Weather Effects on Radar Backscatter from Crops

Abstract

There has been recent interest in the effect of weather on radar backscatter. This paper presents work done in understanding the effect of weather, especially rain on backscatter intensity and phase coherence from crop canopies, in particular, crops typical of South-Eastern England (e.g. wheat and beans).

Current work centres on the use of a computer model to investigate the effect of rain on radar backscatter from crops over the growing season. These results will then be compared with the weather data, crop observations and SAR data for the local area. The study integrates locally collected datasets including crop observations, meteorological measurements, ERS SAR images and computational modelling in order to better understand how the change in crop radar backscatter throughout the growing season is perturbed by rainfall events.

The ultimate aim of the study will be to indicate quantitatively, the disturbance effects of rain (and possibly wind) on SAR images of crops in order to improve the monitoring and classification of crops by microwave remote sensing.

Keywords: radar backscatter, phase coherence, crops, computer modelling, ERS SAR, weather.

1 Introduction

Satellite-borne radar sensors are being used to monitor crops. In particular, the SAR data from the ERS series of satellites are being used to distinguish different crop types and identify changes in vegetation biomass. Although ERS SAR data can be obtained in all weather conditions, there is an awareness that the weather can affect the microwave backscatter from agricultural surfaces. The availability of rainfall information before or during a satellite pass is essential for the correct interpretation of data (Lichtenegger, 1996). Hence, the effects of rain and wind on radar backscatter must be corrected for in order to reduce the errors in the interpretation of crop radar signatures for agricultural monitoring.

Fluctuations in backscatter during the growing season may be explained by wind and rainfall events (Nieuwenhuis & Kramer, 1995). Rainfall of about 10 mm prior to SAR data acquisition has led to increases in backscatter of 1-4 dB from grass, bare soil and wheat-bare soil cover types (Wooding et al., 1992). Moreover, there may be some evidence that a period of low observed backscatter from winter wheat corresponds to a period of low rainfall, although there may have been other causes (Wooding et al., 1993).

The project described in this paper aims to quantify the weather effects on backscatter intensity and phase coherence for typical UK agricultural crops. Work is underway and the paper presents aspects of the database being compiled as well as discussing the modelling approach used.

2 Methodology

A database of crop, satellite and weather data is being created. . The database includes observations of crop development and field conditions at local test sites, weather records, and a series of ERS-1 and 2 SAR images. The data collection has been designed to monitor weather-dependent radar backscatter, within the resources available, as opposed to classifying or monitoring crops. The database is being used to develop and validate a model of weather effects on radar backscatter from crops.

In addition, this database is being extended to include the 1996/1997 growing season.

2.1 Crop sampling

Measurements of the nearby winter wheat and bean crops in the Cranfield area were made (52 04' N -0 37' E) during the growing season with a particular emphasis on crop height and moisture content during the period May-August 1996, during which time the crops developed rapidly. In addition, photographs of the crops were taken to provide a record of crop structure during the growing season.

2.2 ERS SAR data acquisition

ERS SAR images have been obtained from two sources. ERS-1 Precision Images (ERS-1.SAR.PRI) of the local area were obtained from the RAIDS service over a period spanning July 1995 to May 1996 (Graves, 1995). In addition, SAR images from the ERS-1 and ERS-2 Tandem Mission were ordered.

2.2.1 RAIDS SAR Imagery

Details of the images from the RAIDS service are shown in Table 1a. These are 16-bit images with a pixel size of 12.5 m x 12.5 m covering areas of approximately 25 km x 25 km. The spatial resolution of the images is about 30 m. A field that was used to grow winter wheat and a second field that was used to grow beans were identified within this set of images. Also, a field of cropped grass was used as a form of control: the small changes in the grass plants, that were deliberately kept short, over the growing season would be expected to have a relatively constant radar backscatter profile.

