Spectral reflectance in the Tunesian desert

Satellites provide the possibility to give a synoptical view of the earth surface at regular time intervals. Satellites operating in the optical wavelengths have however as disadvantage that monitoring of the surface characteristics becomes impossible as soon as clouds are present. Deserts and desert margins are for that reason much more appropriate for monitoring by optical satellites than temperate and wet tropical areas. Potential hazards, possibilities and often inaccessibility makes use of optical remote sensing very reasonable.Landsat Thematic Mapper (TM) satellites provide a much better spectral resolution than other satellites at reasonable spatial and temporal resolution. Ile presented research was hence focused on the spectral possibilities of Landsat Thematic Mapper for determining surface characteristics and their dynamics in desert areas. Field measurements of reflectance in different seasons were performed to evaluate the effect of different factors and their dynamics on reflectance. A field radiometer (MMR) with TM compatible bands was used. The study was performed in southern Tunisia in an area with large variation: footslopes, dunes and dynamical salt plains, all with few or absent vegetation cover. Dominant mineralogy is representative for many and areas and comprises gypsum, carbonate, quartz and halite.In order to compare results of field reflectance measurements with Landsat Thematic Mapper data, adequate processing of both data sets is necessary.General accepted assumptions that reference plates are ideal reflectors have to be rejected. Both wavelength and insolation angle dependant reflectance of the panel has to be determined. For large solar zenith angles also corrections have to be made for influence of diffuse irradiation (Chapter 2).Use of calibration coefficients given for Landsat Thematic Mapper data processed in Fucino will cause large errors in calculating reflectance data. It turned out that these coefficients were not updated for deterioration of the sensors during the flight and that calculations were made based on two different definitions of bandwidth. In order to achieve adequate values of reflectance these errors were evaluated and corrected (Chapter 8).The evaluation of factors affecting field reflectance can be separated in external and internal ones.The influence of external factors, solar zenith angle and atmosphere was evaluated (Chapter 3-4). Solar zenith angle dependant reflectance turned out to be limited for this area, if measurements are performed with solar zenith angles up to 65 degrees. In spring on the footslopes reflectance values at noon were about 10% higher than those at 65 degrees. Differences on the playa were even less. Both based on accuracy and applicability of the results (observations of this area at these latitudes with Landsat TM take place with angles ranging between 28 in June and 63 degrees in December), 65 degrees can be considered as a useful limit for performing adequate measurements.Influence of atmosphere on field reflectance is limited. The evaluation of the external factors leads to the conclusion that all measurements with a solar zenith angle less than 65 degrees and on clear days could be used for obtaining a field data reference set.Field measurements showed that Landsat TM-like bands are very useful in detecting the surface characteristics ( internal factors ) in this area (Chapters 5 - 7).Gypsum has absorption bands in both middle infrared bands (comparable with TM bands 6 and 7) and carbonate in TM band 7. Since on footslopes and in dunes quartz is the other important mineral, a high reflectance in these bands points to presence of quartz. Standard field reflectance measurements showed that on footslopes gypsum, quartz and carbonate dominated areas have a different spectral signature. Differences in eolian deposits are even more clear: relations between gypsum content and indices derived from spectral reflectance could be established under field conditions.Presence of halite on the playas could be detected by relatively high reflectance in the visible part of the spectrum, especially in the blue band. Field reflectance on plots showed that moisture content induces a relatively low reflectance in all bands with an extra low reflectance in the middle infrared bands. Linear relations between volumetric moisture content and reflectance in individual bands in near and middle infrared turned out to be feasible.Most dynamical parts of the area turned out to be the playas, where after storms moisture contents of the top layer were affected for a long time and halite efflorescences occurred shortly after the storms. Although dust slowly covering the surface, even in spring plots with higher halite content could be derived from the spectral signature. Field plots in dune parts showed a large variation in reflectance between November and May too due to changes by wind. Dynamics on footslopes were much less important than in other areas. Variation in vegetation appears to be relatively small, while also effect of storms was not visible for more than a few days after a storm.Results of field reflectance were extrapolated to Landsat TM satellite data (Chapter 9). It was possible to derive directly from Landsat TM data a number of useful classes for playas, footslopes and eolian material, having variation in surface mineralogy (gypsum, carbonate, quartz, halite) and variation in surface type. Also dynamics of factors like moisture and halite could be derived using multitemporal Landsat TM data.The presented methodology, implementing an extensive field reflectance measurement campaign, gives insight in possibilities of Landsat TM under a range of conditions. It corroborates that for operational application in and areas Landsat TM data will be a useful source of information in addition to other types of remote sensing as for instance aerial photography.