Alexandre Gonçalves

Hydrodynamic simulation of estuaries requires a single digital elevation model (DEM) resulting from merging of both topographic and bathymetric data. These two datasets are usually produced using different technologies, co-ordinate systems and datums. Intertidal data in particular are often lacking due to the difficulty of data acquisition using conventional survey techniques. This paper presents a fast, accurate and low-cost methodology to fill this gap and highlights the effect of the digital elevation model characteristics, such as the interpolation method and spatial resolution, on modelled water levels and flooded areas. The Lima river estuary, located in North-western Portugal, is used as a case study. Validation tests for commonly available spatial interpolators showed ordinary kriging to be the most adequate interpolator. Digital elevation models with regular grids of 5 m and 50 m resolution were used, together with the original (not interpolated) elevation dataset, as input to a finite element hydrodynamic model for astronomic tide simulation. Results indicate that the larger differences between using different elevation models occur at low tide during spring tide, marginally impacting the flood modelling. The effect of a vertical offset of the chart datum with respect to a part of the digital elevation model was finally investigated, showing a limited influence in the determination of the water levels.

A method to assess the coherence of defensive systems based on the individual and combined visual coverage of fortifications is presented. The case study is the complex fortification system of Lines of Torres Vedras, around Lisbon, Portugal, activated in the first decade of the 19th century to halt the French armies of Napoleon in the Peninsular War. As messages between fortifications are known to have been exchanged visually by semaphore, GIS-based viewshed calculations enabled the assessment of the system's coherence and the exploration of how the undocumented communication with the defensive forces' headquarters in Lisbon might have been achieved. Results show that the defensive system with its two lines was highly effective per se with respect to the visual coverage of the terrain and the exchange of messages between fortifications, but it could have been improved with minor changes to the location of communication devices. It is also found that communication with the headquarters in Lisbon was not adequate, suggesting that additional devices were needed.

In this paper we address the problem of planning a network of weather monitoring stations observing average air temperature (AAT). Assuming the network planning scenario as a location problem, an optimization model and an operative methodology are proposed.  The model uses the geostatistical uncertainty of estimation and the indicator formalism to consider in the location process a variable demand surface, depending on the spatial arrangement of the stations. This surface is also used to express a spatial representativeness value for each element in the network. It is then possible to locate such a network using optimization techniques, such as the used methods of simulated annealing and construction heuristics.
This new approach was applied in the optimization of the Portuguese network of weather stations monitoring the AAT variable. In this case study, scenarios of reduction in the number of stations were generated and analyzed: the uncertainty of estimation was computed, interpreted and applied to model the varying demand surface that is used in the optimization process. Along with the determination of spatial representativeness value of individual stations, simulated annealing was used to detect redundancies on the existing network and establish the base for its expansion. Using a greedy algorithm, a new network for monitoring average temperature in the selected study area is proposed and its effectiveness is compared with the current distribution of stations. For this proposed network distribution maps of the uncertainty of estimation and the temperature distribution were created.

The computation of least-cost paths over a cost surface is a well-known and widely used capability of raster geographic information systems (GISs). It consists in finding the path with the lowest accumulated cost between two locations in a raster model of a cost surface, which results in a string-like, thin and long sequence of cells. In this article, a new extension of raster-based least-cost path modelling is proposed. The new modelling approach allows the computation of paths or corridors with a fixed width, larger than one cell. These swaths are called wide paths and may be useful in circumstances where the detail level of the raster cost surfaces is higher than the width of the desired path or corridor. The wide path model presented in the article is independent of the choice of least-cost algorithms, because the transformation from regular to wide paths is applied to the construction of nodes and edges of an induced graph. The article gives the foundations and discusses the particularities of such paths, regardless of the imposed width, and explores the difference from the usual least-cost path model. Test cases were included, one hypothetical and the other with real data. The results are coherent and indicative of the applicability of wide paths.