A GEO-LEO (Geostationary and Low-Earth-Orbiting) Synergistic Approach to Aerosol Mapping: Towards a More Detailed and Higher Accuracy Aerosol Spatial and Temporal Distributions

Aerosols have a multitude of impacts on society, from daily life, to our environment and the climate. Aerosols remain one of the largest sources of uncertainty in climate modelling. Air pollution in the form of aerosols have major impacts on human health. Aerosols provide critical information for the correction of atmospheric effects on remote sensing imagery. Remote sensing provides a practical means to measure aerosols at global scales and at high spatial resolution. The new generation of geostationary satellites such as Himawari-8/9, HY-4, GOES-R and MTG-I, presents new opportunities to map aerosols in high temporal resolution with potentially high accuracy. However, aerosol mapping over land, especially over bright land such as the Australia continent, has been challenging due the dominating land surface which is highly variable over time, sun and view angles, and spectral wavelength. This work presents a multi-satellite approach combining GEO-LEO satellites to simultaneously retrieve aerosols and surface BRDF (Bidirectional Reflectance Distribution Function) over land. GEO-LEO Virtual Dual View (VDV) Sensor Construction A major challenge in aerosol retrieval is insufficient measurements from remote sensors, especially the most commonly available single-view sensors. One way to overcome this difficulty is by combining multiple satellites to effectively increase the simultaneous measurements. In this work, a procedure has been developed to construct virtual dual-view sensors to acquire dualview observations, similar to a physical dual/multiview sensor such as AATSR and MISR. This is Figure 1 Flow chart showing the process implemented by Qin and McVicar (2018) to construct virtual-dual-view sensors from GEOLEO sensors. White boxes represent the input images, and green texted boxes are the outputs of the process. achieved by pairing AHI (Himawari-8/9) with low-earth-orbiting sensors such as MODIS. Due to the much higher (10 minutes) temporal resolution of the AHI, it is now possible to match the sensors within a very short time difference less than 5 minutes. However, the spectral bands of the involved sensors are very different, and the radiometric calibration may also be different. To reduce these differences, a procedure to unify the spectral bands of the sensors followed by a radiometric intercalibration has been developed (Qin and McVicar 2018). Three virtual dual-view sensors, AHIMODIS/Aqua, AHI-MODIS/Terra and AHI-VIIRS, have been constructed, with AHI-SGLI (GCOM-C) to be completed. The two-step procedure to construct VDV sensors is outlined in Figure 1 and discussed below. In the first step, over 1270 representative Hyperion passes, uniformly spreading over the Australian continent, were used to simulate the spectral bands of the GEO and LEO sensors. These Hyperion images are aggregated from their original 30 meter resolution to 960 meters, approximating the resolution of MODIS and VIIRS, resulting in a total 1.3 million sample points. All valid sample pixels, including those over waters and a small amount of residual clouds, are used and together they represent a large range of spectral types. Using these simulated simultaneous GEO-LEO observations, relationships to estimate the radiance ( ) of each of the five GEO bands (1 to 5) using one or more nearby LEO bands were established. Table 1 shows the LEO bands used, the conversion coefficients and the conversion error matrix. Among all the GEO-LEO band combinations, a maximum relative error of 2.74% at 95% of the sample points was found, and the majority (9) of the 15 cases have maximum relative error less than 1% at 95% sample, and only 2 are above 2%. The mean bias is negligible for all cases. The representativeness of the Hyperion sample, the low regression error and negligible bias indicate that the conversion is reliable and accurate over the whole continent. In the second step, ray-matching was used to collect near-simultaneous observations from each of the GEO-LEO sensor pairs. Around the GEO’s sub-satellite point, and for each LEO orbit, there is a moment when the observed point and the LEO and GEO satellites are linearly aligned providing a pair AHI Bands (μm) LEO Sensor & Bands Conversion Coefficients Mean Error Mean Bias 95% Error Abs. 104 Rel. % Abs. 104 Rel. % Abs. 104 Rel. % MODIS/Aqua 0.47 b03 b04 -0.5969 0.9748 0.0316 3.9 0.25 -0.7 -0.02 10.2 0.65 0.51 b11 b10 -0.9021 0.4864 0.5170 8.0 0.53 -0.5 0.04 22.4 1.61 0.64 b01 b04 -0.0342 0.9544 0.0541 3.0 0.17 0.0 0.00 9.5 0.44 0.86 b02 -0.0022 1.0041 1.4 0.07 0.0 -0.02 3.8 0.19 1.60 b06 b07 -0.0506 0.9423 0.1630 12.8 0.79 1.1 -0.17 32.7 2.57 MODIS/Terra 0.47 b03 b04 -0.6256 0.9736 0.0344 4.2 0.26 -0.4 -0.00 10.8 0.68 0.51 b11 b10 -0.9015 0.4963 0.5065 8.0 0.53 -0.5 0.04 22.5 1.62 0.64 b01 b04 -0.0335 0.9525 0.0564 3.2 0.18 0.0 0.00 10.2 0.47 0.86 b02 -0.0066 1.0040 1.6 0.05 0.0 -0.00 5.0 0.13 1.60 b06 b07 -0.0526 0.9408 0.1717 13.8 0.85 1.2 -0.17 35.4 2.74