AGRICULTURE AND SOILS
The main objective of the land studies is to derive estimates of land surface biochemical (chlorophyll content) and biophysical (LAI/biomass, albedo, fAPAR) properties by inverting semi-empirical and physically based BRDF models against sample sets of hyperspectral directional reflectance measurements acquired by CHRIS. In addition to specific BRDF studies, the hyperspectral data will also be used to investigate 'red edge' shifts in wavelength as a function of canopy biochemistry (chlorophyll content/concentration).
One can consider some advantages and disadvantages of three different types of BRDF model:
- Physically based
- based on approximations to radiative-transfer equations
- computationally intensive
- numerical inversion techniques sensitive to initial estimates of model parameters
- Empirical
- computationally simple
- poor extrapolation beyond solar and sensor angles for which reflectance data are available
- Semi-Empirical
- computationally efficient analytical solutions to model inversion using matrix-based techniques
- scale linearly
The overall aim is to assess the ability of multi-angle reflectance data sampled using the CHRIS sensor to retrieve/estimate land surface biophysical properties (LAI, albedo etc). An element of this will include direct comparison of the retrieved parameter values against ground-based measurements of the same parameters. The surface biophysical parameters will be estimated from the CHRIS data using a number of different techniques, ranging from traditional (the so-called vegetation indices), through more recent developments (such as red-edge position (REP), to more advanced techniques (notably BRDF model inversion). The validation exercise will look at the accuracy/precision with which atmospherically-corrected CHRIS data can be used to determine the basic surface radiation fields, as well as at the ability to retrieve properties such as LAI, albedo, etc. through the application of vegetation indices, REP techniques and BRDF model inversion.
Test sites
Test sites have been selected to include the major terrestrial biomes. Provisionally these include East Anglia, UK, Niamey, Niger, Jornada, New Mexico, USA, Boreas, canada, La Crau, France and Southern Africa. More details are provided below concerning experiment plans relating to the test sites in East Anglia and Southern Africa.
The test site in East Anglia, UK, encompasses a range of land cover types typical of northern Europe, including cereal crops (wheat and barley), root crops (potato, sugar beet, etc.), permanent pasture and both coniferous and deciduous woodland. Sample sites appropriate to the size of images that will be acquired by the CHRIS instrument have already been identified, each covering a different sub-set of the full range of cover types present throughout the whole study area. The sub-sites have also been chosen to reflect the principal differences in soil type and solid geology within East Anglia, so that differences in the substrate can be assessed in terms of their impact of the observed bidirectional reflectance trends.
There is a long-term fieldwork programme running throughout 1999 and 2000 (March-October) at the East Anglia test site. A wide range of ground-level, airborne and satellite sensor data are being acquired, and there is an extensive network of contacts with local farmers etc. to permit access to individual fields.
The ground-level data consists of detailed measurements of vegetation and soil properties, notably Leaf Area Index (LAI) using an LAI-2000 and hemispherical photography, spectral reflectance (using a range of field spectroradiometers), albedo (using Kipp and Zonen net radiometers), and observations of plant growth stage and percentage ground cover. Measurements of atmospheric optical properties will also be acquired using a CIMEL sun photometer. The latter will provide information on the atmospheric phase function that is required to estimate surface biophysical properties from the CHRIS data. An extensive sample of land cover/land use at 1800 locations within East Anglia has also been carried out, and these data will be used to generate a large-area land cover map from a combination of Landsat-5 TM and Landsat-7 ETM+ image data that we are acquiring under separate funding. This programme of fieldwork will be repeated through the year 2000.
In addition to the fieldwork programme, which involves measurement of the relevant properties every 7-14 days throughout March-October, a range of airborne and satellite sensor data are also being acquired. Currently there are negotiations with the Infoterra Ltd. to acquire hyperspectral image data using the new HyMap system over the one of the main sample sites (Barton Bendish, East Anglia). This will provide hyperspectral image data in the range 400nm-2500nm at 2-3nm spectral resolution. These data will be of value in assessing the impact of selecting specific spectral channels in the CHRIS data. They will also assist in 'scaling up' the point-based ground-level measurements to the spatial resolution of the CHRIS instrument. Should the HyMap data not be available , there are archive data (1997) acquired by an airborne CASI (compact airborne spectrographic imager) and Daedalus AADS-1286 multispectral scanner that could be used as a substitute (since cropping patterns do not differ dramatically from year-to-year in this region).
Various types of satellite sensor data will also be acquired, including Landsat-7 ETM+, NOAA-AVHRR, SPOT-VGT and MODIS. Data from every single overpass of the NOAA-AVHRR sensors are available through the U.K. receiving station based at Dundee (Scotland). This will provide high temporal (albeit low spatial information) that can be used to follow the development of the crops throughout the annual growing cycle. An end-to-end data processing chain has been developed that takes the raw AVHRR data from Dundee, calibrates it, performs cloud-masking and atmospheric correction, geometric registration and BRDF model inversion. While, ultimately, the radiometric calibration of the AVHRR sensors limits their use for precise determination of surface reflectance fields, they can provide a useful indication of the relative temporal variations in surface reflectance, and it is planned to compare the BRDF model parameter retrieved using these data with spatially aggregated values determined from the CHRIS instrument. Data from the SPOT-VGT instrument are available for similar ends through involvement in the SPOT Vegetation Preparatory programme. This will provide information over much of western Europe that can be used as means of cross-calibration/validation with CHRIS data from other proposed European data collection sites. Later this year, data from NASA's MODIS sensors will also come on stream and will be made available through our involvement in the MODIS Science Team. In connection with the latter, regular Landsat-7 ETM+ data will be provided over East Anglia and a CIMEL sun photometer is to be based permanently at the main Barton Bendish site.
Southern Africa
It is hope to extend field validation to a number of other (albeit less intensively instrumented) sites elsewhere. To maximise the potential benefits accruing from this exercise, it is planned to co-locate our experiments with larger field measurement programmes. One possibility for next year will be to take part in SAFARI 2000, a programme being coordinated by colleagues in the USA but taking place in southern Africa. The details associated with this are still being finalised.
Data requirements
The basic requirement for BRDF studies is a set of hyperspectral images obtained with different viewing geometries. Across-track viewing (up to 30°) will be used to obtain these additional angular samples and to increase the range of solar zenith angles. The default will be 3 sets of 5 images with along track ground incidence angles of 0°, +/-25° and +/-44°, acquired within a 16 day period using different across track viewing angles. These images will be acquired at 25m x 25m resolution in 19 spectral bands. Sets of images are required in different growing seasons, normally 4 times a year.
