A Method Using Different Remote Sensing Techniques for Estimating Grassland Bio-physical Variables

For efficient grassland management, information on the spatial variation of the crop within fields is of the utmost importance. Currently, this mainly depends on qualitative expert knowledge. Quantitative information on the actual status of grass swards at the right moment in the season is important for management decisions, like nitrogen supply, water supply or harvesting. Remote sensing has proven to be a useful technique for estimating and mapping the spatial variation of various biophysical variables. A whole range of vegetation indices has been developed for estimating variables like biomass, leaf area index and the fraction of absorbed photosynthetically active radiation for a range of vegetation types. However, calibration of the image data is crucial in the performance and applicability of this technique. The aim of this paper is to show the possibility to calibrate image data using fast non-destructive close-range sensing instruments, hence being able to build models to assess important plant characteristics on large areas. A homogeneously managed grassland field of about one hectare was used as a test site. It was subdivided into 20 plots of 15 by 3 meters. End of July 2004 measurements were performed with a close-range sensing device, the so-called Imspector Mobile. This imaging platform consists of two imaging spectrographs, covering the spectral range of 440-960 (Imspector V9) and 850-1680 nm (Imspector N17), and a 3-CCD camera, equipped with special band filters (centre wavelengths are at 600, 710 and 800 nm). The platform is further equipped with artificial light sources. An airborne campaign with the UltraCam digital CCD camera (centre wavelength of the four bands are at 460, 520, 660 and 740 nm) was used for extrapolation to large scale areas. Plots were harvested and variables like fresh and dry biomass, and leaf nitrogen content were determined. Results showed that the Imspector Mobile could be used for estimating crop variables of the grassland field with a grass-clover mixture. Partial least squares (PLS) models combining spectral and spatial information from the Imspector Mobile yielded acceptable results in predicting crop variables. Subsequently, the predicted field variables were used to build a prediction model using the reflectance values of the UltraCam images. These were then compared with the measured field variables and the model proved to have acceptable predictive power. INTRODUCTION Energetically speaking, current agricultural cropping systems have a low efficiency. Under Dutch growing conditions with highly developed agricultural practices – still no more than 1% of the energy of the sun reaching the earth surface is fixed in plant biomass (chemical energy in crops). The net yield of grassland is estimated at no more than 8.5 ton dry matter (DM) yield per hectare on average each year, whereas the potential gross yield with the same agricultural practices on experimental plots is estimated at 15 ton DM per ha per year. If we can lower this so-called 'yield gap', the same or more biomass can be produced on a smaller area. Since the area of grassland in the densely populated Netherlands is over 1 million ha, more efficient management lowering the yield gap is extremely useful. However, at this moment it is still not clear what causes this ‘yield © EARSeL and Warsaw University, Warsaw 2005. Proceedings of 4th EARSeL Workshop on Imaging Spectroscopy. New quality in environmental studies. Zagajewski B., Sobczak M., Wrzesień M., (eds) gap’. Insight in the causes is hampered by the lack of fast and accurate methods for determining crop growth. Various techniques, e.g. using imaging spectroscopy, are therefore developed for determining the causes of this yield gap. If we do know the causes, we can optimize current grassland management practices, using precision agriculture. Optimization refers to biomass production of high quality in order to maximize animal production. This requires the need to monitor the crop over large areas. To assess stagnation of growth and impending yield reduction at a large scale, the availability of fast, timely and accurate methods is called for. Furthermore, such methods can also be used for estimating the amount of nitrogen left in the soil, which could be re-used, hence minimizing environmental impacts associated with fertilization. The latter being of increasing concern [1]. The role of imaging spectroscopy for the characterisation of grass swards was studied by Schut [26]. He explored the potential for growth monitoring, detection of nitrogen and drought stress, and assessment of dry matter yield, clover content, nutrient content, feeding value, sward heterogeneity and production capacity using a close-range imaging spectroscopy system applicable in the laboratory or for mini-experiments. From this a mobile system was developed for application to field experiments, the so-called Imspector Mobile [7]. This imaging platform consists of two imaging spectrometers, covering the spectral range of 440-960 (Imspector V9) and 8501680 nm (Imspector N17), and a 3-CCD camera, equipped with special band filters (centre wavelengths are at 600, 710 and 800 nm). The platform is further equipped with artificial light sources and thus can measure independently of external weather conditions. A combination of image parameters and hyperspectral reflectance curves derived from classified images can be used to estimate yields, nutrient contents and feeding value of grass plots [8]. Although the Imspector Mobile is a non destructive and relatively fast measurement device, it is not suitable for monitoring large areas of grassland. Remote sensing has proven to be a useful technique for estimating and mapping vegetation biophysical variables over large scale areas. Both statistical and physical methods have been used for describing the relationship between spectral measurements and biophysical variables. As an example, a whole range of vegetation indices has been developed for estimating variables like biomass, leaf area index and the fraction of absorbed photosynthetically active radiation for a range of vegetation types. For estimating leaf chlorophyll and nitrogen content imaging spectroscopy has shown promising results [9]. However, calibration of the image data is crucial in the performance and applicability of this technique. Therefore we try to combine the dedicated close-range sensing equipment with remote sensing techniques. By using the Imspector Mobile for calibrating the remote sensing data, we hope to be able to extrapolate the results to large areas of grassland. The idea is illustrated in figure 1. Beeri et al. [10] developed an alternative approach for using ground-truth models for sugar beet N-credit and tested these models with satellite images. Figure 1: Schematic representation: destructive and expensive measurements are used to calibrate the close-range sensing device. The close-range sensing device can then characterise the status of the crop at many locations, fast and non-destructively. These data can then be used to calibrate the remote sensing data. After calibration, these data can be used for extrapolation of these characteristics to larger areas. Physical + chemical measurements Close sensing Remote sensing cal. cal.