The most trusted maps of soil moisture come from NASA's SMAP satellite, which has flown only since 2015. A European radar has measured the land since 2007, in a different way. This study teaches a network to turn the radar's soil moisture into what SMAP would have seen, and then reads soil moisture and the density of the vegetation from it as SMAP does. Over the United States the result reaches back to 2007.
At a glance
- Theme: A longer record of soil moisture and vegetation, as SMAP would see them
- Observations: The ASCAT radar, since 2007, and the SMAP radiometer, since 2015; ground stations of six networks for checking
- Approach: A learned translation from the radar to SMAP, followed by SMAP's own physical retrieval
- Area and period: The contiguous United States, 2007 to 2023
- Lead author: Dr. Jaese Lee
- Status: Published

The basics
Two kinds of satellite sense the moisture of the soil with microwaves. SMAP listens to the radiation that the ground gives off by itself, at a long wavelength called L band, which passes through vegetation better than shorter ones. It measures at two polarisations, and from the pair two things can be found: the moisture of the soil and the vegetation optical depth, which says how much the plants weaken the signal and goes with the water and the mass they hold.
ASCAT is a radar on the European Metop weather satellites. It sends pulses at a shorter wavelength, C band, and measures the echo. Its soil moisture record began in 2007, but it is a different kind of measurement, it sees less far into vegetation, and it gives no vegetation optical depth at L band.
The idea
The method has three steps. First, in the years 2015 to 2017, when both satellites flew, two networks learn to turn maps of the radar's soil moisture into the soil moisture that SMAP derives from each of its two polarisations. Second, the physics of microwave emission is used to work out the brightness that SMAP would have measured at both polarisations. Third, SMAP's own retrieval is applied to that brightness and gives soil moisture and vegetation optical depth together.
The first version of the method produced soil moisture only. This second version adds the second polarisation, and with it the vegetation. The last step was run in two ways: with the settings of the standard SMAP product, and with newer settings for how vegetation scatters and how rough the ground is.
What was found
In the years 2019 to 2023, which were kept apart for testing, the maps of soil moisture and of vegetation optical depth resemble those of SMAP: wetter in the east of the United States than in the west, and the densest vegetation over the forests. The real test is the time before SMAP, where only ground stations can judge.

From 2007 to March 2015 the random error against the stations falls from 0.071 m³/m³ for the radar's own product to 0.055 in open land, and from 0.082 to 0.057 in forest. That is close to the ESA CCI record, which merges many satellites. With the newer settings the values are also nearly free of bias, within 0.02 m³/m³, where the standard SMAP product is too wet by about 0.06 in forest.
Three sites with many sensors inside one satellite cell allow a closer look. In Texas the result follows the ground with a correlation of 0.78, against 0.85 for SMAP itself. At Millbrook, a forest site in New York, it reaches 0.71 and is too dry by 0.03 m³/m³, where SMAP is too wet by 0.08. At the forest site in Massachusetts all products do less well; the new one reaches 0.53 with no bias.
The vegetation optical depth follows the seasons much as SMAP's does, most closely over forests.
What the result does not say
The paper is careful about this. The work covers the United States only. The vegetation signal is SMAP-like by construction and has not been checked against independent measurements of vegetation, which hardly exist. The checks of soil moisture lean in part on SMAP itself. And the translation is learned, not derived: it removes known flaws of the radar over dry ground in effect, without describing their physics. Before the record is used for long-term trends of vegetation or carbon, it needs independent tests.
Why it matters
Droughts, the greening of the land and the memory of the soil are questions of decades, and eleven years of SMAP are short. If the older radar can be read in SMAP's terms, the record gains eight years, and the same route could carry it on after SMAP ends. The lab's project on soil moisture under forests takes up one of the open points the paper names: letting the scattering by vegetation change with time.
The paper
Lee, J., Colliander, A., Im, J., and Kim, H. (2026). ASCAT2SMAP-V2: Deriving L-band-like soil moisture and vegetation optical depth from C-band satellite observations. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 19, 29719–29733. The work was done with colleagues at the Finnish Meteorological Institute and UNIST. The article is open access.












