satellite images are always cloud free They aren’t! What effect does a cloud have on an image? • Underneath the cloud • Simple! • Next to the cloud • Harder! Shadows? • Further away from the cloud? • Harder! How big is this effect?
Simulate the passage of light through the atmosphere • Input: Atmospheric conditions, TOA solar intensity • Simulate: Absorption & Scattering by Aerosols & Gases • Output: Intensity at ground ( -> intensity at sensor) • Examples: • 6S • MODTRAN (-> ATCOR, MODO) 3
by which the light intensity is reduced • Optical depth – Used to calculate the transmittances All dependent on wavelength All we need to do is calculate the optical depths! Fundamental equations of RTMs 4 T = e ⌧ Iground = IT OA T1 T2 T3 T4 . . . Tn
model – Have to deal with each ray of light individually as generalised mathematics won’t work for horizontally heterogeneous atmospheres – Run for 10,000’s of rays • Based on a 2D atmospheric grid • Exploratory model – for understanding, not accurate outputs • Based on a very simple RTM – SPCTRAL2 (Bird and Riordan, 1984) 7
molecules (small!) – Probability: • Mie – Aerosols (large!) – Probability: More constant with wavelength, depends on aerosol type and amount • Each cell given a probability, and directions calculated from phase functions 11 / 1 4
General atmospheric gases (O, N, CO2 ) – constant – Water vapour and Ozone – spatially variable • Method – Assign each cell an amount of water vapour and ozone – Sum for each cell the ray passes through – Add on other gases at the end (proportional to path length) 13
has – Water vapour content – Aerosol type (Maritime, Urban, Cloud etc) – Aerosol amount (particles per cell) • Can add a cloud (generated from a cellular-automaton model) anywhere in the grid – just change parameters for those cells • Parameterisation developed from standard profiles, plus random element 14
away • Used a large grid • Moved cloud away from sensor • Location defined by the Cloud Location Ratio: – Designed to allow transfer to the real world • Conversion to standard remote sensing output: NDVI 21 CLR = Horizontal distance from sensor to cloud Vertical height of cloud base
mean? • Converted NDVIs to Net Primary Productivity (Ruimy et al., 1994) • Maximum difference in NDVI • Reduction of NPP by 2.8g/m2 – Over a whole Landsat scene that is 89,000 tonnes! • The CLR suggests this could happen even when the cloud is a long way away – CLR of 16, with real-world height of 2000m, gives distance of 32km! 23
for this! Possible methods are under research…. 24 Cloud here Could affect irradiance here Causing a change of up to 5% in NDVI Nothing is done about this at the moment – clear skies are always assumed!