3
1. Introduction
MAIAC is a new advanced algorithm which uses time series analysis and a combination of pixel-
and image-based processing to improve accuracy of cloud detection, aerosol retrievals and
atmospheric correction (Lyapustin et al., 2011a,b; 2012; publication on current MAIAC is under
preparation). The underlying physical idea behind MAIAC is simple: because surface changes
slowly in time compared to aerosols and clouds given the daily rate of global MODIS observations,
we focus on extensive characterization of the surface background in order to improve all stages of
MAIAC processing. MAIAC starts with gridding MODIS measurements (L1B data) to a fixed grid
at 1km resolution in order to observe the same grid cell over time and work with polar-orbiting
observations as if they were “geostationary”. In this regard, this approach is fundamentally different
from the conventional swath-based processing where the footprint changes with orbit and view
geometry (scan angle) making it difficult to characterize an always changing surface background.
To enable the time series analysis, MAIAC implements the sliding window technique by storing
from 4 (at poles) to 16 (at equator) days of past observations in operational memory. This helps us
retrieve surface BRDF from accumulated multi-angle set of observations, and detect seasonal (slow)
and rapid surface change. A detailed knowledge of the previous surface state also helps MAIAC’s
internal dynamic land-water-snow classification including snow detection and characterization.
Consistently with the entire C6 MODIS land processing, the top-of-atmosphere (TOA) L1B
reflectance includes standard C6 calibration (Toller et al., 2014) augmented with polarization
correction for MODIS Terra (Meister et al., 2012), residual de-trending and MODIS Terra-to-
Aqua cross-calibration (Lyapustin et.al, 2014). The L1B data are first gridded into 1km MODIS
sinusoidal grid using area-weighted method (Wolfe et al., 1998). Due to cross-calibration, MAIAC
processes MODIS Terra and Aqua jointly as a single sensor.
2. Overview of MAIAC products
MAIAC provides a suite of atmospheric and surface products in HDF4 format, including: (1) daily
MCD19A1 (spectral BRF, or surface reflectance), (2) daily MCD19A2 (atmospheric properties),
and (3) 8-day MCD19A3 (spectral BRDF/albedo).
2.1 Tiled File Structure and Naming Convention
Products are generated on a 1km sinusoidal grid. The sinusoidal projection is not optimal due to
distortions at high latitudes and off the grid-center, but it is a tradeoff made by the MODIS land
team for the global data processing. The gridded data are divided into 1200x1200km
2
standard
MODIS tiles shown in Figure 1.
The current dataset presents data per orbit (we do not provide a daily composite image as in
standard MODIS surface reflectance product MOD09). Each daily file name follows the standard
MODIS name convention, for instance:
MCD19A1.DayOfObservation.TileNumber.Collection.TimeOfCreation.hdf.
DayOfObservation has the format “AYYYYDDD”, where YYYY is year, DDD is Julian day.
TileNumber has the standard format, e.g. h11v05 for the east coast USA.
