SOIL MOISTURE INVESTIGATION FOR THE DIFFERENT AGRICULTURAL CROPS USING ERS-1 AND ERS-2 DATA

Abstract

Within the Project AO2.PL102 sponsored by ESA the following paper has been prepared. The data applied for the project include images obtained for Pilot Project PP.PL-4 for 1992-1994. SAR and ATSR-1 data are used for soil moisture assessment for different agricultural crops. For 1992-1996 growing seasons the various soil-vegetation ground measurements have been carried out at the time of ERS-1/2 overpasses. Also, survey was undertaken to record crop type with their actual development stage and crop condition. Meteorological parameters as air temperature, wind speed, solar and net radiation were also measured. Obtained SAR.PRI and ATSR-1 descending images have been rectified to Landsat TM georeferenced image. Backscattering coefficient and ATSR data (albedo and surface temperature) have been extracted for the chosen agricultural test site. SAR data were taken as an average value for the each whole plot as well as for measurement point (from block of 9x9 pixels). Also NOAA/AVHRR data have been registered closed to ERS-1/2 overpasses and processed to obtain NDVI and surface temperature values. ATSR and AVHRR/NOAA pixels covered nearly the same area on Landsat TM image. From SAR data average backscatter for the area of each NOAA and ATSR pixel has been extracted. Comparison of ground data with SAR data gave good results for soil moisture under separated surface vegetation roughness classes represented by LAI values. From AVHRR/NOAA and ATSR data soil moisture index WDI has been calculated. This index which reflects crop water condition has been compared to SAR backscatter values for the area of different crops and growing vegetation stages. The results are going to be presented at the 3rd ERS Symposium.

INTRODUCTION

The project has been carried out for the test site situated in South-West part of Poland in the Obra Valley. The considered area is covered by agricultural fields with the following dominant crops: wheat, rye, triticale, barley, oat, rape, maize, sugar beet and potatoes. The crop types for the considered fields varied from 1992 to 1996. The first ERS-1 SAR images for years 1992-1995 were acquired (towards the end of May) when winter crops were in their heading or flowering stage, spring crops were in their tillering or jointing stage, root crops and maize were in their emergence or germination stage and grass in the flowering stage before harvest. The ATSR-1 images were available for 1995. The predominate soil types for crops are sandy and sandy loam. The area of the test site is flat.

Throughout each of the growing seasons and simultaneously to ERS-1,2 descending overpasses the measurements of soil and vegetation parameters were carried out at 30 points chosen for different crops. The parameters measured were:

- volumetric soil moisture [%],

- wet and dry biomass [g/m²],

- gravimetric vegetation moisture [%],

- leaf area index (LAI),

- height of the vegetation [m].

These measurements were based on samples taken at the sites which represent the whole field. Also, surveys were undertaken to record crop type with their actual development stage and grass growing condition.

ERS-1.SAR.PRI data acquired during the 35-day repeat orbit were obtained from ESA for Project AO2.PL.102 and PP.PL-4. From each of the ERS-1.SAR images a sub-scenes corresponding to the area of the test site were geometrically transformed to the georeferenced Landsat TM image.

Backscatter coefficient values (sigma) for each of the ground truth points were calculated and averaged for blocks of pixels 9x9. The ground sample point was placed in the middle of each block. Also for the whole field the backscatter coefficient was calculated as well as for the area covered by each of AVHRR/NOAA and ATSR pixel. Such backscatter signature values have been included in the subsequent data analysis.

RESULTS AND DISCUSSION

The intensive study during the previous years tried to answer the question to what extend the value of backscatter depends on surface roughens and soil plant moisture. In this paper we wanted to present the spatial distribution of backscatter and its relation to soil moisture and roughness for different roughness-moisture conditions in order to determine the relative influence of the parameters on backscattering signal and perform the statistical analysis. The first part of the studies presents the variation of backscattering coefficient and soil moisture as point measurements for 1992 - 1996. The second part presents statistical analyses for satellite data (ERS-1/2.SAR, ATSR, NOAA/AVHRR).

Generally the simple correlation between sigma and soil moisture for each of the crop during vegetation season was poor. Only for spring triticale (r=0.7) and spring wheat (r=0.5) correlation was stronger.

The Figure 1 presents the relation between backscattering coefficient and soil moisture for the fields covered by cereals for the period of 1992-1996. The surface roughness has not been considered. The correlation was poor. It is difficult to present the surface roughness when the soil is covered by vegetation.

