The real missing variable of the water cycle obtained from remote sensing is the root zone soil moisture. We have surface soil moisture from microwave sensors (e.g. ASCAT, AMSR, SMOS, SMAP), total water storage from gravity (e.g., GRACE), but we have no measurements for root-zone soil moisture.

An indirect estimate of root-zone soil moisture from remote sensing needs some modelling: 1) simplified, e.g. the exponential filter (https://www.researchgate.net/publication/233946180_A_physically_based_approach_for_the_estimation_of_root-zone_soil_moisture_from_surface_measurements?ev=prf_pub), 2) land surface modelling, e.g. the SMAP root-zone product (L4_SM, http://smap.jpl.nasa.gov/data/)

Article A physically based approach for the estimation of root-zone ...

Hi, you can also use Landsat L5 imagery from USGS.

http://earthexplorer.usgs.gov/

It's available from 1984. Just register and download is free. Than you can use GIS or ENVI software 

Hi and thanks every one. :)

Dear Luca,

First of all thanks for your help, in fact I was looking for some readily available product for root zone soil moisture but as Mateusz mentioned it seems that there is no such data for year 2010 and before, so now I am reading your paper and trying to apply the simplified model to a surface soil moisture product (probably SMOS)

Dear Levan

thanks for your answer, can u explain it more?

Dear Mateusz

thank you very much, actually I first try to use H-SAF, but it's Archive root zone data starts from 2012, I couldn't get the ASCAT 2010 data from ECMWF either. in the case of the GLDAS I am not sure whether it is my fault or the model's, so I am reconsidering the results to find the possible error.

How about ERA-5 volumetric soil water data, which include four layers.

Ehsan, Hi

Root zone soil moisture can actually be higher than FC. But if you are saying that GLDAS values are higher than 1 (if it is relative saturation) or more than soil porosity (~50%), then one thing that you wanna do is to just convert any values over the Field Capacity to 1. Models tend to generate unreal resutls, but it's OK. It happens mostly because the soil medium is spatially very heterogeneous so it is very hard to match the real in-situ parameters to the satellite scale and vice versa.