Remote Sensing from Space

Remote sensing from space is only able to retrieve a subset of aerosol properties, mainly AOD and sometimes its spectral dependence, for AnP estimates. These retrievals require a priori assumptions of other aerosol properties (at least aerosol absorption) and environmental properties. Without sufficient accuracy of these assumed properties (or choices in lookup tables), aerosol retrievals are too uncertain to be useful. An example of this is the common failure to retrieve AOD from visible reflectance data over brighter surfaces (e.g., desert, sun glint). In addition, sampling is spatially or temporally limited, depending on the satellite orbit. Over the equator, (geo)stationary platforms have the potential for high temporal sampling; however, because of the geometric aspects of their position, remote sensing is restricted to lower latitudes, and at least five different satellites with five different sensors are needed to achieve global coverage. Polar-orbiting platforms, in contrast, achieve global coverage with a single sensor and thus have been a preferred choice for aerosol sensor platforms. With limitations to the swath, polar orbiters provide at best a single daytime overpass. Temporal sampling frequency is further reduced by the aerosol retrieval requirement for a cloud-free scene. This is particularly a problem if the region associated with the smallest pixel size is large. For polar orbiters, the number of successful samples becomes often so sparse that statistical significance is not guaranteed, even for monthly averages at a spatial resolution of 1°x1° longitude/latitude. Thus, for typical AOD maps from polar-orbiting remote sensing, only the available multiyear sensor datasets are considered and compared to AERONET statistics (Figure 3.1). For the displayed multiannual averages, periods of enhanced stratospheric aerosols are excluded.

Figure 3.1 demonstrates the significant differences that exist among regional annual values for AOD. The retrieved dataset is sensitive to retrieval assumptions, as demonstrated by comparisons of different processing versions for AVHRR, MODIS, and TOMS. For example, a change to the ocean reflectance in the TOMS retrieval causes a switch from formerly the largest to now the smallest background AOD over oceans. Obviously, the retrieval performance in reference to AERONET varies regionally. To compensate, a satellite retrieval composite for AOD has been developed based on the best overall scores obtained regionally for bias, spatial variability, and seasonality. The combined product appears superior over any individual satellite dataset. The resulting seasonal AOD distribution is presented in Figure 3.2.

Even the best regional AOD satellite retrievals of this composite are far from perfect. Moreover, next to AOD, satellite retrievals provide only limited information on particle size and almost never on particulate absorption. In retrievals developed for different satellite sensors, these assumed aerosol properties as well as their environmental properties are often inconsistent. This complicates comparisons of retrieved AOD values and can be responsible for qualitative differences (e.g., different spatial patterns). Improvements are expected with sensor standards, sensor side-by-side deployments on the same platform, and consistency in assumptions for ancillary and/or a priori data. Useful insights on retrieval capabilities and limitations are expected from careful comparisons to quality reference data (e.g., sun/sky photometry). However, this requires a commitment to compare, evaluate, and reprocess the data plus the necessary time.

0 0

Post a comment