Skip navigation.
Home

Algorithm Description

Printer-friendly version

The merging algorithm

The ESA CCI SM algorithm generates consistent, quality-controlled, long-term (1978-2018) soil moisture climate data records by harmonising and merging soil moisture retrievals from multiple satellites into an active-microwave-based only (ACTIVE), a passive-microwave-based only (PASSIVE) and combined active-passive (COMBINED) product.

Since its first release in 2012 the ESA CCI SM merging algorithm has undergone substantial improvements. In essence, the merging algorithm is responsible for merging soil moisture retrievals from various satellites that have finite lifetimes and significantly different instrument characteristics (frequency, spatial resolution, temporal coverage, polarisation, revisit time etc) into three consistent multi-decadal data sets. This process faces innumerable scientific challenges and is therefore subject to continuous research and development. The merging scheme of the most recent version of the algorithm released at the end of 2018 (v04.4) is illustrated and described below.

 

Merging scheme of the ESA CCI SM v4 algorithm. From v03.2 (2015) onwards, soil moisture products from four active-microwave-based instruments and seven passive-microwave-based instruments are merged into ESA CCI SM data sets. For more information on the product generation please refer to Gruber et al. (2019).

1. Soil moisture product retrieval and gridding

All ESA CCI SM algorithms merge pre-processed level 2 (L2) data, that is, gridded soil moisture products retrieved from radiometrically calibrated backscatter or brightness temperature measurements. These are later resampled to daily global images on a 0.25o regular grid.

2. Level 2 data harmonisation into COMBINED product

All L2 input data sets are directly scaled against GLDAS Noah land surface model to harmonize their climatology. 

3. Uncertainty estimation

Uncertainties are estimated for each individual product and used to construct error covariance matrices for all merging periods depending on the sensor availability during these periods. Uncertainty estimates and hence merging weights largely follow VOD patterns.

4. Quality control

For the generation of both the PASSIVE and the COMBINED product, correlation significance levels are used to completely mask out individual L2 input products that are deemed unreliable at a particular location and during a particular merging period (where more than one product is available for merging).