SHARE: Soil Moisture for Hydrometeorological Applications

Doubková Marcela, Bartsch Annett, Wolfgang Wagner,
Institute of Photogrammery and Remote Sensing, TU WIEN, mdo@ipf.tuwien.ac.at

Introduction

Soil moisture represents a switch that controls the proportion of rainfall that percolates, runs off, or evaporates from land. Since the 1970s, a variety of coarse resolution soil moisture datasets have become available from active and passive microwave systems (i.e. ERS-1/2, METOP ASCAT, AMSR-E and SMOS) at coarse (>25km) spatial resolution. These have been applied to improve flood forecasting, numerical weather predictions and rainfall estimates as well as to study soil moisture trends and anomalies in relation to climate change [1–4].

While of excellent radiometric accuracy, the coarse spatial resolution datasets remained a constraint for data users operating at medium scale (<1km). It became obvious that applications such as coupled crop-climate modeling or soil moisture monitoring over heterogeneous landscapes or river runoff prediction at sub-basins scale may benefit the establishment of medium resolution (<1km) soil moisture dataset [5–7]. SHARE (Soil moisture for hydrometeorologic applications), the ESA’s DUE Tiger Initiative project, answered the need of hydrological and agricultural community for improved Earth Observation products by providing medium resolution (1 km) soil moisture service derived from the Advanced Synthetic Aperture Radar (ASAR) onboard ENVISAT [9]. Since its start in 2005 the SHARE service extended over Australian and portions of African and South American continent.

The algorithm for the ASAR Global Mode (GM) soil moisture product has been adopted from the already existing change detection algorithm for the ERS-1/2 scatterometer [8]. The basic idea behind the change detection is that the backscatter cross section of natural surfaces changes over short timescales mainly due to variations in soil moisture, while vegetation or surface roughness are assumed to be constant or only slowly varying [9]. It should be noted that the ASAR GM soil moisture product is an index scaled between 0 (dry conditions) and 1 (saturated conditions) and its conversion to absolute values may be required.

The SHARE project demonstrated in two important ways how data from medium resolution microwave instruments can be used to support flood monitoring efforts. Firstly, the data can continuously monitor how much water is stored in the soil (Figure 1) and thus determine the amount of runoff resulting from rain. Secondly, the data can support monitoring of inundated areas during a flood due to its capabilities to penetrate clouds and even rain.

Five maps showing The ASAR GM relative soil moisture product (monthly mean) over Victoria in Feb. 2007-2011.

Figure 1. The ASAR GM relative soil moisture product (monthly mean) over Victoria in Feb. 2007-2011.

Importantly, given the similar characteristics of the ASAR GM and the future Sentinel-1 sensor it is anticipated that the ASAR GM algorithm can be transformed to a potential soil moisture product retrieved from Sentinel-1.

Toward operational products

The development of operational water monitoring services is progressing rapidly. The requirement on the operationality is thus becoming a standard also for the Earth observation products. The ASAR GM soil moisture product is available semi-operationally, in other words it is automatically processed on a monthly basis. The development of the product includes algorithm development, data processing, data validation, algorithm improvement, automatization of the data processing and delivery, capacity building and support to data users. A fully automatic processing chain has been generated at the TU WIEN that reprocesses the ASAR Level 1 data into soil moisture Level 3 datasets and make previews available via the ASAR GM data viewer with a 1 month delay. The potential minimum delay is however only several hours and compares to the latency of the near-real-time coarse resolution soil moisture datasets from the SMOS, ASCAT and AMSR-E sensors. The ASAR GM georeferenced soil moisture product is available on request at no coast at the institute website.

The ASAR GM soil moisture product development and validation was in detail summarized elsewhere [9–11]. The latter works demonstrated a good potential of ASAR C-band observations to monitor variations in soil moisture on a quasi-operational basis. Additional works demonstrated a good agreement of ASAR GM soil moisture and the soil moisture output from an independent AWRA-L landscape hydrological model developed within the Australian Water Resources Assessment system (AWRA) [11], [12] over the Australian continent (Figure 2, left). Further, the observational error of the ASAR GM dataset was evaluated [11] (Figure 2, right) using the independent estimates from the AWRA-L model. The error estimates were less (25%) for forested areas and areas covered with rock outcrops in western, northern, and eastern coastal Australia. The percentage represents the maximum relative soil moisture that can be accounted by the ASAR GM error. The good understanding of the error together with the knowledge of the relationship between remotely-sensed and model variables are critical for a successful application of the product [13].

