Data

The Trend Detective reads 34 data layers in 17 categories, from 10 NASA and NOAA product families, averaged over river sub-basins month by month. This page lists every layer, how the basin series are made, how the data is versioned and whom to credit.

Every layer

Each of the 17 variable categories has a default layer, used unless you switch product in the analysis view. A basin only offers the layers that have a series there: sea-surface temperature, for example, is missing over most land basins. Whenever a layer has a caveat, the analysis view shows it too.

SO₂ is annual. The SO₂ layer is a yearly point-source emissions catalogue whose value is repeated in every month of the year. It has no monthly variation, so the detective does not offer it for monthly anomaly, rank or trend tests, even through a typed URL.

Every data layer in the catalogue, grouped by variable category; the default layer of each category is listed first
Layer	Product	Agency	Type	Unit	Period	Time step	Caveat
PrecipitationHow much rain and snow falls from the sky.
Precipitation`p_imerg`Default	GPM IMERG monthly	NASA	Satellite, gauge-adjusted	mm/day	Jan 1998 – Sep 2025	Monthly
Precipitation`p_gpcp`	GPCP v2.3 (published by NOAA PSL)	NOAA	Satellite and gauges	mm/day	Jan 1979 – Oct 2026	Monthly	GPCP is published by NOAA PSL, not through a NASA portal.
Precipitation (bias-corrected)`p_m2`	MERRA-2 M2TMNXFLX PRECTOTCORR	NASA	Reanalysis	mm/day	Jan 1980 – Sep 2026	Monthly
Precipitation over land`p_m2land`	MERRA-2 M2TMNXLND PRECTOTLAND	NASA	Reanalysis	mm/day	Jan 1980 – Sep 2026	Monthly
EvapotranspirationWater that soil and plants send back into the air.
Evaporation`et_m2`Default	MERRA-2 M2TMNXFLX EVAP	NASA	Reanalysis	mm/day	Jan 1980 – Sep 2026	Monthly
Evaporation over land`et_m2land`	MERRA-2 M2TMNXLND EVLAND	NASA	Reanalysis	mm/day	Jan 1980 – Sep 2026	Monthly
Evapotranspiration`et_gldas`	GLDAS Noah 2.1 0.25°→0.5°	NASA	Land-surface model	mm/day	Jan 2000 – Jul 2026	Monthly	GLDAS-2.1 evapotranspiration has a known long-term drift, so its trend is unreliable.
Soil moistureHow much water is in the soil where plant roots grow.
Root-zone soil moisture`sm_root`Default	GLDAS Noah 2.1	NASA	Land-surface model	mm	Jan 2000 – Jul 2026	Monthly
Root-zone soil wetness (0–1)`sm_m2`	MERRA-2 M2TMNXLND GWETROOT	NASA	Reanalysis	fraction	Jan 1980 – Sep 2026	Monthly
Soil moisture, 0–2 m`sm_total`	GLDAS Noah 2.1	NASA	Land-surface model	mm	Jan 2000 – Jul 2026	Monthly
Top-layer soil moisture, 0–10 cm`g1_soil_top`	GLDAS_NOAH10_M 2.1 (1°)	NASA	Land-surface model	mm	Jan 2000 – Jul 2026	Monthly
Terrestrial water storageChange in all water on land: surface, soil and groundwater.
Terrestrial water storage anomaly`tws`Default	GRACE/GRACE-FO JPL mascon	NASA	Satellite gravimetry	cm	Apr 2002 – Sep 2026	Monthly	GRACE resolves roughly 300 km, so small basins are smoothed. Large earthquakes cause gravity jumps that are not hydrological.
RunoffRain that the ground does not soak up and flows to rivers.
Runoff`runoff`Default	GLDAS Noah 2.1	NASA	Land-surface model	mm/day	Jan 2000 – Jul 2026	Monthly
Runoff`ro_m2`	MERRA-2 M2TMNXLND RUNOFF+BASEFLOW	NASA	Reanalysis	mm/day	Jan 1980 – Sep 2026	Monthly
Snow waterHow much water the snow on the ground would melt into.
Snow water equivalent`swe`Default	GLDAS Noah 2.1	NASA	Land-surface model	mm	Jan 2000 – Jul 2026	Monthly
Snow-cover fraction`snow_cover`	MODIS Terra MOD10CM v6.1 monthly snow cover (CMG 0.05°)	NASA	Satellite	%	Mar 2000 – Oct 2026	Monthly	Only valid values 0 to 100 are used. Cloud, polar night and water are excluded, so high-latitude winter regions often have no value. Antarctica is fixed as snow by the product and is masked. No file exists for 2000-08, 2001-06, 2002-03, 2003-12, 2016-02 and 2022-10 (empty).
Air temperatureTemperature of the air near the ground.
Near-surface air temperature`tair`Default	GLDAS Noah 2.1 forcing	NASA	Land-surface model	°C	Jan 2000 – Jul 2026	Monthly
Surface temperature`tsurf_m2`	MERRA-2 M2TMNXLND TSURF	NASA	Reanalysis	°C	Jan 1980 – Sep 2026	Monthly