2.2.2 ERS Tandem Mission Imagery

Image pairs from the ERS Tandem Mission [Duchossois and Martin (1995)] are being obtained for the growing season 1995/96. Up to 25 image pairs may be required since our test sites are close to a frame edge. Each image pair covers an area approximately 50 km square. These images will be used primarily for the backscatter coherence studies although they also enhance the quality of data available for the backscatter intensity work.

2.2.3 SAR Data Analysis

Average relative backscatter intensities were obtained for the wheat and bean fields before and during the growing season. These values were converted to decibels (dB). It was not possible to calculate the absolute average backscatter intensity values for the crop fields due to the inappropriate value of the calibration constant, K supplied with the RAIDS ERS-1.SAR.PRI data. However, since the changes in backscatter due to weather effects, over the growing season, are of primary interest, analysis of relative backscatter intensity values in decibels is unaffected. The only consequence of using relative backscatter values is the large linear offset in the backscatter time series graphs.

The ISAR software developed by the Politecnico di Milano is being used to process the Tandem Mission images (Koskinen, 1995). Coherence images for each overpass are the primary output required, thus phase unwrapping is not necessary. The basic SLC images allow the calibration constant for coincident RAIDS images to be evaluated.

2.3 Meteorological data collection

An automatic weather station at Cranfield University, within 5 km of the fields under study, was used to acquire most of the weather measurements during the SAR image acquisitions and crop sampling. In addition, some data from Silsoe College (approximately 15 kilometres east of Cranfield) were also acquired. Parameters recorded include wind speed, wind direction, rainfall, solar and net radiation, air temperature, and humidity.

2.4 Radar backscatter modelling

Figure 1 describes the conceptual model underlying the radar backscatter modelling. The aim of the research is to investigate the effect of weather on backscatter intensity and phase coherence (particularly for radars similar to those of ERS-1 and 2). Previously published work allows most of the components identified to be quantified, at least approximately, and our current modelling work is concerned with refining specific components. Relative permittivity models for soil and vegetation based on those published by Dobson et al. (1985) and Ulaby & El-Rayes (1987) are useful since they identify the separate contribution of free and bound water. Rain / dew is assumed to modify the effective relative permittivity of the soil or vegetation by its contribution to the free water component.

A key part of the model is the component which calculates backscatter coefficient from knowledge of the scatterers' relative permittivity and their geometry. Several approaches are being evaluated, including empirical relationships, full electromagnetic scattering models (e.g. MIMICS) and simplified scattering models.

Hobbs (1996) describes the coherence simulator used. For this, all scatterers within the sample volume are assumed to belong to one of several classes defined by their variability in response to weather (e.g. soil might be regarded as invariant, large branches to be slightly variable, and twigs and leaves to be highly variable). The simulator's inputs are the relative strengths of these different classes and their variability (phase and amplitude). A similar approach for forest applications is being developed at Chalmers University (e.g. Smith et al., 1996).

3 Results

Table 1 and Figure 2 are examples of the information available from the database being compiled. Table 1a gives the mean relative backscatter values and standard deviations from the winter wheat, bean, and grass fields throughout the growing season. The grass fields were used as a form of control. As these were cropped, it was expected that the radar backscatter due to changes in vegetation would be relatively small and other factors would be more influential. Table 1b shows the mean vegetation heights of the species under study, which is a significant vegetation parameter for radar backscatter.

Figure 2 shows the mean relative temporal backscatter profile and standard deviations for the winter wheat, bean, and grass fields.

4 Discussion

The results in Figure 2 show a clear contrast between the relatively constant backscatter of the airfield's grassed areas and the annual cycle of the farmed test sites. The most significant effects appear to be related to the soil preparation in the autumn and the gradual weathering of the soil surface over the winter. Relative to this, the backscatter intensity changes during the growing season are small. These patterns are similar to those observed by other workers (e.g. Nieuwenhuis & Kramer, 1995). Similar plots for the backscatter coherence magnitude have not yet been obtained, but when available will provide an interesting comparison with the plots backscatter intensity.

