Comm. Appl. Biol. Sci, Ghent University, 70/2, 2005 1 SOIL MOISTURE RETRIEVAL FROM OPTICAL AND THERMAL SPACEBORNE REMOTE SENSING W.W. VERSTRAETEN 1,2 ; F. VEROUSTRAETE 2 ; J. FEYEN 1 1 Laboratory of Soil and Water Management, Katholieke Universiteit Leuven, Vital Decosterstraat 102, B-3000 Leuven; 2 Remote Sensing and Earth Observation Processes, VITO, Boeretang 200, B-2400 Mol BACKGROUND & OBJECTIVES Soil moisture is the water held in the soil within reach of plant roots and is one of the most important land environmental variables in perspective with land surface climatology, hydrology, and ecology. The soil moisture content (SMC) generally refers to the water contained in the upper 1-2 m of soil, which can potentially evaporate into the atmosphere. Variations in SMC have strong impacts on changes in the surface energy balance, regional runoff, and vegetation productivity (crop yield potential). SMC conditions may also serve as a warning for flooding. In areas of active deforestation, SMC estimates help to predict run-off as well as soil erosion. Despite the importance of SMC, the main problem is that the gravimetric method (standard procedure of SMC deteration), on which all other methods are ultimately calibrated, is essentially a point measurement. Local scale variations in soil properties, terrain, and vegetation cover make selection of representative field sites difficult if not impossible. Contrarily, remote sensing (RS) techniques may be compromising because of the spatial ability and the relatively low cost (Wagner et al., 1999). Zazueta and Xin (1994) classify well known general measurement techniques to detere soil moisture content: (i) gravimetric techniques, (ii) nuclear (neutron scattering, gamma attenuation, nuclear magnetic resonance), (iii) electromagnetic (resistive and capacitive sensors, time and frequency domain reflectometer), (iv) tensiometric, (v) hydrometric, (vi) remote sensing (passive and active microwave, thermal infrared) and (vii) optical techniques (polarized light, fibre optic sensors, near-infrared). A more basic field method (ix) is the Feel and Appearance method using a soil moisture interpretation chart based on texture classification and squeezing of soil samples (Miles, 1998). We demonstrate that optical and thermal information from an existing operational spaceborne sensor (METEOSAT) can be used to detere SMC on 10-daily intervals without the need of excessive ancillary data or direct coupling with SVAT models. We present the methodology, the product validation on 2 EUROFLUX sites (Valentini et al., 2000) and a comparison with active microwave RS using the ERS Scatterometer (Wagner et al., 1999). METHODS & DATASETS Similar to the Integral NOAA/AVHRR-imagery processing Chain (Verstraeten et al. 2004, 2005) the Integral METEOSAT-imagery processing Chain (ime- TEOSAT-Chain) was build to produce spatial information for the detection of water limiting conditions and hydrological variables such as evaporation and SMC in an operational mode in the framework of plant carbon relations
2 (Veroustraete et al., 2002; Verstraeten et al., submitted). With this ime- TEOSAT-Chain SMC is estimated combining visible (surface albedo, α 0 ) and thermal (day and night land surface temperatures, LST0,d, LST0,n) satellite information on the regional scale with the hydrometric thermal inertia (TI) method (datasets: March-October 1997 of Europe and EUROFLUX data). Figure 1 offers an example of the electromagnetic (EM) spectrum. Figure 1. The electromagnetic spectrum. The visible info is between 400 and 700 nm and the thermal (infrared) info is around 10-12 µm. TI is a body property of materials describing their resistance to temperature variations. Essentially, a change detection technique is applied in analogy with the ERS Scatterometer derived Soil Water Index (SWI). We opted for the approach of Mitra and Majumdar (2004) using the apparent thermal inertia (ATI) which can directly be inferred from remote sensing imagery. Equations (1) to (4) are used to derive SMC. The temporal and thermal resolution of METEOSAT and ERS Scatterometer imagery are 30 and ± 20 km², 3-4 days and 2500 km², respectively. 1 α0 ATI = C (1); ATI( t) ATI SMSI 0( t) = (2); ΔLST ATI ATI 0 max SMSI ( t) = t SMSI ( t ) e t 0 e i t ti T t ti T (3); θ ( t ) = SMSI ( t) ( θ max θ ) + θ (4) In Equations (1) to (4) α0 is the (hemispherical) surface albedo, LST0 is the day and night land surface temperature difference, C is a correction parameter and function of latitude and the solar declination; ATI, ATImax and ATI(t) are respectively the imum, maximum and apparent thermal inertia index at time t; SMSI0(t) and SMSI(t) are the Soil Moisture Saturation Index at the surface soil layer and for the 1 m soil profile using a characteristic time length T; θ(t), θ and θmax are the SMC at time t and the imum
