March/April, 1996
The question of the feasibility of extracting soil moisture information from satellite data came up in discussing monitoring requirements for the Global Climate Observing System (GCOS). Given the uncertainties in this field, it was suggested that an initial step should be to produce a consensus on the state-of-the-art and likely future prospects.
As a step in this direction, a list of statements/propositions was prepared attempting to capture what we are reasonably certain of; what we suspect/have fair evidence on; and what we do not know, with some suggested next steps. This statement was sent to leading active researchers or research programme managers in this field, including:
Evert Attema, ESA/ESTEC; Craig Dobson, University of Michigan; Ted Engman, NASA; Thomas Jackson, USDA; Terry Pultz, CCRS; Jon Ranson, NASA; Sasan Saatchi, JPL; Soroosh Sorooshian, University of Arizona; Thomas Schmugge, USDA; and Ying Ming Wei, NASA.
The six responses received were sorted out with minimal editing according to the individual questions, and reproduced below.
1. What we know:
1.1 Microwave techniques are the only remote sensing method for assured, frequent data acquisition on which soil moisture estimation from space could be based (others suffer from atmospheric interference in an unacceptable way).
1: Agree;
2: Agree;
3: Agree;
4: True, thermal infrared can give useful information but the atmosphere provided too many interruptions;
5: Uncertain - they are the most operational methods certainly;
6: Agree.
1.2 Any microwave technique practicable from space platforms will provide information on soil moisture only for depths 0-5 cm or so (because of the dielectric contrast between the air and soil, and the ionospheric interference at much lower wavelengths).
1: Agree;
2: Agree;
3: Using direct observations;
4: It is not believed that direct observations are obtained of 0-5 cm at C-Band. It is necessary to go to longer wavelengths, e.g., L-Band to get this;
5: Agree;
6: Agree. Deeper depths may be modelled.
1.3 The soil moisture signal is contaminated by surface soil roughness and vegetation (live or dead biomass).
1: Agree;
2: Agree;
3: Agree;
4: Roughness is more of a factor for radar and vegetation effect is stronger for radiometers;
5: Doesn’t this depend on active or passive method? How about incidence angle and topography?
6: Agree.
1.4 Vegetation moisture content, total biomass and canopy structure are the key parameters affecting the degree of soil signal contamination.
1: Agree;
2: Agree;
3: Agree;
4: For radiometric observations moisture content is the dominant factor;
5: See above - what about roughness and incidence angle?
6: Agree.
1.5 Estimation of relative changes in soil moisture between subsequent measurements is more easily achievable than that of absolute moisture content (provided that surface soil moisture is the only variable changed).
1: Agree;
2: Agree;
3: This is not true for passive; for active (especially current satellites) it is agreed;
4: True;
5: How could one disagree?
6: Agree.
2. What we suspect/think:
2.1 Multipolarization/multifrequency radar data will be required to obtain satisfactory soil moisture information for various ecosystems/phenological conditions.
1: Agree;
2: Agree;
3: Is it assumed that radar is the best solution? For active approaches, it is agreed;
4: Not believed to be true. An L-Band radiometer has been shown to be very effective for soil moisture in a wider range of conditions than the synthetic aperture radar (SAR) and is much simpler to interpret and can provide the data at a spatial scale relevant to climate change studies;
5: Agree. Soil moisture cannot be sensed below wet canopies so don’t even bother this is a method for semi-arid and arid regions where soil moisture varies with time;
6: True if one requires a very robust approach for various ecosystems/phenological conditions. Single Channel SAR, such as Radarsat, may be sufficient in some environments.
2.2 Semi-empirical or empirical models must be used to extract soil moisture from microwave data.
1: Agree;
2: Most probably true, though perhaps sometime in the future we can find the ‘Holy Grail’ of soil moisture through an elegant (but simple) inversion;
3: Not true, we have only a limited understanding and incomplete databases;
4: The basic dielectric models have given reasonable agreement with radiometric observations in HAPEX-Sahel, Washita 92 etc., so the level of empiricism is low for radiometric observations;
5: Agree - without fully specifying the physical properties of the land cover, we must improvise;
6: Agree.
2.3 The influence of vegetation with soil signals may be accounted for by a single parameter (related to optical thickness of the canopy).
1: Agreed;
2: Quite possible;
3: For passive (long wavelengths): yes. For active: some structural information is needed;
4: True;
5: Not known;
6: Unlikely.
2.4 DEM data with sufficient x/y and z resolution will be required if radar data are used for soil moisture estimation.
1: Agree;
2: Agree and suspect that 1:100,000 hypsography is adequate for most places;
3: There is some doubt that for satellite scale footprints and larger look angles that this is true;
4: Not necessary for the radiometer;
5: Agree - see 1.3;
6: Agree. Canadian Prairies may be an exception (flat).