5 Conclusions & Further work

The next stages of the project are to (1) continue building the database for our test sites, (2) refine the backscatter models, and (3) derive time series for the coherence magnitude. Our aim is to develop (A) statistical relationships between backscatter intensity or phase coherence and suitable weather parameters, and (B) physically-based models to relate backscatter to weather parameters. The final stage will involve evaluating the relationships derived using our database against other similar databases, ideally for different geographic or climatic regimes (another current project at Cranfield should give access to such a database for southern Europe).

6 Acknowledgements

We would like to thank AEA Technology PLC, the Engineering and Physical Sciences Research Council, the European Space Agency, Matra Marconi Space UK Ltd., The Royal Society and Silsoe College for their support of this research.

7 References

Dobson, M. C., Ulaby, F .T., Hallikainen, M. T. & El-Rayes, M. A., 1985:

Microwave Dielectric Behavior of Wet Soil - Part II: Dielectric Mixing Models. IEEE Transactions on Geoscience and Remote Sensing, Vol. GE-23, No. 1, pp. 35-46.

Duchossois, G. & Martin, P., 1995:

ERS-1 and ERS-2 Tandem Operations. ESA Bulletin, No. 83, August 1995, pp. 54-60.

Graves, A., 1995:

Rapid Information Dissemination System (RAIDS). IN: Proceedings of the Second ERS Applications Workshop, ESA SP-383, 6-8 December 1995, London, United Kingdom, pp. 383-385.

Hobbs, S. E., 1996:

Weather Effects on SAR Backscatter for Agricultural Surfaces. Proc. ESA Workshop on SAR Interferometry, Zurich, 30 Sept - 2 Oct 1996.

Koskinen, J., 1995:

The ISAR-Interferogram Generator Manual. Version 3.0, 7/18/95, 52pp, DEX/ED, ESA-ESRIN.

Lichtenegger, J., 1996:

ERS-1 SAR Images - Mirror of Thunderstorms. Earth Observation Quarterly, No. 53, pp. 7-9.

Nieuwenhuis, G. J. A. & Kramer, H., 1995:

Monitoring of Agricultural Crops with ERS-1 and JERS-1 Multi-Temporal SAR Data Crop Growth Modelling. IN: Proceedings of the Second ERS Applications Workshop, ESA SP-383, 6-8 December 1995, London, United Kingdom, pp. 383-385.

Smith, G., Dammert, P.B.G., and Askne, J., 1996.

Decorrelation mechanisms in C-band SAR interferometry over boreal forest. Proc. ESA Workshop on SAR interferometry, Zurich, 30 Sep - 2 Oct 1996.

Ulaby, F. T. & El-Rayes, M. A., 1987:

Microwave Dielectric Spectrum of Vegetation - Part II: Dual-Dispersion Model. IEEE Transactions on Geoscience and Remote Sensing, Vol. GE-25, No. 5, pp. 550-557.

Wooding, M. G., Griffiths, G. H., Evans, R., Bird, P., Kenward, D. & Keyte, G. E.,.1992:

Temporal Monitoring of Soil Moisture using ERS-1 SAR Data. IN: Proceedings of the First ERS-1 Symposium - Space at the Service of our Environment, ESA SP-359, 4-6 November 1992, Cannes, France, pp. 641-648.

Wooding, M. G., Zmuda, A. D. & Griffiths, G. H., 1993:

Crop Discrimination using Multi-Temporal ERS-1 SAR Data. IN: Proceedings of the Second ERS-1 Symposium - Space at the Service of our Environment, ESA SP-361, 11-14 October 1993, Hamburg, Germany, pp. 51-56.

Figure Caption

Table 1a. The ERS-1.SAR.PRI images used to obtain relative mean vegetation backscatter profiles and standard deviations.

Table 1b. Crop heights during growing season.

Figure 1. The conceptual basis underlying the radar backscatter modelling.

Figure 2. The mean relative temporal backscatter profile and standard deviations for the winter wheat, bean, and grass fields.

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