REGIONAL APPLICATION OF ERS-1/2 IN THE FLEVOLAND AGRICULTURAL AREA IN THE NETHERLANDS.

ABSTRACT

In the scope of searching applications of ERS-1/2 satellite data in agriculture this study has been formulated. This paper summarizes the state-of-the art in crop growth monitoring with use of remote sensing. The study is a follow-up of the ESA study in 1993 (ESA/ESTEC contract # 9837/92/NL/GS) where airborne RS data has been used, this study focuses on the potential of the present satellites for crop growth monitoring. Optical data has been used for calibration and validation purposes. The study has been performed by a Dutch consortium made up by SYNOPTICS AB-DLO, FEL-TNO and WAU.

Keywords: Crop monitoring, modeling, ERS-1/2

1. INTRODUCTION

In search of application of ERS-1/2 satellite data in agriculture the ESA study "Vegetation retrieval by combined microwave and optical remote sensing" (Contract nr. 11154/94/NL/NB) had been formulated. The summary of this study reflects the state of the art in crop growth monitoring with remote sensing.

For the first time a complete time-series of three successive years of the same test site in the Southern Flevoland Province in the Netherlands offered a opportunity to the Dutch research community to study the potential of radar satellite observations to crop growth monitoring. In this study three major crops in the Netherlands were subject of study: wheat, potato and sugar beet. Regional averages of the backscatter of these crops were used to calibrate and validate the semi-empirical 'Cloud-model'.

2. METHODOLOGY

The Flevoland test site in the Netherlands has been observed by ERS-1/2 for the three successive years: 1992, 1993 and 1994. In 1994 JERS-1 data has been used in order to study the potential for crop growth monitoring as well. Extensive data sets from former studies as well ground based and airborne campaigns like ROVE, Agriscatt, MAESTRO and MAC Europe, were used in order to support the theoretical microwave modeling and crop growth modeling.

Remote sensing models were selected and used in this study. The development of RS models is highly dependent on the crop type and on the physical assumptions underlying the modeling in a specific part of the spectral domain. Furthermore, in general, one can say that the modeling activities in the optical region are better understood compared to that of the microwave region. In this study three major crops in the Netherlands were subject of study: wheat, potato and sugar beet. Regional averages of the backscatter of these crops were used to calibrate and validate the semi-empirical 'Cloud-model'. The more complicated 'WSRC-model' (FEL-TNO) based on the radiative transfer theory was used in combination with well initialized crop growth models (AB-DLO) for the simulation of parameters of the 'Cloud-model' as well.

The major findings of the study and potentials for application of ERS-1/2 :

The potential of ERS-1/2 for crop growth monitoring is mainly dependent on the amount of satellite overpasses in time and on the moment of observation during the growing season. This differs for each crop. It is promising that the parameterized Cloud model for regional applications can be applied for different years. However, it is difficult to use the calibrated Cloud model for accurate yield prediction as the standard deviation of the regional backscatter is rather large. In this respect estimation of biomass by inversion of the Cloud model is troublesome and therefore not accurate enough for calibrating the crop growth model. Due to the low accuracy of estimation of biomass (plant water), calibration of the crop growth model is not possible.

Figure 1. Comparison of ERS observations of 1992 and backscatter simulations with a CLOUD model calibrated on soil moisture derived from fruit trees backscatter (open vegetation) for winter wheat. (1-layer Cloud model Wheat, ERS 1992-1994: Soil G: 0.045 Crop C : 0.005 Soil K : 0.078 Crop D : 0.2)

Moreover, the crop growth model imposes high requirements on the accuracy of estimation of biomass from remote sensing as the crop growth model can reproduce very accurate growth information already, when the sowing date and meteorological information is present. Exhaustive studies in the past have resulted into accurate physiological descriptions of crop growth. Retrieved information from remote sensing like biomass should not exceed an accuracy of at most one unit (in kg/m²) or a shift of few weeks in time. From simulations, it appeared that even a shift of 50 days in sowing date could not be detected by the microwave time-series. On the other hand the standard deviation in time showed interesting applications in finding the moment of regional closure of leafy crops, like potato and sugar beet.

Fig 2. ERS-1 signatures of standard deviation in backscatter of sugar beet for the 1992, 1993 and 1994 growing season

Soil properties (especially soil moisture) appeared to have more impact on total backscatter of ERS-1/2 then was expected from the modeling. The positive result of this intensive study is that ERS-1/2 satellite time-series can provide soil moisture information under vegetated conditions, which can be of particular interest for water balance studies. The latter is very important under semi-arid climatic conditions, where the soil water balance is of major importance to crop growth and possible growth limitations. The conditions in the Flevopolder are favorable and no water stress and with that growth limitation is expected to occur. Interesting is the use of crop growth models extended with a good hydrological model in those semi-arid conditions. Microwave time-series could provide these models by estimating top surface moisture conditions which are an indication of the water availability over the soil profile.

3. CONCLUSIONS

The ESA study resulted into three major conclusions relevant for applications of ERS data in the agribusiness sector:

1. ERS-1 backscatter from agricultural crops like potato and sugar beet show still clear dependency from soil moisture during the growing season. However, during the vegetative and the begin of generative growth of the winter wheat crop (day 120 till 175) almost no correlation can be noticed with soil moisture. Apparently, the canopy structure (in combination with soil) dominates the backscatter behavior.
2. The presentation of the regional standard deviation (100 fields per crop type) in backscatter gives a clear dip in backscatter during the moment of regional crop closure. While the regional intensity information of the winter wheat crop gives a clear signature in time. This information on agricultural crops are of particular interest to the agribusiness industries operating on a regional administrative level.
3. A new regional approach had been developed to model the canopy and soil contribution of the agricultural crops by using a regional crop growth model and a simplified backscatter model like the Cloud model. The regression parameters of the Cloud model appeared to be valid also between the successive years. Modeling results on regional level show remarkable well agreement with the actual measurements.

Figure 3. Average soil moisture contents for the three crops compared with the average radar backscatter signatures for the three crops in 1992.

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