This repository contains soil moisture, vegetation optical depth (VOD), and single-scattering albedo retrievals from the SMAP Multi-Temporal Dual-Channel Algorithm. 

The algorithm is described in the following paper:
Konings, A.G., M. Piles, N. Das, and D. Entekhabi (2017). L-band vegetation optical depth and effective scattering albedo estimation from SMAP. Remote Sensing of Environment, 198:460-470
https://doi.org/10.1016/j.rse.2017.06.037

And is based on principles explained in more detail in a paper describing the original algorithm development using Aquarius observations

Konings, A.G.*, M. Piles*, K. Rötzer, K.A. McColl, S. Chan, and D. Entekhabi (2016). Vegetation optical depth and scattering albedo retrieval using time-series of dual-polarized L-band radiometer observations. Remote Sensing of Environment. 172, 178-189. 
https://doi.org/10.1016/j.rse.2015.11.009

See also the related publication

Konings, A.G., K.A. McColl, M. Piles and D. Entekhabi (2015): How many parameters can be maximally estimated from a set of measurements? IEEE Geoscience and Remote Sensing Letters, 12(5), 1081-1085.
https://doi.org/10.1109/LGRS.2014.2381641

In short, the algorithmic approach uses both horizontally and vertically polarized brightness temperatures to retrieve soil moisture and VOD simultaneously. The key innovation of the MT-DCA is that it recognizes that classical
dual-channel algorithms are under-determined. Compensating errors between just two unknowns when there are only two correlated observations can add significant errors to retrieval (as illustrated in Konings et al, RSE 2016). 
This adds high-frequency noise to commonly used snapshot dual-channel algorithms. The MT-DCA is based on the recognition that VOD is expected to change more slowly than soil moisture, and combines multiple observation times
to reduce the ratio of the number of unknowns. A moving window (in the SMAP case, of two overpasses approximately 3-days apart) is used. Over each moving window, the soil moisture at both overpasses is retrieved, along with
a constant VOD over the window. This leads to two estimates of each of soil moisture and VOD at any given pixel-time: one where that time was at the beginning of the window, and one where it was at the end. The VOD and soil 
moisture values included here are the averages over those two windows. 

A second key innovation of the MT-DCA is that, because the retrievals are no longer under (or just barely) determined, it is also possible to retrieve a constant single-scattering albedo for each pixel, in contrast to other low-frequency 
retrieval algorithms that assume either a globally uniform value or a value that is uniform per land cover type. The retrieved albedo is also included here. 

Data Formatting

Files are provided of three types:
a) static files with lat and lon of each grid cell (latMTDCA.bin and lonMTDCA.bin)
b) daily snapshots of soil moisture or VOD (of the format mvMTDCAYYYYDDD.bin and vodMTDCAYYYYDDD.bin, respectively)
c) a time-independent single-scattering albedo map (omegaMTDCA.bin)


All data are saved in binary format in column-major order (first the first column, then the entire second column, and so on) as 64-bit floating point numbers in little-endian format. Each file contains a 1624 by 3856 entry map
that represents either a static variable (lat, lon, and albedo) or daily snapshot (VOD and soil moisture).

Retrievals are obtained from the 9km enhanced-resolution brightness temperatures from SMAP obtained using Backus-Gilbert regularization. As such, they are on the SMAP 9km EASE2-grid. The latitude and longitude of the center of 
each grid cell are included in the files latMTDCA.bin and lonMTDCA.bin. Additional information and geolocation tools are available at https://nsidc.org/data/ease/ease_grid2.html

File titles are of the format 'mvMTDCAXXXXYYY.bin', where XXXX is the 4-digit year and YYY is the 3-digit day of year. 

A Note on Retrievals Across Time

The Konings et al RSE 2017 paper describes only the first year of SMAP data, spanning April 1st, 2015 to March 31st, 2016. All data in this directory match those used in that paper. For data beyond this period, a slight tweak was made.
The albedo optimization in the MT-DCA is made over the entire record. This means that all retrievals change every time the record is extended, which is a somewhat undesirable situation from a software stability situation (though it has the 
advantage of continually refining estimates of optimal average albedo). Here, an alternative choice was made. Starting on April 1st, 2016 and beyond, the pixel-specific value of albedo is set as fixed based on the results from the 
retrieval in Year 1. The soil moisture and VOD are then retrieved based on a moving window as in the rest of the MT-DCA retrievals, but without the additional albedo optimization loop. 

Questions?

Email konings@stanford.edu