Two maps showing The simple Pearson correlation coefficient between ASAR GM and AWRA-L soil moisture (left). AWRA-L is a landscape hydrology model that explicitly models soil surface moisture dynamics. The ASAR GM error (right) is estimated by propagating sensor error through the ASAR GM retrieval algorithm.

Figure 2. The simple Pearson correlation coefficient between ASAR GM and AWRA-L soil moisture (left). AWRA-L is a landscape hydrology model that explicitly models soil surface moisture dynamics. The ASAR GM error (right) is estimated by propagating sensor error through the ASAR GM retrieval algorithm.

Demonstrated and planned applications in hydrology

The major applications of the ASAR GM product are expected in hydrology and water management. While the added value of the coarse resolution soil moisture datasets in hydrological models have been demonstrated [3], [6] similar investigation with medium resolution ASAR GM data has began only recently. The preliminary studies demonstrated the potential of the ASAR GM data to identify saturated surfaces (Figure 1) [10], [14] which to a large extent contribute to surface runoff [15]. The ASAR GM data were also implemented to identify bias in precipitation datasets [16] and could resolve spatial patterns not observable in the ERS scatterometer measurements [14].

Further investigations are performed within the scope of the SHARE project; supported by combined efforts of TU WIEN (Vienna University of Technology) and CSIRO (Commonwealth Scientific and Industrial Research Organisation). CSIRO identified remote sensing datasets as crucial for the hydrological observation system (AWRA); this will soon become operational through the Bureau of Meteorology. Preliminary assessments have suggested potential of ASAR GM soil moisture to:

a) Characterise the relative errors of AWRA-L (AWRA landscape hydrological model) soil moisture (Figure 2);
b) Serve as an independent dataset for a multi-objective calibration of the AWRA-L model parameters;
c) Serve as an independent member for a generation of a blended soil moisture product at 5 km spatial resolution;
d) Support monitoring of large scale inundation events.

  An example of the ASAR Wide Swath (WS) and Image Mode (IM) normalized backscatter images over Eastern Queensland, Australia, during dry (April 2010) and wet season (January, 2011). The inundated areas are shown as orange and red, characterized by very low backscatter values owing to the specular reflection of the radar signal on the flooded surface. The dark blue colors, which are a result of the high backscatter values, show inundated vegetation that reflects the signal back to the sensor (double bounce effect).

Figure 3. An example of the ASAR Wide Swath (WS) and Image Mode (IM) normalized backscatter images over Eastern Queensland, Australia, during dry (April 2010) and wet season (January, 2011). The inundated areas are shown as orange and red, characterized by very low backscatter values owing to the specular reflection of the radar signal on the flooded surface. The dark blue colors, which are a result of the high backscatter values, show inundated vegetation that reflects the signal back to the sensor (double bounce effect).

As this work is ongoing and will continue beyond the duration of the SHARE project only first findings are here summarized.

Several different ways of merging observations within the model-data system require different computational overheads. Blended dataset can be used as a stand-alone product for wide range of implications (i.e. agricultural decision making, drought detection). Also, it can be directly assimilated into a model rather than assimilating several datasets with independent error structures and often different spatial resolutions.

While the ability of the ASAR data from higher resolution modes (Wide Swath (WS) and Image Mode (IM) with 150m spatial resolution) to monitor large scale inundation events is evident (Figure 3), a generic classification approach applicable also on the ASAR GM data is under investigation [17–19]. Within the SHARE project a method [20] for inundation extent mapping using the ASAR GM data was developed that uses a thresholding approach to distinguish flooded and non-flooded areas and combines this with the MODIS Open Water Likelihood (OWL) index to retrieve water proportion within each ASAR GM pixel. The method demonstrated the ability of the ASAR GM data to detect open water bodies as well as water under vegetated areas (Figure 4).

Nevertheless, an overclassification of flooded regions was evident that occurred over areas where wet soil got mistaken with flooded vegetation (southeastern corner of Figure 4). Also, the total proportion of water within each pixel differed substantially between the ASAR GM and MODIS algorithms. The latter may be caused by the low spatial resolution of the ASAR GM data that provides mixed signature of flooded and non-flooded regions resulting in a mid-range backscatter; these may be consequently classified as only partly flooded. On the contrary, several irrigated areas were correctly detected by the ASAR GM data that could not be detected using the MODIS OWL index. These results suggest the synergistic combination of several remote sensing methods as the best approach for the characterization of inundation events.

