Chavez PS. An improved dark-object subtraction technique for atmospheric scattering correction of multispectral data. Remote Sensing of Environment

U.S. Geological Survey, 2255 N. Gemini Drive, Flagstaff, Arizona 86001 USA
Remote Sensing of Environment (Impact Factor: 6.39). 04/1988; 24(3):459-479. DOI: 10.1016/0034-4257(88)90019-3

ABSTRACT Digital analysis of remotely sensed data has become an important component of many earth-science studies. These data are often processed through a set of preprocessing or “clean-up” routines that includes a correction for atmospheric scattering, often called haze. Various methods to correct or remove the additive haze component have been developed, including the widely used dark-object subtraction technique. A problem with most of these methods is that the haze values for each spectral band are selected independently. This can create problems because atmospheric scattering is highly wavelength-dependent in the visible part of the electromagnetic spectrum and the scattering values are correlated with each other. Therefore, multispectral data such as from the Landsat Thematic Mapper and Multispectral Scanner must be corrected with haze values that are spectral band dependent. An improved dark-object subtraction technique is demonstrated that allows the user to select a relative atmospheric scattering model to predict the haze values for all the spectral bands from a selected starting band haze value. The improved method normalizes the predicted haze values for the different gain and offset parameters used by the imaging system. Examples of haze value differences between the old and improved methods for Thematic Mapper Bands 1, 2, 3, 4, 5, and 7 are 40.0, 13.0, 12.0, 8.0, 5.0, and 2.0 vs. 40.0, 13.2, 8.9, 4.9, 16.7, and 3.3, respectively, using a relative scattering model of a clear atmosphere. In one Landsat multispectral scanner image the haze value differences for Bands 4, 5, 6, and 7 were 30.0, 50.0, 50.0, and 40.0 for the old method vs. 30.0, 34.4, 43.6, and 6.4 for the new method using a relative scattering model of a hazy atmosphere.

