Xerra News Cat

Calibrating satellite imagery using ground-based data collection

By June 11, 2019 No Comments

Satellite imagery is revolutionising how we see our world in manifold ways. Looking down from orbit has a drawback, however, as the intervening atmosphere, even when cloudless, adds some scattered sunlight for which allowances must be made.

At Xerra, in collaboration with the University of Waikato, NASA and the European Space Agency (ESA), we are gathering data in Aotearoa New Zealand that help the development of methods to remove the effect of the atmosphere. This is part of a global project to improve calibration and exploitation of satellite imagery for environmental monitoring and a range of agricultural and commercial applications.

Sensors onboard the Landsat 8 (NASA/US Geological Survey) and Sentinel-2 and Sentinel-3 satellites (ESA and the European Commission) detect sunlight reflected by the ground. This imagery can be analysed to derive important characteristics of the ground cover, such as the extent of forests and logging operations, the health of crops, the built-up area in development zones, and algal blooms in lakes, rivers and oceans.

However, the data collected by the satellites is not enough to get a full picture. Satellite operators and data providers require information collected at ground level to calibrate and validate the data collected from space. This is called ‘ground-truthing’.

How does the atmosphere impact satellite data collection?

The radiation detected from a satellite contains a large atmospheric signal – sunlight scattered by the air between the orbiting satellite and the ground (just as the sky looks bright and blue to us as we look upwards). Correction for this atmospheric scatter is required if scientists are to be able to identify the precise characteristics of the ground and water below.

Absorption and scattering by air molecules, and particles such as dust and aerosols, change the intensity and the spectral shape (colour) of light received by a satellite. These processes have to be accounted for if the satellite imagery is to be used for deriving information about the target areas below.

Correcting for atmospheric effects over inland and coastal waters is particularly challenging for various reasons, including the relatively low reflectivity of water (it looks darker than land when viewed from high above), stray-light effects from land areas adjacent to the water bodies, and also scattering of light by particles in the water which can mimic aerosol scattering.

So-called atmospheric correction is an active area of research globally, and several approaches are being developed by different teams. Some of these methods, which are applied to Sentinel-2 and Landsat 8 imagery, are formally validated and compared by an international collaborative initiative led jointly by ESA and NASA (ACIX II: the Second Atmospheric Correction Inter-comparison Exercise).

Dr Moritz Lehmann (Xerra and the University of Waikato) is participating in ACIX II by providing ground validation data from New Zealand lakes. Our nation’s remote location in the southern hemisphere provides a rare opportunity for data collection in a vastly under-sampled part of the world. Moreover, our country’s diverse environment (in terms of geology, topography, soil types, and anthropogenic pressures) produces a plethora of lake types and atmospheric conditions, which provides a useful challenge for generating globally-applicable methods for atmospheric correction.

The data Moritz has provided includes lake-surface spectral measurements of reflected sunlight (i.e., the same quantity measured by satellites but without the confusing influence of the atmosphere). This ground-truth data represents the ‘gold standard’. If the atmospherically-corrected satellite imagery matches the ground-truthing data we know that the correction method used is accurate.

Collecting ground-truth data

The trick to collecting the best ground-truth data is to be on a lake under a cloudless sky on the same day as a satellite passes over to take an image. Landsat 8 has a 16-day revisit period, and the two Sentinel-2 satellites return every five days on the same orbital path, meaning fieldwork can be planned ahead of time, in theory… the weather is the real wildcard in New Zealand.

Moritz and his team have so far collected two years of data from almost 100 lakes (Figure 1). For at least 20 lakes, the satellite orbits and weather conditions aligned to produce data suitable for ground-truthing (Figure 2).

Current international research

The information from New Zealand lakes has been combined with similar datasets from international collaborators. Developers of the atmospheric corrections methods receive only the geographic coordinates and the times of sampling from this pool of data. They will then be asked to estimate spectral reflectance at ground level from Landsat 8 and Sentinel 2 imagery.

Once the results are in, the atmospheric correction methods will be compared carefully and the results will be peer-reviewed, published and presented at various international conferences. The outcome will provide valuable guidance to the scientific community on choosing the right method for particular applications and directing future research effort.

Photograph of recording light measurements on Lake Ohau in Canterbury, New Zealand.

Figure 1: Light measurements on Lake Ohau (Canterbury) under perfect atmospheric conditions. Photo: Xerra Earth Observation Institute.

One map of New Zealand showing locations where data was collected, and a graph showing reflectance spectra from lakes at locations shown in the map.

Figure 2: Top: Map of New Zealand showing the locations of data collected simultaneously with overpasses of Landsat 8 and/or Sentinel 2 satellites. Bottom: Ground-truthing data (reflectance spectra) from lakes at locations shown in the map on the left. Credit: Xerra Earth Observation Institute.

Collaborators: Xerra, University of Waikato, NASA, European Space Agency

Funding by:
Lakes Resilience Programme (MBIE Grant UOWX1503)
Eye on Lakes Smart Ideas Project (MBIE Grant UOWX1802)
Xerra Earth Observation Institute Limited

Acknowledgements:
Dr Uyen Nguyen (post-doctoral researcher coordinating data collection)
Warrick Powrie (Technician, fieldwork)
Dean Sandwell (Technician, fieldwork)
Daniel Bellamy (summer student, field and lab work)
Caitlyn Gillard (summer student, lab work)
Moshu Xie (summer student, fieldwork)