Daytime land-surface temperature`lst_day`	MODIS Terra LST (orbit drift after 2020)	NASA	Satellite	°C	Feb 2000 – Oct 2026	Monthly	MODIS Terra's orbit has drifted since 2020, which creates a false cooling trend.
Night-time land-surface temperature`lst_night`	MODIS Terra MOD11C3 LST_Night (0.05° to 0.25° to 0.5°)	NASA	Satellite	°C	Feb 2000 – Oct 2026	Monthly	Terra's orbit has drifted since 2020, so its overpass time is earlier.
Air dryness (VPD)How thirsty the air is; higher pulls water out faster.
Vapour pressure deficit`vpd`Default	GLDAS Noah 2.1 forcing (derived)	NASA	Land-surface model	kPa	Jan 2000 – Jul 2026	Monthly
Potential evapotranspiration`pet_gldas`	GLDAS Noah 2.1	NASA	Land-surface model	mm/day	Jan 2000 – Jul 2026	Monthly	GLDAS-2.1 potential evapotranspiration has a known long-term drift, so its trend is unreliable.
Net radiationHeat energy the ground gains from sunlight and the sky.
Net radiation`rn`Default	GLDAS Noah 2.1	NASA	Land-surface model	W/m2	Jan 2000 – Jul 2026	Monthly
Net radiation`rn_m2`	MERRA-2 M2TMNXLND SWLAND+LWLAND	NASA	Reanalysis	W/m2	Jan 1980 – Sep 2026	Monthly
Downward shortwave radiation`swdown`	GLDAS Noah 2.1 forcing	NASA	Land-surface model	W/m2	Jan 2000 – Jul 2026	Monthly
Surface net radiation`rn_ceres`	CERES EBAF Ed4.2.1 sfc_net_tot_all_mon	NASA	Satellite and radiative model	W/m2	Mar 2000 – Jul 2026	Monthly	Surface fluxes are inferred from satellite top-of-atmosphere radiation with a radiative transfer model, on a 1° grid, so small regions are poorly represented.
Wind speedHow fast the wind blows near the ground.
Wind speed`wind`Default	GLDAS Noah 2.1 forcing	NASA	Land-surface model	m/s	Jan 2000 – Jul 2026	Monthly
Vegetation greennessHow green and leafy the plants are.
Vegetation index (NDVI)`ndvi`Default	MODIS (0.05° to 0.5°)	NASA	Satellite	NDVI	Feb 2000 – Oct 2026	Monthly
Sea surface temperatureHow warm the water at the sea surface is.
Sea surface temperature`sst`Default	NOAA OISST v2.1	NOAA	Satellite and in situ	°C	Jan 1982 – Oct 2026	Monthly	NOAA OISST, not a NASA product.
Aerosol optical depthHow strongly particles in the atmospheric column block light.
Aerosol optical depth (550 nm)`aod`Default	MERRA-2 M2TMNXAER TOTEXTTAU	NASA	Reanalysis	1	Jan 1980 – Sep 2026	Monthly	A reanalysis model that assimilates satellite AOD (MODIS and others). Before 2000 fewer observations could be assimilated.
Total-column ozoneThe total amount of ozone in the atmospheric column.
Total-column ozone`to3_m2`Default	MERRA-2 M2TMNXSLV TO3	NASA	Reanalysis	DU	Jan 1980 – Sep 2026	Monthly
Atmospheric water vapourHow much water all the water vapour in the atmospheric column would make if condensed.
Precipitable water vapour`tqv`Default	MERRA-2 M2TMNXSLV TQV	NASA	Reanalysis	mm	Jan 1980 – Sep 2026	Monthly
Annual SO₂ emissionsHow much sulfur dioxide large point sources emit each year.
SO₂ point-source emissions`so2_emis`Default	MSAQSO2L4 v2 multi-satellite SO₂ emission catalogue (OMI, OMPS, TROPOMI)	NASA	Satellite-derived catalogue	kt/yr	2005 – 2025	AnnualNot used for monthly analysis	This is an annual point-source catalogue (780 large polluters and volcanoes), not gridded concentration. Each year's value is repeated across all 12 months, so there is no monthly variation. Regions with no catalogued source are empty. Overall uncertainty is about 50%. The regional totals may count missing sources as 0, and the raw files were not reprocessed. Monthly anomaly and trend tests are not provided.
Vegetation-fire detectionsHow many vegetation-fire detections satellites record each month.
Vegetation-fire detections`fire_count`Default	MODIS FIRMS fire hotspots (Terra, national zips merged and de-duplicated)	NASA	Satellite detections	count/month	Nov 2000 – Dec 2025	Monthly	Counts Terra MODIS vegetation fires only (type 0; volcano, industrial and offshore excluded), with no confidence filter. Only Terra is used so that Aqua joining in 2002-07 does not double the counts, but Terra's orbit has drifted since about 2020 and its overpass is earlier, so later counts may be slightly off. A hotspot is a detection at the moment of the satellite pass, not a burned area. Values are empty before 2000-11; months with no detection in the region are 0.

