Global Change, 1992-1994

A project has been set up to develop physically based models and numerical algorithms to extract quantitative information from satellite remote sensing data. The project investigated biosphere atmosphere interactions, assessing the effect of climatic variability and changes on the biosphere. The approach has centred around the extraction of physical environment parameters from remote sensing data and around the development of physical biosphere models using such data as a source of information. The presence of live vegetation in the environment can be detected on the basis of remote sensing data by exploiting the strong spectral contrast that plants exhibit near the 0.7 um wavelength. The simple spectral indices which exploit this contrast are also sensitive to various perturbing factors such as the presence of atmosphere and the reflectance of the underlying soil. The Global Environment Monitoring Index (GEMI) was designed to address this issue. The reduced sensitivity of GEMI to atmospheric and soil effects was confirmed and the performance of this nonlinear index was documented. One of the major drawbacks associated with remote sensing data from existing satellite sensors is that the combination of a directional solar radiation with a directional viewing, together with the anisotropy of the terrestrial surfaces, results in measurements which are intrinsically dependent on the particular angles of illumination and viewing. Bidirectional reflectance models have been developed to both take advantage of this variability and to allow the correction of the data for angular effects. A modification of one of these models was used to describe the anisotropy of quasi lambertian surfaces used for the calibration of field and laboratory instruments. Members of the project are participating in the design of data analysis algorithms to extract information from new satellite instruments.

Techniques have been developed for global tropical forest inventory using advanced very high resolution radiometer (AVHRR) and earth remote sensing satellite 1 (ERS-1) as the main sources of data. Special attention was given to the detection and monitoring of active deforestation. A comprehensive tropical forest information system has been developed to support the modelling of tropical deforestation dynamics. An intensive pan tropical data collection effort was made. Intensive observation of the equatorial belt by satellite has led in three years to the assembly of an almost adequate set of data. The problem of vegetation classifiction was addressed. Patchy AVHRR data was supplemented by intensive analysis. With single image classifiction and subsequent synthesis, most of the area of interest was covered. Maps at 1 to 3 000 000 scale were distributed to forest experts for evaluation feedback. Extensive field work has been carried out by regional specialists to support the thematic mapper (TM) analysis. Equal area estimates were found for the 20 to 30% forest cover situation; for lower cover percentage the AVHRR underestimates the true cover; for more dense cover, AVHRR overestimates the actual percentage. Other sampling approaches have been tested theoretically. A coherent geographical information system has been further developed. Regions, image windows and local data sets have been identified. Software and hardware were developed. Several preprocessing and processing steps such as calibration, spatial averaging, speckle filtering and classification have been developed in a prototype suite of tools. A geocoding/geolocation softwarepackage has been developed and is currently in a validation phase. A method of obtaining elevation data from synthetic aperture radar (SAR) interferometry has been described. The sensitivity of vegetation mapping from SAR data with respect to environmental parameters was studied and correction procedures for distortions developed.

A project has been initiated to develop remote sensing based methods for the monitoring of fires in the tropical belt. A multiscale approach (regional to global) was adopted and the development undertaken of tools and methods allowing a quantitative assessment of some effects of fire on the environment. Documenting and characterizing biomass burning patterns has been performed at 3 different scales: global, continental and regional. At global scale the project received a test version of the first international geosphere biosphere programme (IGBP) DIS global 1 km data set. A computing environment, suitable for handling these large data sets, has been created. At continental scale, fire distribution patterns over a 4 year period have been analysed, on a seasonal and interannual basis, as a function of vegetation cover and pluviometric conditions. Strong interannual variability in fire activity has been observed. It cannot be explained only by the monthly pluviometric patterns. It therefore seems to result from additional meteorological or anthropogenetic parameters. At regional scale, a field experiment performed in West Africa tested the possibilities of a mobile personal computer (PC) based high resolution picture transmission (HRPT) receiving station for fire related studies and management. The system provided in situ advanced very high resolution radiometer (AVHRR) imagery in real time: the time series thus collected were analysed for: physical characterization of controlled experimental fires; validation of algorithms for fire detection; characterization of diurnal cycle in fire activity for this part of Africa; definition of regional patterns of fire distribution over Guinea, Ivory Coast and Ghana. A methodological investigation has been initiated to study the parameterization of heterogeneity in the spatio temporal distribution of fires for the African continent. The project has described and evaluated what would be the main components of an information system related to vegetation fires and their effects at continental scale and the computing environment required for its installation.

A project has been set up to develop remote sensing based methods of biomass productivity assessment in Sahelian countries at scales ranging from the regional to the subcontinental level. The satellite data set over Niger and Burkina Faso contains over 1000 images which were pre processed for geometric and radiometric correction, and for geophysical parameter computation. Extensive sets of ground data were analysed in order to assess the relationship between green biomass, as perceived by the satellite, and grain production, which is the significance information for the end user. The study was carried out on a series of more than 2800 measurements collected directly in farmers fields. These data were collected in a wide range of ecological conditions, in 3 different countries and during 4 agricultural seasons. The main conclusion of the analysis was that there is a very wide variability between observations. The observed differences cannot be fully explained by environmental or agricultural conditions, and may be due to experimental conditions. The normalized difference vegetation index (NDVI) values derived from satellite measurements show very little difference between the agricultural and the nonagricultural domains. it was found that field data collection carried out on a village basis does significantly improve the calibration of NDVI for yield mapping. It is better to collect average yield figures at regional level, and to compare them to NDVI averages from cloud free pixels which can be found more easily over broader areas. A preliminary evaluation was carried out on the possible assessment of actual evapotranspiration (ET) from advanced very high resolution radiometer (AVHRR) data. Any accuracy assessment was carried out on a satellite derived digital elevation model, in order to better assess its usefulness for topographic description in moderately hilly regions. The resulting quality not only depends upon the geometric properties of the satellite data, but also upon the intrinsic qualities of the algorithms used. Flexibility in the processing of AVHRR data has been introduced by developing a new data and image management environment and by porting available routines into it.