One of the approach that has been considered in this research was to measure Leaf Area Index (LAI) for different vegetation growing stages. LAI values change during the growth of vegetation from the low at the beginning of growth reaching maximum at heading stage of cereals and then at the stage of ripening. In our research LAI values represent surface roughness. Figure 2 presents the relationship between backscattering coefficient of ERS/SAR and calculated using LAI and soil moisture values. The correlation is better, correlation coefficient equals to 0.70.

Also considering different LAI classes for cereals the relationship between soil moisture and backscattering coefficient has been considered. The Figure 3 presents the relation between backscattering coefficient and soil moisture for one of LAI class i.e. 3-4. For separated LAI classes the correlation between sigma and soil moisture for cereals is as follows:

LAI from 1 to 2 - r= -0.68,

LAI from 2 to 3 - r= 0.46,

LAI from 3 to 4 - r= 0.62,

LAI>4 - r= 0.82.

In order to eliminate the surface roughness or to consider it as constant the relationship between backscattering coefficient and soil moisture has been examined. The best correlation between sigma and soil moisture for cereals was for high LAI values when crops are in their maximum biomass (and of jointing to heading). At the stage of similar surface roughness (heading/graining and ripening for the year 1992 - 1996) the soil moisture values varied from 3% to 19%. The results of correlation between backscatter and soil moisture at this stage of the crop for spring wheat were significantly better, r= 0.73, Fig. 4.

At the same time of ERS-1/2 overpasses the ground measurements of surface albedo and radiative temperature have been carried out in order to calculate Normalized Vegetation Index (NDVI) and Water Deficit Index (WDI). These indices were considered as an indication of surface roughness (NDVI) and soil moisture (WDI) which also controls the ability of plant soil system to water loss to the atmosphere. Water Deficit Index (WDI) developed by Moran et al. indicates soil - vegetation status. It takes into account the actual soil water conditions in relation to potential and drought. The index considers the percentage of vegetation cover. The Fig. 5 presents the relationship between backscattering coefficient and WDI (calculated using measurements from the ground level) for spring wheat (between 1992 - 1996). The best relation between backscattering coefficient and Water Deficit Index is for the ripening stage of vegetation what confirmed the results presented at the Fig 4.

The second approach to this study has been to use only satellite data to determine the relation between backscattering coefficient and soil-vegetation moisture and roughness. For the study area covered by ERS/SAR image the NOAA/AVHRR data were introduced. Block of ERS/SAR pixels was extended to the size of NOAA/AVHRR pixels. The grid of NOAA pixels was overlaid on ERS/SAR image. For each of NOAA pixels NDVI and WDI indices were calculated. The NDVI represented surface roughness and WDI soil - vegetation water deficit. The Figure 6 presents three curves. First represents the values of backscattering coefficient averaged for the area of each of NOAA pixels. The next curve represents the values of NDVI and the third the values of WDI for the same pixel. The study area was registered by SAR on 20 05 95 and by NOAA on 23 05 95. The backscattering coefficient fluctuated from -13 to -8 dB. The high values of the coefficient i.e. -8 dB represented wet soil water condition, where Water Deficit Index was 0.75 (pixel 44-48). Dry soil water conditions occurred for the area represented by pixel 68-70, for which backscatter was equal to -13 dB and WDI was close to 1. The NDVI values represented area roughness. One as the surface structure which was constant (like tillage direction or infrastructure), and the other which depend on crop structure and vegetation status. The Fig. 7 shows the differentiation of backscatter values for the year 1992 - 1996. The 1992 and 1996 year was the most dry, the sigma values were the lowest. The examination of changes of backscatter values for different years for the same pixels will show the changes of soil - vegetation conditions across the years.

The examination of relation of ERS/SAR data to NOAA data gave the possibility to examine the potential to apply the ATSR data for better understanding of soil - vegetation water conditions.

For the arable area we have examined radiative temperature measured by Along - Track Scanning Radiometer (ATSR) for two dates i.e 20.05.1995 and 28.07.1995. The grid of ATSR pixels (20.05.1995) was overlaid on ERS/SAR image (20.05.1995). The sigma values were averaged within the block of 80 SAR pixels which corresponded to an area of one ATSR pixel.

The Figure 8 presents the variation of three curves. One represents the backscattering coefficient, second radiative temperature and the third Water Deficit Index (WDI).