How a basin series is made

Basins are HydroBASINS level-4 river sub-basins; 1,313 of them have series. Their outlines are rasterised once onto a 0.5° grid, and every layer uses that same mask.
Each gridded product is brought onto the 0.5° grid (finer products such as MODIS at 0.05° are aggregated). A basin’s monthly value is the mean over its cells, weighted by the cosine of latitude so each cell counts in proportion to its area.
If valid data cover less than 30 % of the basin’s area in a month, that month is missing. Missing months stay missing: they are never filled in or treated as zero, charts leave a gap, and the statistics skip them.
Fire detections are point records: the count of Terra MODIS vegetation-fire detections inside the basin each month, with duplicates between country files removed. A zero means no recorded detection, not proof that there was no fire. SO₂ sums the catalogued point sources inside the basin for each year.
Layers carried over from the project’s first data snapshot were made with an earlier script on the same grid. Rebuilding five GLDAS layers and IMERG precipitation with the current pipeline reproduced them (median correlation 1.000 for GLDAS, at least 0.999 for IMERG).

Nothing is precomputed: the detective downloads the basin’s series file and computes anomalies, tests and trends in your browser. See Methods.

How the data is versioned

Processed data is not kept in git. It is published as one archive attached to a GitHub Release, and web/data.lock.json pins which one: repository, release tag, file name and SHA-256 checksum. npm run setup downloads that archive and refuses to unpack it if the checksum differs. Changing the data therefore means a new release and a new lock file, visible in the project history.

Release: data-v2
Archive: aqua-trend-data-v2.tar.gz
SHA-256: 3fc543130244d98661050844b853b1984a92f654f4f1f4c2028e12e72e193b89

The release manifest checks file structure and index consistency. It does not independently reproduce the scientific values; the processing scripts and their inputs are documented in the repository’s pipeline/ folder.

Credits and citations

GLDAS: Beaudoing, H. and M. Rodell, NASA/GSFC/HSL, GLDAS Noah Land Surface Model L4 monthly V2.1 (0.25°; the 1° product for top-layer soil moisture). NASA Goddard Earth Sciences Data and Information Services Center (GES DISC).
MERRA-2: NASA Global Modeling and Assimilation Office reanalysis, monthly collections M2TMNXFLX, M2TMNXLND, M2TMNXAER and M2TMNXSLV, from GES DISC.
GPM IMERG monthly precipitation, NASA, from GES DISC.
GRACE and GRACE-FO JPL mascon terrestrial water storage, NASA JPL.
MODIS: vegetation index (NDVI), Terra land-surface temperature (MOD11C3) and MOD10CM Collection 6.1 snow cover.
CERES EBAF Edition 4.2.1 surface net radiation, NASA.
NASA FIRMS MODIS Collection 6.1 active-fire archive.
MSAQSO2L4 V2 multi-satellite SO₂ emissions catalogue (OMI, OMPS, TROPOMI), NASA.
NOAA: GPCP v2.3 precipitation, published by NOAA PSL, and NOAA OISST v2.1 sea-surface temperature. These are not NASA products and are labelled NOAA wherever they appear.
Basin outlines: HydroBASINS levels 3 and 4.
NASA Earthdata (CMR search and DAAC downloads) for the pipeline; Giovanni for spot cross-checks of single regions.

Software: Next.js, React, three.js and d3 (MIT licence). Fonts: Literata, Public Sans, Fredoka and Figtree (SIL Open Font License), self-hosted. Variable suggestions: the OneJev 1.6 0.8B model, run locally.

This is a NASA Space Apps Challenge 2026 entry, not a NASA product, and it does not use NASA branding.

AI disclosure. The challenge rules ask entrants to disclose AI use. The team set the research questions, product requirements and scientific rules. AI tools helped draft and check documentation and wrote much of this site’s code and interface copy, which the team reviewed. An AI model suggests which variables to look at; it never produces a number. Every anomaly, p value, rank and trend is computed from the data files through the steps on this page and on Methods.