The MODIS band composite (7, 2, 1) (left), the ASAR GM water map (percentage of water within each pixel) (center) and the corresponding MODIS water map (right) in January 2011.

Figure 4. The MODIS band composite (7, 2, 1) (left), the ASAR GM water map (percentage of water within each pixel) (center) and the corresponding MODIS water map (right) in January 2011.

Data users

It was an explicit aim of SHARE to get the widest possible user community actively involved. Two prime users were identified – the University of Kwazulu Natal (UKZN) and the Australian Commonwealth Scientific and Research Organization (CSIRO). These also acted as a bridgehead to the user community in Australia and Africa.

A data request form has been setup on the SHARE website. Since beginning of the project (December 2005) there have been more than 80 data requests that originated mostly in the African and European continent. The recently published journal papers and the representation of the SHARE project on international meetings raised the awareness on the product also by users from the USA, Australia, and variety of international organizations (Figure 5, right).

Figures showing The number of users and their proposed application of the ASAR GM soil moisture product (left). The origin of the ESA DUE SHARE project data users (right). The numbers represent the number of users

Figure 5. The number of users and their proposed application of the ASAR GM soil moisture product (left). The origin of the ESA DUE SHARE project data users (right). The numbers represent the number of users.

The application of the ASAR GM soil moisture parameter in variety of applied studies has been investigated (Figure 5, left). These range from crop yield estimates, runoff prediction [15] to climate variability studies. A number of comparison and validation studies with in-situ [9], modelled [12] and remote sensing datasets [21] has also been performed.

Final remarks

A continuation of research satellite missions and data service availability on operational bases is needed for successful and meaningful integration of the Earth Observation data into existing models. While the ENVISAT is slowly approaching its end a successive satellite mission – Sentinel – is foreseen to be operated over the period 2013 to 2030 that will provide data at improved spatial, temporal and radiometric resolution.

The results of the SHARE project have well prepared the ground for the future Sentinel SAR sensors by demonstrating the viability of the soil moisture and inundation extent retrieval. The future operationally available medium resolution soil moisture and inundation extent estimates from Sentinel-1 have the potential to be of a great benefit for crop growth and water balance monitoring and modeling in next decades.

[1] Y. Y. Liu, A. I. J. M. Van Dijk, R. A. M. De Jeu, and T. R. H. Holmes, “An analysis of spatiotemporal variations of soil and vegetation moisture from a 29-year satellite-derived data set over mainland Australia,” Water Resources Research, vol. 45, no. 7, p. art. no. W07405, 2009.

[2] W. T. Crow, G. J. Huffman, R. Bindlish, and T. J. Jackson, “Improving satellite-based rainfall accumulation estimates using spaceborne surface soil moisture retrievals,” Journal of Hydrometeorology, vol. 10, no. 1, pp. 199-212, 2009.

[3] L. Brocca et al., “Improving runoff prediction through the assimilation of the ASCAT soil moisture product,” Hydrology and Earth System Sciences Discussions, vol. 7 (4), no. 4, pp. 4113-4144, 2010.

[4] M. Drusch, “Initializing numerical weather prediction models with satellite-derived surface soil moisture: Data assimilation experiments with ECMWF’s Integrated Forecast System and the TMI soil moisture data set,” Journal of Geophysical Research, vol. 112, no. 3, pp. 1-14, 2007.

[5] T. Osborne, J. Slingo, D. Lawrence, and T. Wheeler, “Examining the interaction of growing crops with local climate using a coupled crop-climate model,” Journal of Climate, vol. 22, no. 6, pp. 1393-1411, 2009.

[6] J. Parajka, V. Naeimi, G. Blöschl, and J. Komma, “Matching ERS scatterometer based soil moisture patterns with simulations of a conceptual dual layer hydrologic model over Austria,” Hydrology and Earth System Sciences, vol. 13, no. 2, pp. 259-271, 2009.

[7] P. Meier, A. Frömelt, and W. Kinzelbach, “Hydrological real-time modeling using remote sensing data,” Hydrology and Earth System Sciences Discussions, vol. 7, no. 6, pp. 8809-8835, 2010.