3. What we do not know:
3.1 We do not know if soil moisture information for the top 5 cm of soil would be useful for climate and climate change purposes (can be determined by modellers); and if so, whether remote sensing offers the best means of obtaining this information.
1: Agree;
2: Agree;
3: There is enough evidence in the literature to support the utility of this observation. The most obvious applications are the least demanding in terms of spatial resolution;
4: The repetitive observations of the moisture state for the upper boundary layer of the soil has to be useful. For one thing it can give a clear indication of regions of recent rainfall and a qualitative estimate of how much;
5: Agree, but move to Section 2;
6: Agree.
3.2 We do not know if relative change information would be useful for climate and climate change purposes (can be determined by modellers).
1: Agree;
2: Agree;
3: It would be useful. Climate modellers need a direct observation of state and/or flux variables. They need to reduce the degrees of freedom, not increase the uncertainty;
4: See above;
5: Agree;
6: It is understood that modellers are interested in these data. However, this information comes to me second-hand and through the literature.
3.3 We do not know if soil moisture signals can be isolated and quantified with sufficient accuracy under the variety of vegetation cover on a global basis and for various phases of the phenological cycle (requires much experimentation, likely further model development and validation, operational trials).
1: By averaging over a large number of pixels (say to a resolution of 500 x 500 meters) the investigators could get high correlation of ERS-1 SAR data with soil moisture (accumulated rain) by averaging over the interfering roughness and canopy variations. This is clearly dependent on land use and will be a ‘regional’ algorithm, not global;
2: Agree;
3: Agree;
4: More studies are needed. This is what they say when they don’t want to do anything. Radiometric observations with the PBMR and ESTAR over the past ten years have demonstrated its capabilities over a wide range of conditions from FIFE to HAPEX-Sahel;
5: This is what we know (soil moisture cannot be sensed below wet canopies so don’t even bother - this is a method for semi-arid and arid regions where soil moisture varies with time);
6: Agree.
3.4 We do not know which technique (active or passive) is preferable for global climate applications (could be answered through a dialogue between earth observation and climate scientists but we may need to know more about the performance of each technique).
1: Agree;
2: Agree - also need to know the requirements of each application;
3: Depending upon the application objectives (spatial/temporal) the choice of a system is quite clear;
4: As was pointed out, the radiometer works best for soil moisture where it is needed most, i.e., in sparse vegetation conditions where direct evaporation from the soil is a major factor in the surface energy balance;
5: Active offers the best signal - but at multiple costs;
6: Agree;
3.5 We do not know which ancillary data would be required for operational inversion of the microwave data to obtain soil moisture information and whether these can be obtained in practice (requires better understanding of the two methods and their respective data needs).
1: Agree;
2: Agree;
3: 0;
4: I think Jackson et al recent papers (Remote Sensing of Environment 53: 27-37; IEEE Transactions on Geoscience and Remote Sensing 31 (4): 836-841; Hydrological Processes 7: 139-152) have laid out an algorithm which indicates the types of ancillary data that are needed;
5. Again, we know what the main parameters are but sensitivity is another issue;
6: Agree.
4. Conclusions
4.1 The usefulness of near-surface soil information for climate monitoring purposes and the required spatial resolution should be established (as absolute quantities or in relative units).
1: Agree;
2: Agree. This should be a focus for some of the modelling efforts. Sensitivity studies could also be done that feed into 4.2 below;
3: There are ongoing activities;
4: The spatial resolution should indicate the major changes in soil moisture, i.e., the rainfall variations. The studies we had done in the late 1970’s showed that most of the variation could be accounted with resolutions in the order of 5 to 10 km;
5: Suggested action: support modellers;
6: Agree, is this being addressed in GEWEX?
4.2 Based on the requirements, a comprehensive evaluation of an end-to-end system (raw data to global data sets) should be undertaken, and critical elements identified.
1: Agree;
2: Agree. Effort should include a complete error analysis for each step (this would also be a good mechanism for comparison of competing techniques);
3: There are ongoing activities;
4: For the radiometer with 5 to 10 km, resolution this would seem to be a trivial problem;
5: Partially agree;
6: Agree.
4.3 Sponsorship for further work on the critical elements should be sought through coordinated national efforts.
1: Agree;
2: Agree. This could be done within existing programmes, but an emphasis really needs to be placed upon ‘coordinated’ (not fragmented) efforts on critical parts of 1 (or more) possible end-to-end systems;
3: Agree;
4: Push NASA to fly an L-Band radiometer;
5: Agree;
6: ?
