Dr Ireneusz Baran
Spatial Information Specialist
Email: [email protected]
In the last decade, applications of digital terrain models (DTM) and digital surface models (DSM) have rapidly increased. DTMs are routinely used in engineering applications and environmental studies, risk analysis and disaster monitoring. Growing demand and technological advances have accelerated the development of new techniques capable of delivering rapid, high resolution and accurate terrain definitions.
Both LiDAR (Light Detection and Ranging, also known as Airborne Laser Scanning) and InSAR (Interferometric Synthetic Aperture Radar, also known as IFSAR) are active sensors which transmit pulses of electromagnetic energy and record the backscattered signal to derive spatial location of the survey target.
LiDAR is now well known and accepted in the commercial environment, with many papers describing its technology, benefits and applications (eg. Jonas, 2007). Although InSAR has also been around for decades, it is being considered only now for adoption in conventional mapping applications.
InSAR and LiDAR explained
InSAR sensors are usually installed on a fast moving aircraft capable of flying at high altitudes. Usually two-side looking radar antennas (separated by known baseline) are mounted. In this "single pass" configuration, one antenna transmits radio waves and both antennas receive backscattered signal (Fig 1B). Such a configuration enables the system to scan the same target simultaneously from two different antenna positions. Advanced SAR data processing enables the system to generate a pair of high resolution images/2009/june of the same scene. Each pixel preserves amplitude and phase of the backscattered signal. This information is exploited in the interferometry process where both images/2009/june are differentiated. The resulting phase differences are then unwrapped and converted to heights and finally a DSM (Digital Surface Model). Although it is possible to process images/2009/june acquired at different times ("repeat pass" configuration), simultaneous acquisition has significant advantages as it mitigates temporal decorrelation to improve data quality. There are only a few commercially available airborne InSAR systems.
The commercially available GeoSAR systems operate Xand P-band simultaneously on both sides of the airplane. Moreover, the recently developed Ku-Band InSAR system has spatial resolution of 0.3m and vertical accuracy better than 0.5m (Okada et al, 2007).
The radar pulse is typically transmitted at a 20 to 50 degree look angle. As the pulse spreads across the flying path, it hits targets along its way and the system records the corresponding returns (Fig 1B). This means that different targets positioned at the same distance from the sensor cannot be resolved. This is known as foreshortening and layover. These phenomena, together with shadows and multipath of radar signal, are the major limitations of InSAR system.
InSAR sampling cell contains individual (volume) scatters on the ground and above. This causes the 'noisy' nature of InSAR elevation data and often introduces unwanted biases. It is especially problematic in urban and forested areas where, for example, a building or tree together with ground is present in a single sampling cell.
Fig 1. (A) LiDAR and (B) InSAR basic acquisition geometry
The long wavelength of the InSAR system offers its biggest advantage as it can penetrate through clouds, haze and dust. This means that InSAR can operate virtually under all weather conditions. By using different wavelengths, the system penetration capabilities can be altered. For example, the X-band will reflect from the vegetation where P-band will penetrate to the ground.
InSAR data processing appears more complex than LiDAR, even though this technology is relatively well developed and the processing algorithms very robust. The processing of InSAR data requires highly sophisticated software and – as with LiDAR – trained data processing personnel. Some system and processing induced errors are still problematic and present in the data as noise. Advanced data filtering techniques are applied to improve the data quality (eg. Baran et al. 2003).
Although both systems are capable of recording more data along a single range (LiDAR records more than one return; InSAR uses multiple frequency and/or polarisations) in principle, they capture digital surface definition. Further processing is required to extract digital terrain definition. Many algorithms have been developed to automate this process. However, none of the available algorithms are fully reliable and expensive and time consuming manual data filtering and QA are still required.
Fig. 2 LiDAR derived DTM. Sample size: 2x10km (after Norheim et al, 2002)
Fig. 3 InSAR Derived DTM. Sample size: 2x10km (after Norheim et al, 2002)
Extracting the bare ground from under the vegetation is problematic and both systems tackle this problem differently. X-band InSAR system will not penetrate the canopies and reach the ground. Moreover, its large resolution cell requires much large patches of unobscured ground in order to capture it reliably. Using longer waves such as P-band allows the InSAR systems to penetrate the vegetation and often deep into the ground introducing some unwanted biases. P-band is also more difficult to operate and process as it is less immune to interference and attitude errors. LiDAR's very narrow laser beam is more effective. LiDAR's nadir looking configuration and ability to pass between the gaps in the canopy allows the system to penetrate to the ground even through thick vegetation canopies.
Data characteristics and accuracy
The vertical accuracy of LiDAR derived DTMs is typically in the range of 0.10m-0.20m (one sigma) and depends on many systematic errors as well as data calibration and classification. If carefully planned and properly calibrated, a LiDAR system can achieve vertical accuracy better than 0.10m. InSAR too can achieve this accuracy but the high costs involved limits its application.
Both systems are sensitive to terrain variations and land cover. LiDAR derived DTMs will be less accurate under heavy vegetation than on clear ground due to reduced laser penetration. The accuracy of InSAR derived DTMs will significantly depend on the band used, terrain variation and land cover. If X-band is used, then there will be no penetration of the vegetation so the "DTM" can only be derived by taking the tree heights and subtracting an estimated tree height. If P-band is used, the penetration of the vegetation is possible, however, the measured range is less accurate. Moreover, P-band tends to penetrate into the ground.
The average error of InSAR derived DTM over vegetated areas may be several times larger than on open ground. This is primarily caused by contributions of many individual scatters being within a much larger resolution cell. However, it is also due to the fact that radio waves interact differently with different materials depending on their conductivity properties.
There are two important parameters that determine DTM suitability for a specific application: spatial resolution and accuracy. Usually, accuracy refers to the vertical and horizontal aspects. However, as the horizontal accuracy (especially absolute) may not be critical, often DTM accuracy refers to the vertical component only.
Spatial resolution and vertical accuracy (rather than acquisition method) should always decide the methodology to meet the application requirements. This 'bottom up' approach will quickly differentiate between different products and ultimately help select a preferred data acquisition technique. In cases when both LiDAR and InSAR accuracies are suitable, differences in data characteristic caused by different acquisition method, as explained in the previous paragraphs, as well as acquisition costs, should be used as a guide.
The two aerial survey techniques generally complement each other along a continuum of requirements. LiDAR is the preferred technology at the high end of the accuracy scale: for engineering applications, earthwork volumes, drainage studies, where localised terrain shapes, vegetation penetration and high degree of reliability is required. InSAR is the preferred technology at the lower end of the accuracy scale: for topographic applications, national mapping programmes, conceptual planning, where high reliability of terrain heights is not cost justified over very large areas.
In between sit those projects which require a closer costbenefit analysis. A flood study's whole of catchment DTM can often be satisfied by InSAR at the regional level, but the LiDAR survey will likely be required later anyway if the project has to move to an engineering component or if modelling at a specific location is required. InSAR is valuable when deciding broad planning of remote road corridors … which side of the mountain should one use … but LiDAR will be required if and when the project requires accurate earthwork volumes. InSAR will provide overall DSM information of a cityscape, but the project may benefit from finer definitions of individual buildings, trees or actual powerline conductors. (Refer to Gamba et al 2003 for detailed comparison.)
Like all survey planning, the geospatial professional should analyse the requirements and available budget to decide the most appropriate survey technique(s).