Surface Displacement Measurement from Remote Sensing Images. Olivier Cavalie

but some occlusion occurs (hidden faces, hidden streets, etc.). Spatial resolution decreases as the viewing angle increases. Hence, spatial resolutions are given at the nadir of the satellite.

      Impact of the spatial resolution: The spatial resolution of the satellite directly drives the analysis that can be performed with the images. The definition of optical missions (HR, VHR, etc.) depends on the spatial resolution:

       – 1,000–3,000 m: Analysis of the atmosphere, aerosols and land surface emissivity/temperature. For example: MSG, METOP;

       – 250–1,000 m: Analysis of the atmosphere, aerosols, land surface emissivity/temperature. For example: MODIS, Sentinel-3;

       – 30–60 m: Analysis of land cover (agriculture, forestry, cartography, geology, cryosphere, etc.). For example: Landsat1–7;

       – 15–20 m (medium resolution (MR)): Analysis of large-scale land cover, agriculture, forestry and highway infrastructures and computation of digital terrain models. For example: Sentinel-2 (red edge, shortwave infrared – 20 m);

       – 1.5–10 m (high resolution (HR)): Analysis of land cover, agriculture, regional to city-level coverage, road infrastructures, digital elevation model extraction and medium or large objects, such as boats, and computation of digital terrain models. For example: SPOT-6–7, Sentinel-2 (blue, green, red, near-infrared – 10 m);

       – 0.5–1 m (very high resolution (VHR)): Analysis of many objects in the images. At this resolution, a large number of objects are visible, such as urban elements, houses, vehicles, buildings affected by natural disasters and archeological objects. Extraction of digital surface models can also be undertaken. For example: Pléiades;

       – < 0.5 m (ultrahigh resolution (UHR)): Detailed analysis of elements of the scene. Many objects are visible in the images with more details, including small vehicles, buildings, archeological objects, etc. Computation of digital surface models can be undertaken. For example: WorldView-3, Cartosat-2.

      The main optical satellites that are used for DEM extraction have spatial resolutions better than 10 m. For precise estimation of DEMs, submetric satellites are the best candidates.

      Resolution versus swath versus revisit time: The size of telescope mirrors was the limiting factor in reaching high resolutions for the first generations of optical satellites. With the evolution of space technology, the first civil satellites were launched at the beginning of the 2000s with resolutions lower than 1 m. Ikonos was launched in September 1999 and was the first commercial satellite with a resolution of less than 1 m; it was followed by satellites such as Quickbird-2, Eros B, Kompsat-2 and WorldView-1. A satellite with a very high resolution has a reduced swath, in contrast to decametric resolution satellites. A medium-resolution satellite usually has a wider swath and a shorter revisit time. Conversely, meteorological satellites such as the Meteosat Second Generation (MSG) and Meteorological Operational Satellite (MetOp) family usually have a low spatial resolution around 1 or 3 km, with a wide swath between one and several thousands of kilometers, in order to give a view of the meteorological situation. The number of spectral bands, from 500 nm to 15,000 nm, allows a fine analysis of clouds and aerosols. These satellites have a low spatial resolution but a high temporal resolution. Their spatial resolution is not compatible with DEM extraction.

Choice of the spectral bands: The choice of spectral bands combined with the spatial resolution directly drive the applications of a satellite. Many satellite systems exist with many different configurations (PAN+XS configuration, only PAN configuration, only XS configuration, etc.). A significant number of UHR and VHR satellites have four bands (blue, green, red and near-infrared) associated with a panchromatic band. The panchromatic band (P or PAN) usually has a spatial resolution four times higher than the multispectral ones (XS). Pan-sharpening techniques are employed in order to build panchromatic plus multispectral (P+XS) products, which means fused multispectral products at a panchromatic spatial resolution.

Table 1.5. Description of spectral bands

Band name	Spectral interval
UV band	300–400 nm
Visible bands (VIS)	400–700 nm
Near-infrared (NIR) bands	700–1,000 nm
Short-wavelength infrared (SWIR)	1,000–2,500 nm
Medium-wavelength infrared (MWIR)	3,000–5,000 nm
Thermal infrared (TIR)/long-wavelength (LWIR)	8,000–15,000 nm

      High-resolution (HR) satellites work with lower resolution but include additive infrared bands useful for land cover, water color or atmospheric analysis. To compute digital elevation models and to measure displacements, the spectral band with the highest spatial resolution and the best SNR is generally chosen. For many satellites, this is usually the panchromatic band. For some HR systems with only XS bands, multispectral bands may be used (e.g. Sentinel-2).

1.2. Past and on-going SAR missions

      This section presents spaceborne missions that have been used or are used to compute displacement fields by interferometric processing. Some peculiarities or issues are pointed out to help the reader understand them. This is thus not an exhaustive list of SAR space missions. Note that the Magellan mission began to map Venus in September 1990 with an SAR instrument, but its topography was computed using a radargrammetry technique, combining two images from either the same side or from opposite sides: this is rather far from the topic of this book, and thus this very successful mission will not be further detailed here. We must note that most of these missions were organized and designed before the interferometric technique emerged, around the beginning of the 1990s. Thus, certain constraints, such as orbit housekeeping, yaw steering and burst synchronization for ScanSAR modes, were sometimes not taken into account in the designs of the spacecraft or payload.

      After a series of C-band and L-band satellites at medium resolution in the 1990s, a new generation of SAR satellites operating in the X-band arrived in Europe at the national level (Germany, Italy, Spain) from mid-2007 onwards: although targeting different goals, they all had significantly improved resolutions and repeat pass capacities, which gave them serious advantages when focusing on localized areas.

      With small satellite missions, such as ICEYE, interferometry is usually possible with a short time period: maintaining a small orbital tube is too expensive for the length of the mission and archives on large scales are out of scope; however, such constellations can gather significant data over cities and PS techniques can be considered. NovaSAR is not addressed here, as it is a single satellite on a short-term mission with limited applications for interferometry. Some Chinese and Indian missions are also not addressed due to the limited availability of the datasets. Finally, a significant improvement in interferometric abilities and coverage arrived with missions such as Sentinel-1 and ALOS-2.

      1.2.1. ERS-1, ERS-2 and Envisat

      ERS-1 (1991–2000) and ERS-2 (1995–2011), the European remote sensing satellites, began the era of SAR instruments in Europe and demonstrated the first applications in radar interferometry. The five-year overlap provided the opportunity to develop very nice orbital combinations, such as the so-called tandem phase, allowing acquisitions of the same area within just a one-day interval, which was much more favorable than the native 35-day repeat cycle for InSAR on rapidly changing surfaces.

      The Environmental Satellite (Envisat) continued the C-band SAR missions for Europe with