Land-surface temperature
Hotter land often points toward drier conditions, especially where vegetation is sparse.
Can satellites see water hidden in the soil?
A few sensors cannot describe an entire watershed. This research asked whether heat, vegetation, and satellite observations could reveal a sharper picture of where the land is wet—and where it is drying.
Begin the story51,876 km²Urmia Lake basin
3ground stations
10–25 kmoriginal pixels
≈1 kmdownscaled maps
8 daysMODIS interval
02
The blurry satellite view
A basin inside a pixel
Satellites saw everything—at the scale of a city.
Microwave satellite and model products can estimate soil moisture over enormous regions. Their weakness is detail: one pixel may be 10 to 25 kilometres wide.
Inside that one block may be farms, bare land, mountains, roads, settlements, and water. The regional average is useful, but it is too coarse for many agricultural and basin decisions.
Original view10–25 km
→Downscaled view≈1 km
03
The landscape clues
Heat · Vegetation · Moisture
The land leaves two clues.
Dry ground usually becomes hotter. Wet soil often remains cooler because energy is used to evaporate water.
Vegetation provides a second clue. Healthy plants alter both surface temperature and reflectance, helping distinguish bare soil, partial cover, and dense vegetation.
Vegetation condition
Greenness helps separate bare soil, partial cover, and dense vegetation across the basin.
TVDI
The Temperature Vegetation Dryness Index positions each pixel between estimated wet and dry edges.
Mixed conditions
The heat and vegetation clues suggest neither an extremely dry nor an extremely wet surface.
04
Sharpening the picture
The UCLA method
One broad measurement. Thousands of local clues.
The large satellite pixel supplied the overall moisture signal. Local heat and vegetation patterns suggested how that moisture should vary inside it.
The result was a set of maps with an approximate spatial resolution of 1 km—much sharper than the original 10–25 km products.
Start broad
Collect soil-moisture products with pixels around 10–25 km wide.
Read the land
Use MODIS to map surface temperature and vegetation every eight days.
Estimate dryness
Combine heat and vegetation into the Temperature Vegetation Dryness Index.
Downscale
Redistribute the broad signal into a sharper map at approximately 1 km resolution.
↔
Sharper local variationCoarse regional blocks
05
A basin through the seasons
2010 · 2014
The map changed as winter moisture faded and irrigation took over.
Moisture did not simply rise or fall everywhere at once. Every land use followed its own seasonal story.
Highlands began wetter, much of the basin dried as temperatures rose, and irrigated agricultural plains could stay wetter later into summer.
Seasonal reading
The highlands began wetter.
Mountains and elevated areas held more surface moisture because of winter precipitation and snow cover.
One basin-wide average can hide very different local conditions.
06
The reality check
Satellite vs ground
A detailed map matters only if it agrees with reality.
The downscaled estimates were compared with soil-moisture observations measured at approximately 5 cm depth.
The comparison used three available ground stations and evaluated correlation together with several error statistics.
2010Best overall result
LPRM
At the Tabriz University research farm, LPRM had the strongest overall combination of correlation and error statistics.
- Correlation
- 0.94
- MAE
- 3.87%
- RMSE
- 5.13%
2014Best error performance
ESA-CCI
Across the two validation stations, ESA-CCI produced the strongest overall error performance, although GLDAS sometimes showed higher correlation.
- Best MAE*
- 2.23%
- Best RMSE*
- 2.59%
- R at Khosrowshah
- 0.70
The honest result
No single product was best everywhere and in every year. Performance depended on the sensor, model, land cover, and available ground observations. The research supports careful use—not blind trust.
07
Why it matters
From map to decision
A sharper map can lead to sharper water decisions.
The method does not create water or replace field instruments. It helps reveal where attention, measurement, and management may be needed most.
Irrigation planning
Identify areas showing water stress and compare conditions across agricultural land.
Drought monitoring
Track where dry conditions are expanding and how different land uses respond.
Runoff modelling
Improve understanding of how much rainfall may infiltrate or become surface runoff.
Field priorities
Decide where new monitoring stations and ground investigations may add the most value.
Water management
See basin-scale patterns without losing the local variations that shape decisions.
Research design
Target future measurements toward places where satellite products disagree or uncertainty is high.
08
Keeping the story honest
What the maps cannot see
Powerful evidence, with visible limits.
Remote sensing expands our view, but it remains an estimate of near-surface conditions—not a perfect measurement of every root zone and field.
The method is best treated as a powerful complement to field monitoring, never a replacement for it.
Near-surface moisture
Validation focused on observations around 5 cm depth. Deeper root-zone moisture may behave differently.
Limited ground network
Only three stations were available. A denser network would provide a stronger and more representative test.
Sharper, not field-scale
About 1 km is a major improvement over 10–25 km, but one mapped cell can still contain several fields and land-cover types.
Optional technical layer
For readers who want to look beneath the story.
What are NDVI, LST, and TVDI?
NDVI is a vegetation index derived from red and near-infrared reflectance. LST is land-surface temperature derived from thermal observations. TVDI positions each pixel between estimated wet and dry edges in LST–vegetation space.
Which soil-moisture products were compared?
The study evaluated ESA-CCI, LPRM, AMSR-E/AMSR2, and GLDAS products, depending on data availability in 2010 and 2014. Their original spatial resolution was typically about 10 or 25 km.
How did the UCLA downscaling work?
The method used the ratio between MODIS-scale TVDI and the average TVDI within each coarse product pixel as a spatial scaling factor. In simple terms, the broad soil-moisture value was redistributed using local wetness and dryness patterns while preserving the larger-scale signal.
How were the results evaluated?
Ground observations were compared with the estimates using correlation, mean bias error, mean absolute error, root mean square error, centred RMSE, and model efficiency. Full values appear in Table 5 of the published paper.
Published research · 2023
Soil moisture monitoring by downscaling remote sensing products
Water Supply · Volume 23 · Issue 2
The research behind this story
From scientific paper to public understanding.
- Authors
- Amin Rostami, Mahmoud Raeini-Sarjaz, Jafar Chabokpour, and Aaron Anil Chadee
- Citation
- Rostami, A. et al. (2023). Soil moisture monitoring by downscaling of remote sensing products using LST/VI space derived from MODIS products. Water Supply, 23(2), 688–705.
- DOI
- 10.2166/ws.2023.002
Most visuals in this story were created specifically for Hydroism. One original figure from the paper is retained for geographic and land-use context.