The values of backscattering coefficient for the pixel 72 and 73 were equal to -12 dB and for the pixel 64 and 65 were equal to -10 - 10.5 dB. The soil - vegetation moisture for the area covered by pixels 64 and 65 were much better than these at the area covered by the pixel 72 and 73, what reflected in lower values of Water Deficit Index (WDI).

The Figure 9 presents the curves of NDVI values calculated from AVHRR data for ATSR pixels and radiative temperature from ATSR and backscatter values obtained from ERS/SAR for 28, 29 July. For the area of temperature increment what indicated the decrease of soil moisture the values of backscatering coefficient lowered.

The visible and infrared range of electromagnetic spectrum gives the information about crop status and allows to examine and understand better the influence of soil - vegetation complex on radar signal.

The project is being continued and the other data of SAR, ATSR and ground observations will be added.

LITERATURE

Bamler, R., 1992, A comparison of Range-Doppler and Wavenumber Domain SAR focusing algorithms, IEEE Trans. on Geoscience and Remote Sensing, Vol. 30.

Dabrowska-Zielinska, K., Gruszczynska, M., Janowska, M., Stankiewicz, K., Bochenek, Zb., 1994, Use of ERS-1 SAR data for soil moisture assessment, Proc. of the First Workshop on ERS-1 Pilot Projects, Toledo, s. 79-84.

Dabrowska-Zielinska, K., Moran, M. S, Janowska M., Gruszczynska M., Stankiewicz K. 1995, Visible, infrared and microwave data as a source of information about vegetation status. Proc. the Meteorological Satellite Data Users Conference, Winchester, U.K.

Laur, H., 1992, Derivation of backscattering coefficient in ERS-1 SAR.PRI products, ESA Esrin.

Moran, M. S., Clarke, T. R., Inoue, Y., Vidal, A., 1994, Estimating crop water deficit using the relation between surface-air temperature and spectral vegetation index. Remote Sensing of Environment, s. 246-263.

Nghiem S.V., T. Le Toan, J.A. Kong, H.C.Han, M. Borgeaud 1993, Layer model with random spheroidal scatterers for remote sensing of vegetation canopy. Journal of Electromagnetic Waves and Applications, Vol.7, No.1,pp.49-75.

Wooding M.G., G.H. Griftiths, R., Evans, P. Bird, D. Kenward, G.E. Keyte 1992, Temporal monitoring of soil moisture using ERS-1 SAR data. Proc. First ERS-1 Symposium, Cannes, pp 641-648

T. Le Toan, P. Smacchia, J.C. Souyris, A. Beaudoin, M.Merdas, M.Wooding, J.Lichteneger 1993, On the retrieval of soil moisture from ERS-1 SAR data. Proc. Second ERS-1 Symposium

Ulaby, F.T., 1980 Microwave response of vegetation ; 23rd Ann. Conf. of Committee on Space Res. (COSPAR Adv. Space Res., vol.1 pp 55-70

Figure Caption

Fig.1. Relation between ERS-1/2.SAR backscatter (SIGMA) and soil moisture from ground measurements for cereals.

Fig.2. Relation between backscatter observed from ERS-1/2.SAR (SIGMA) and backscatter predicted from soil moisture and LAI point measurements (1992-1996).

Fig.3. Relation between backscatter from ERS-1/2.SAR (9x9) and soil moisture from point measurements.

Fig. 4. Relation between backscatter and soil moisture (SM) from point measurements (days 165 and 195 from 1992-1996) for spring wheat.

Fig. 5. Relation between backscatter from ERS-1/2.SAR (SIGMA-9x9) and WDI from point measurements for spring wheat for 1992 - 1996.

Fig. 6. Temporal evolution of backscatter averaged for NOAA pixels and NDVI and WDI from NOAA.

Fig. 7. Temporal evolution of backscatter from May 1992-1996 averaged for NOAA pixels.

Fig. 8. Plot of backscatter (SIGMA) averaged for ATSR pixels and surface temperature and WDI from ATSR.

Fig. 9. Plot of backscatter (SIGMA) averaged for ATSR pixels, surface temperature from ATSR and NDVI from NOAA.

Keywords: ESA European Space Agency - Agence spatiale europeenne, observation de la terre, earth observation, satellite remote sensing, teledetection, geophysique, altimetrie, radar, chimique atmospherique, geophysics, altimetry, radar, atmospheric chemistry