[8] W. Wagner, G. Lemoine, and H. Rott, “A Method for Estimating Soil Moisture from ERS Scatterometer and Soil Data,” Remote Sensing of Environment, vol. 70, no. 2, pp. 191-207, 1999.

[9] C. Pathe, W. Wagner, D. Sabel, M. Doubkova, and J. Basara, “Using ENVISAT ASAR Global Mode Data for Surface Soil Moisture Retrieval Over Oklahoma, USA,” IEEE Transactions on Geoscience and Remote Sensing, vol. 47, no. 2, pp. 468-480, 2009.

[10] I. Mladenova, V. Lakshmi, J. P. Walker, R. Panciera, W. Wagner, and M. Doubkova, “Validation of the ASAR global monitoring mode soil moisture product using the NAFE’05 data set,” IEEE Transactions on Geoscience and Remote Sensing, vol. 48, no. 6, pp. 2498-2508, 2010.

[11] M. Doubková, A. I. J. M. Van Dijk, G. Blöschl, D. Sabel, and W. Wagner, “Evaluation of predicted soil moisture retrieval error from C-Band SAR by comparison against soil moisture estimates over Australia,” Remote Sensing of Environment, 2011.

[12] A. I. J. M. Van Dijk and G. A. Warren, “AWRA Technical Report 4. Evaluation Against Observations.,” WIRADA/CSIRO Water for a Healthy Country Flagship, Canberra, 2010.

[13] A. I. J. M. van Dijk and L. J. Renzullo, “Water resource monitoring systems and the role of satellite observations,” Hydrology and Earth System Sciences, vol. 15, no. 1, pp. 39-55, Jan. 2011.

[14] C. Pathe, W. Wagner, D. Sabel, Z. Bartalis, M. Doubkova, and V. Naeimi, “Scatterometer and ScanSAR soil moisture observations of the contiguous United States,” in Proceedings of the IEEE National Radar Conference, IEEE National Radar Conference, 2009.

[15] A. Bartsch, M. Doubkova, C. Pathe, D. Sabel, P. Wolski, and W. Wagner, “River flow & wetland monitoring with ENVISAT ASAR Global mode in the Okavango Basin and Delta,” Proceedings of the Second IASTED Africa Conference, Water Resource Management (AfricaWRM 2008). Gaborone, Botswana, 8-10 September, 2008, pp. 152-156, 2008.

[16] C. Milzow, P. E. Krogh, and P. Bauer-Gottwein, “Combining satellite radar altimetry, SAR surface soil moisture and GRACE total storage changes for model calibration and validation in a large ungauged catchment,” Hydrology and Earth System Sciences Discussions, vol. 7, no. 6, pp. 9123-9154, 2010.

[17] S. Martinis, a. Twele, and S. Voigt, “Towards operational near real-time flood detection using a split-based automatic thresholding procedure on high resolution TerraSAR-X data,” Natural Hazards and Earth System Science, vol. 9, no. 2, pp. 303-314, Mar. 2009.

[18] P. Matgen et al., “Towards the sequential assimilation of SAR-derived water stages into hydraulic models using the particle Filter: Proof of concept,” Hydrology and Earth System Sciences Discussions, vol. 7, no. 2, pp. 1785-1819, 2010.

[19] D. O’Grady, M. Leblanc, and D. Gillieson, “Use of ENVISAT ASAR Global Monitoring Mode to complement optical data in the mapping of rapid broad-scale flooding in Pakistan,” Hydrology and Earth System Sciences Discussions, vol. 8, no. 3, pp. 5769-5809, Jun. 2011.

[20] C. J. Ticehurst, A. Bartsch, M. Doubkova, and A. I. J. M. van Dijk, “Comparison of ENVISAT ASAR GM, AMSR-E passive microwave, and MODIS optical remote sensing for flood monitoring in Australia,” in Earth Observation and Water Cycle Science Symposium, 2009, vol. ESA Specia, p. 8.

[21] D. Sabel et al., “Synergistic use of Scatterometer and ScanSAR Data for Extraction of Surface Soil Moisture Information in Australia,” in EUMETSAT Meteorological Satellite Conference, 8-12 September, 2008, Darmstadt, Germany, 2008.