 14 same as dekadal frequency
Dekadal dominant class vista_dominant_class_<dekad> 1 dominant_class
Monthly dominant class vista_dominant_class_monthly_<center-dekad> 1 dominant_class

Count products use class_* names with N_* summary bands and hold counts. Frequency products use freq_* names throughout and hold proportions in the 0.0-1.0 range. Dekadal and monthly products of the same family always share a schema.

To confirm the schema of any asset before relying on it:

print(ee.Image('projects/fao-oer/assets/vista_products/dekadal/frequency/vista_frequency_01d1')
  .bandNames());

Comments and limitations

  • VISTA inherits the strengths and limitations of VIIRS, a daily global optical product at approximately 375 m resolution. Where VIIRS cannot observe floods, because of cloud cover, very short-lived events, events smaller than the sensor can resolve, or other optical/sensor constraints, VISTA has the same limitation.
  • The products inherit uncertainties from the NOAA VIIRS source data, including source classification uncertainty and missing source composites.
  • Cloud and no-data observations can strongly affect interpretation. Use N_tot, N_valid, class_9_cloud, and class_0_nodata in count products, and freq_valid, freq_9_cloud, and freq_0_nodata in frequency products, to assess reliability.
  • Dekadal products are aggregated from available 5-day VIIRS composites using VISTA priority rules. They are not daily products.
  • Monthly products are centered rolling 3-dekad composites, not strict calendar months.
  • Multi-year statistics depend on the currently available archive years. Check N_tot and N_valid before comparing places or periods.
  • Dominant-class products intentionally merge all flood severities into flood_any class 12; use count or frequency products when flood severity matters.
  1. Open VISTA Explorer and choose Compare a dekad. Select an available year and dekad, inspect the observed classes, then choose the reference window and flood definition. Adjust the reference-frequency threshold to examine how the seasonal and exceptional classifications change.
  2. Click pixels to check the observed class, historical frequencies, valid-observation support and annual class strip before interpreting the comparison. Treat cloud as obscured and pixels without sufficient reference support as unclassified.
  3. Browse or download source products in the VISTA bucket browser when GIS analysis or local processing is required.
  4. Use count products to answer concrete recurrence questions, for example: “In how many years was this pixel flooded during the first dekad of January?”
  5. Use frequency products for the same type of recurrence information, normalized by the number of valid observations, so pixels and regions can be compared more consistently.
  6. Use dominant-class products for quick cartographic summaries, but return to count or frequency products for quantitative interpretation.