A strategy for detection and measurement of the cliff retreat in the coast of Algarve (Portugal)

Sea cliff retreat is mainly caused by the occurrence of slope mass movements of different types and sizes, which are a significant constraint for human activities and a source of considerable natural hazard and risk.
With the increasing use of cliffed coastal areas in recent decades, mainly with houses, urban areas and leisure resorts, this problem has growing importance in many areas of the world, in terms of natural risk reduction and of related environmental, landscape and heritage preservation issues.
By analogy with landslide hazard assessment, cliff instability hazard assessment should include spatial, time and magnitude components, computed with mainly empirical models based on the analysis of systematic inventories of past events, which are scarce in the literature. The Algarve cliffs are one exception, because they were covered by detailed studies based on systematic comparison of aerial photos from 1947 to 1991 performed with simplified methods. The methods used provided a very rich database, but are very dependent on the skill of the user and do not enable the assessment of the measurement errors involved.
Current digital photogrammetric software packages enable the enhancement of multi-temporal studies, provided that special procedures are used to deal with problems involved in the use of old aerial photos, where usually the necessary complementary information is not available.
To overcome these problems and to set up general use methods for cliff instability identification and measurement with the required accuracy, a photogrammetric study was made in the 16 km long Burgau-Lagos cliff dominated coast in southern Algarve (Portugal). From west to east, the cliffs are composed by alternating beds of Cretaceous marly limestones and marls, overlain by Miocene weak calcarenites, heavily affected by karst features, which are partially filled with Plio-Pleistocene sandy deposits. The cliff height is variable between 20m and more than 100m, and corresponds mainly to irregular cross profile slopes, with a general dip of 60º to 80º in the Cretaceous rocks, and near vertical with frequent overhanging sections in the Miocene rocks.
The frequency of cliff instabilities is not very high (45 events between 1947 and 1991), providing average cliff retreat rates of 10-2 to 10-3 m/year. As an indication of the magnitude of the cliff failures to identify and measure, the horizontal area lost at the cliff top in each event varied between 3m2 and 76m2, and volumes of displaced material between 50m3 and 12000m3.
Available aerial surveys covering the area date from 1938-1948 (no precise date available), 1947, 1958, 1972, 1980, 1983, 1991, 1995 and 2002, with different scales and quality. The available complementary information includes camera calibration certificates for the more recent surveys and a set of ground control points identifiable in the 2002 flight.
The 1938-48 survey is fundamental for this study because it is the older one and its scale (1:17,000) is much more favourable than the one of the 1947 survey (1:30,000). Unfortunately, this flight presents a lot of obstacles for precise photogrammetric processing. It has a non-standard format of 18x18cm, hardly visible fiducial marks, poor contrast, no calibration certificate, and the original films are not available, so the existing paper copies had to be digitized. Fortunately, the focal distance of the lens was recorded in the frames. The survey is composed by several parallel short strips (2 to 3 photos each) in North-South direction across the shoreline and in many photos more than half of the photographed area corresponds to the sea.
Due to urban development in the region, the landscape changed significantly throughout the decades and most of the 2002 ground control points are not identifiable in the older surveys.
To assess the exterior orientation of the photos, a multi-temporal aerotriangulation has been calculated by means of BLUH (Leibniz University Hanover aerotriangulation program) using the 2002 survey ground control and identical object points in two or more flights.
To detect changes in the cliff morphology, digital terrain models could then be automatically generated for the several epochs, using the photogrammetric software LISA and the orthoengine module from PCI Geomatica. After harmonizing the resolution, a comparison reveals the areas of change. Cliff top, ridges and toe will then be stereo plotted, in order to quantify the changes after refining the local elevation models, and also to provide the shapes of the failed rock masses.