Coastal Masterplan Development Optimization


A large-scale development with hundreds of individual hotel units is planned for construction on a sea coast in Mexico. An investor, at the request of an architect, hired Parametric Support to evaluate buildings distribution and, in a collaboration with the architect, to develop a smart masterplan that has the lowest impact on the environment and, at the same time,  brings the highest profits. After careful examination we realized that in the coastal zones of Mexico views and landscape are two factors that are influencing the masterplan and the building distribution the most.  

fig. 1 Initial design


Views are one of crucial aspects affecting the price of a property. Academia survey shows that a water view is the most desirable and can increase the value of a property – a plot or an apartment/office space up to 67%. In Cancun, for example, if you compare the price of a garden/interior facing room it costs about 23% less than an oceanfront room (having 180° of unobstructed view to the sea) and about 14% less than a sea view room (having a partial view to the sea). Other important aspect of views are privacy (other observers cannot see the observer), open view (obstruction – natural and man-made) and an angle of a view. To our best knowledge there are no academia or publicly available studies on the price of a view, therefore we decided to only to look for an unobstructed view¹.

To perform a calculation we defined an array of points on terraces and on the centers of windows (living room and bedroom). From each point (14 in total) an array of viewing vectors (18) was generated, each vector rotated by 10 degrees from the previous one. In total, we were evaluating 252 rays for each unit.The algorithm was looking to maximize the views from each point and, at the same time,  balance a view distribution.

fig. 2 View rays

The second objective was to minimize intervention in the landscape. Minimizing dirt movement is beneficial from the point of view of economy and sustainability, as it requires less earthworks, workers, excavators, tipper trucks to add or remove dirt from a construction site, etc. A good balance between excavated and added dirt will reduce the direct cost of the preliminary phase of construction.To perform such optimization a detailed map of topography and geology is required. The precision of data will reflect in the quality of optimization.


The preliminary distribution of units made by an architect was our starting point. We identified three constraints that would exclude a solution. The first was to keep the position of the first row (though the units there could rotate freely around their weight centerpoint). The front row units are the premium ones and enjoy panoramic view to the bay, they are much larger and have different topology. The second constraint was to ensure the privacy on the terraces. Basing on Christopher Alexander’s ² research we decided that a terrace can be called private if there is no direct view from any other terrace that is closer than 300 m.

The last constraint was to distribute the units with equal density, maintaining 30 meters of free space around each one.

Project Workflow

The information the architect received from the surveyor were the positions of points (X, Y, Z), the tool we developed (Topomesh) translated these points into three-dimensional mesh and to topolines. Three types of ground are present on the plot: sand, clay and limestone rocks. Each of them has different properties and excavation cost and difficulty depends on its weight and density. This characteristic was translated to factors, where sand is the least laborious, its earthwork factor is 1, for clay it is 1.35 and for limestone rocks – 1.5. The information on topography was enriched with geological data and was our reference for calculation.


fig. 3 Topography and geology

Later we parametrized the units and defined a fitness function. We were searching for the solutions that were minimizing dirt movement and maximizing the views from each point. The units started to move freely through the landscape searching for the best position.

For every geographical distribution of the 40 units the algorithm was simultaneously checking the views, the amount of earthwork (excavated and added), the privacy and the distance. More than 30 000 solutions were checked. The Pareto Fronts of 30 runs of the algorithm were superimposed and the 80 best solutions were selected.

fig. 3 superimposed Pareto Fronts

fig. 4 80 best solutions

Result and comparison
To  built a Catalogue we ran a process based on Machine Learning which classifies the solutions according to their geometric differences. To provide the most variable and best set of solutions we were searching for the most dissimilar geometrical properties, comparing the position and rotation angles.

The catalogue has four best solutions.

Initial Design V_01 V_02 V_03 V_04
Earthwork [m3] 17,458.00 11,347.70 13,617.24 13,791.82 14,490.14
improvement [%] 35% 22% 21% 17%
Average View [°] 23 26 27 32 38
improvement [%] 13% 17% 39% 65%

fig. 5 Catalogue of the best solutions

The table shows a clear improvement. The best distribution can reduce earthwork by 35% which is 6100 m3 or 640 truck tippers, the earthworks will take significantly less time and will require less machine and people involved in the process. On the other hand, we also managed to improve the quality of views, the best solution has on average 15° more view than in the initial design. Moreover, when we analyzed the data for every single unit we found that some of them got 40° more of unobstructed sea view.

An automated, optimized approach let us discover better solutions; informed design let us take decisions based on validated data and make our decisions more conscious. We can ensure the client that the proposed solution has the lowest possible impact on the environment and that its return on investment is the highest.


  1. The Value of a water view: variability over 25 years in a coastal housing market;Julia L. Hansen, Earl D. Benson, The Coastal Business Journal,Spring 2013, Volume 12, Number 1
  2. Pattern Language, Christopher Alexander, Oxford, 1977

About the Author: Adrian Krężlik