Comm. Appl. Biol. Sci, Ghent University, 70/2, 2005 3 and maximum SMC based on soil hydrological properties of saturation, field capacity and wilting point. First the ATI is calculated for each pixel of the METEOSAT image. Then for each pixel a time profile is constructed and the imum and maximum ATI is selected. These values are combined in the SMSI0 of Equation (2). To convert the SMSI for the surface layer to the 1 m soil profile a Markov type of temporal (red noise) filter is applied based on the autocorrelation function of the water balance equation. Finally, the retrieved SMSI(t) from Equation (3) is combined with soil physical information to obtain SMC. RESULTS & DISCUSSION For the Vielsalm, and Brasschaat EUROFLUX sites the scatterograms of the 10 daily SMC values derived from the thermal inertia approach (SMC-TI) against the EUROFLUX measurements (SMC-EUROFLUX) are given in Figure 2 (March-Oct 1997). The latlon coordinates, the elevation above sea level, the doant tree species and the validation statistics (slope, intercept and R² of the linear fit, and the root mean square error, RMSE) are also shown. The temporal profiles of the SMC-TI and SMC-EUROFLUX are given in Figure 3. In this figure also the SWI-ERS Scatterometer and SMSI-METEOSAT temporal profiles are depicted. The RMSE for the SWI against SMSI is 0.048 and 0.094 for the Vielsalm and Brasschaat site respectively. From Figures 2 and 3 it is observed that the SMC-TI fits satisfactorily with the SMC observations for the Vielsalm and Brasschaat case with a maximum R² value of 0.88. From the temporal SWI and SMSI profiles we can observe some similarity in trends but they are indubitably not that optimal then observed in the SMC-TI versus SMC-EUROFLUX case. The slope, intercept and R² for the Vielsalm site is 1.078, -0.037 and 0.71 and for the Brasschaat site 0.742, 0.119 and 0.57. From this preliary analysis it may be concluded that the results of this methodology are promising. SMC-TI [cm 3.cm -3 ] 0.28 0.26 0.24 Vielsalm (Picea ) 450 m asl, 50 18'N 6 00'E RMSE = 0.01 cm 3.cm -3 0.22 y = 1.0285x - 0.0021 R 2 = 0.8745 0.22 0.24 0.26 0.28 SMC-EUROFLUX [cm 3.cm -3 ] SMC-TI [cm 3.cm -3 ] Brasschaat (Pinus ) 10 m asl, 51 18'N 4 31'E RMSE = 0.02 cm 3.cm -3 y = 1.1166x - 0.0292 R 2 = 0.7077 SMC-EUROFLUX [cm 3.cm -3 ] Figure 2. The scatterogram of the SMC-TI versus the SMC-EUROFLUX for Vielsalm (left) and Brasschaat (right) with the latlon coordinates, elevation
1 4 above sea level (asl), tree species and the validation statistics (slope, intercept, R², RMSE). 10 daily values between March and Oct. 1997 were used. SMC-EUROFLUX SMC-TI SWI ERS Scatterometer SMSI-METEOSAT 0.45 Vielsalm Brasschaat Soil moisture content [cm 3.cm -3 ] 0.35 Soil water indicator [-] 0.10 0.05 1 st March to 31 st October 1997 1 st March to 31 st October 1997 EUROFLUX site days Figure 3. The SMC time series from METEOSAT (SMC-TI) and from measurements (SMC-EUROFLUX) (large symbols); The time series of the SWI from the ERS Scatterometer and SMSI from METEOSAT imagery (small symbols). REFERENCES Miles, D.L. (1998). Estimating Soil Moisture. Crop series, irrigation, no. 4.700, Colorado State University Cooperative Extension, 9/98. Mitra, D.S., Majumdar, T.J. (2004). Thermal inertia mapping over the Brahmaputra basin, India using NOAA-AVHRR data and its possible geological applications. International Journal of Remote Sensing, 225(16), 3245-3260. Valentini, R. et al. (2000). Respiration as the main deterant of carbon balance in European forests. Nature, 404, 861-865. Veroustraete, F., Sabbe, H., Eerens H. (2002). Estimation of carbon mass fluxes over Europe using the C-Fix model and Euroflux data. Remote Sensing of Environment, 83, 376-399. Verstraeten, W.W., Veroustraete, F., Feyen, J. (2004). Estimating European forest evapotranspiration by processing NOAA/AVHRR and METEOSAT imagery. Communications in Agricultural and Applied Biological Sciences, Universiteit Gent, Vol 69(2), 333-336. Verstraeten, W.W., Veroustraete, F., Feyen, J. (2005). Estimating evapotranspiration of European forests from NOAA-imagery at satellite overpass time: Towards an operational processing chain for integrated optical and thermal sensor data products. Remote Sensing of Environment, 96(2), 256-276. Verstraeten, W.W., Veroustraete, F., Feyen, J. (2005). On temperature and water limitation in the estimation net ecosystem productivity: Implementation in the PEM C- Fix. Submitted to Ecological Modelling.
Comm. Appl. Biol. Sci, Ghent University, 70/2, 2005 5 Wagner, W., Lemoine, G., Rott, H. (1999). A Method for Estimating Soil Moisture from ERS Scatterometer and Soil Data. Remote Sensing of Environment, 70, 191-207. Zazueta, F.S., Xin, J. (1994). Soil Moisture Sensors. Bulletin 292, Florida Cooperative Extension Service, Institute of Food and Agricultural Sciences, University of Florida.