Download full-text


Available from: Pat S. Chavez, Jr., Feb 09, 2015
  • Source
    • "Song et al., 2001). In an ideal situation, a radiometrically 'dark' object (e.g. a clear water body) produces zero radiance in all wavelengths and hence any radiance received at the sensor for a dark object pixel is due to atmospheric path radiance (Chavez, 1988). Thus, pixels over Lake Albert containing the lowest Digital Number (DN) values were selected (as the dark objects), and their representative value was subtracted from the DNs across the whole scene to reduce scattering influences (Song et al., 2001). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Deforestation within and outside protected areas is widespread in Western Uganda, but quantification of such forest changes is rare. In this study, spatio–temporal forest cover patterns in the Northern Albertine Rift Landscape were reconstructed for the period 1985–2014, over a range extending from Bugoma forest in the South of the region to as far as Murchison Falls National Park in the North, an area approximately 225km North-South by 63km East-West. We examine both the spatial and temporal heterogeneity of the land cover changes. Seven 30 x 30m resolution, ortho–rectified, cloud–free Landsat images obtained from the USGS archive were analysed at the landscape– and three smaller scales. Forest classification using Landsat imagery appears robust; similar amounts are obtained from a UK-DMCii image (22m resolution) taken a day before the Landsat scene in Dec, 2010. Our results show that larger–scale aggregate measures of total change can obscure more local patterns, in which protected areas and the national park maintain or grow forest cover, whilst the forest corridor areas that are not protected suffer drastic losses. Time–series show that the loss continues nearly linearly into the present around Bugoma, but seems to level off around Budongo Forest after 2010, apparently because almost all forested corridor areas have been cleared. At the landscape scale, between 1985 and 2014, the data suggest approximately 0.4% of initial cover was lost per year. However, this was mostly a result of the large protected forest blocks remaining relatively stable; deforestation was mostly situated in the corridor and riverine areas. Local–scale losses were most prominent in unprotected forests around Budongo and Bugoma, with annual losses at a much higher average rate about of 3.3% per year in each case. The annual rates of loss are higher than Uganda’s average (1-3%). Forest cover in the protected zones expanded only marginally, with annual average increases of order 0.03% and 0.5% in Budongo and Bugoma reserves respectively. Our results suggest that forest protection in the gazetted areas is successful, and the protection policy is working, but these forests are being isolated by large losses immediately outside the protected zones, in the forest corridors. This may have severe social and ecological consequences – both within and outside protected forests.
    Land Use Policy 08/2015; 49:236-251. DOI:10.1016/j.landusepol.2015.07.013 · 3.13 Impact Factor
  • Source
    • "Atmospheric and radiometric corrections to improve the image quality and the accuracy of the analysis were applied (Chavez, 1988, 1996). In order to detect and identify mobile/stabilized sands and improve decision making in the classification procedure, the indices and calculations specified in Table 2 were applied to the ETM+ images. "
    Dataset: AR SM
  • Source
    • "Path radiance (L khaze ) was obtained from pixels of known zero reflectance such as large deep water bodies and it was assumed that any value in the raw image in these areas other than zero represented haze effects. This value was subsequently verified through a band specific histogram analysis by examining the step in pixel radiance values (Chavez Jr., 1988). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Accurate modelling of leaf chlorophyll content over a range of spatial and temporal scales is central to monitoring vegetation stress and physiological condition, and vegetation response to different ecological, climatic and anthropogenic drivers. A process-based modelling approach can account for variation in other factors affecting canopy reflectance, providing a more accurate estimate of chlorophyll content across different vegetation species, time-frames, and broader spatial extents. However, physically-based modelling studies usually use hyperspectral data, neglecting a wealth of data from broadband and multispectral sources. In this study, we assessed the potential for using canopy (4-Scale) and leaf radiative transfer (PROSPECT4/5) models to estimate leaf chlorophyll content using canopy Landsat satellite data and simulated Landsat bands from leaf level hyperspectral reflectance data. Over 600 leaf samples were used to test the performance of PROSPECT for different vegetation species, including black spruce (Picea mariana), sugar maple (Acer saccharum), trembling aspen (Populus tremuloides) and jack pine (Pinus banksiana). At the leaf level, hyperspectral and simulated Landsat bands showed very similar results to laboratory measured chlorophyll (R2 = 0.77 and R2 = 0.75, respectively). Comparisons between PROSPECT4 modelled chlorophyll from simulated Landsat and hyperspectral spectra showed a very close correspondence (R2 = 0.97, root mean square error (RMSE) = 3.01 μg/cm2), as did simulated reflectance bands from other broadband and narrowband sensors (MODIS: R2 = 0.99, RMSE = 1.80 μg/cm2; MERIS: R2 = 0.97, RMSE = 2.50 μg/cm2 and SPOT5 HRG: R2 = 0.96, RMSE = 5.38 μg/cm2). Modelled leaf chlorophyll content from Landsat 5 TM canopy reflectance data, acquired from over 40 ground validation sites, demonstrated a strong relationship with measured leaf chlorophyll content (R2 = 0.78, RMSE = 8.73 μg/cm2, p < 0.001), and a high linearity with negligible systematic bias. Study results demonstrate the small number of input bands required for PROSPECT inversion and provide a theoretical and operational basis for the future retrieval of leaf chlorophyll content using broadband or multispectral sensors within a physically-based approach.
    ISPRS Journal of Photogrammetry and Remote Sensing 04/2015; 102. DOI:10.1016/j.isprsjprs.2015.01.008 · 3.13 Impact Factor
Show more

Questions & Answers about this publication

  • Pat S. Chavez, Jr. added an answer in Remote Sensing Applications:
    How does sun irradiance affect satellite imagery, and how does it vary among different satellites?

    Why do we need to account for top-atmosphere reflectance?

    This is in relation to the methods by Stapleton, et. al (2014) from Polar Bears from Space: Assessing Satellite Imagery as a Tool to Track Artic Wildlife.

    Pat S. Chavez, Jr. · Northern Arizona University (retired USGS)


    Depending on the application radiometric calibration (including corrections for Earth-Sun distance and atmospheric conditions, plus sun elevation angle) may or may not be needed.  For most, not all, applications that use multi-temporal or multi-sensor data sets radiometric calibration are needed and/or help with the analysis.

    Of course, there are various methods that have been developed over the last few decades for radiometric calibration and the links below are to 2 papers that discuss  the DOS and COST methods that I developed.  You may want to take a look at them and see if perhaps they may apply for your data sets and applications.

    Good Luck,