Where to afforest? Single and multiple criteria evaluation methods for spatio-temporal decision support, with application to afforestation

Land evaluation and land use planning are scientific disciplines and professional practices that aim at promoting sustainable territorial development across the globe. One of the cornerstone elements in these disciplines and practices is the notion of land unit. A land unit is an area of land for which the within-unit variability of its functional qualities is smaller than the between-unit variability. With this concept in place, the basic purpose of land evaluation is to establish a matching between the qualities of a land unit and the requirements of a land use type with a view to transferring this information to the subsequent land use planning. This matching can consist in determining the most suitable land units for a given land use type or finding the optimal land use type that should be applied to a given land unit, with the ultimate goal of achieving a high performing, flexible and sustainable land use and management regime for the benefit of society. Several methods originating from decision sciences have recently been proposed to match land unit’s qualities to land use type’s requirements. Essentially, decision analysis is about selecting the best option among a number of feasible alternatives. Such selection involves in most cases the consideration of several conflicting objectives. In these situations the application of multi-criteria decision making methods, which perform a trade-off among the considered criteria to find the best alternative, becomes a requirement. In this dissertation several multi-criteria decision making methods are studied and applied to the spatial planning of new Pinus and Eucalyptus plantations in the southern Andes region of Ecuador. In the context of these studies, the considered criteria are termed land performance attributes, which can be seen as indicators of productive or regulatory ecosystem services. These attributes are categorized as either on-site, when the attribute level is conditioned only by inherent characteristics of the considered land unit (e.g., the stock of carbon stored in the soil), or off-site, when the attribute level depends on the state and behaviour of neighbouring or even distant land units (e.g., the amount of sediment trapped in a land unit depends on the sediment produced, transported and deposited in the land units upstream). A ‘per land unit’ land evaluation focused on on-site attributes considers a land unit as an independent entity, and assesses its performance without considering any relationship with other land units. The performance of the study region is then merely the sum of the per land unit performances. On the other hand, land performance can be evaluated at a regional scale. In this approach land units are considered to be interacting components of a larger system, e.g., a river catchment. This interaction must be taken into account in the assessment of regional land performance. In the first part of this study, afforestation planning was tackled using the existing a method called CAMF, which aims at locating the cells within a rasterized river catchment that should be afforested in order to minimize a single, off-site criterion, i.e., the sediment yield of the catchment. We proposed a first variant of this method with the goal of enhancing its efficiency in terms of execution time by approaching the problem from an on-site land evaluation perspective. It was observed that this variant was able to select areas for afforestation in a small fraction of the time required by the original method, while producing almost identical results. In the second variant we replaced the single flow direction algorithm used by CAMF to simulate sediment flow over the study catchment by a multiple flow direction model, in an attempt to make the sediment flow component more realistic. It was shown that after calibration, this new variant produces similar results when compared to WaTEM/SEDEM, a well-tested mechanistic sediment delivery simulation model that was used as a reference. Next, five existing and one newly proposed multi-criteria decision making methods were explored and evaluated in the context of solving the problem of selecting land units for afforestation with a view to optimize several on-site performance attributes. Despite the strong concepts behind the new Iterative Ideal Point Thresholding method, it was found less applicable to this particular case due to its inability to discriminate between closely related alternatives in a reasonable amount of iterations. Afforestation planning is not necessarily restricted to selecting areas within a region to establish forests. From a more strategic perspective the goal can be to determine where forestry land use types should be established and how they should be managed to maximize their benefits considering the full region as one entity. With a view to provide decision support tools that are useful in such strategic contexts, a Mathematical Programming model was formulated, implemented and its applicability was demonstrated. In a first stage, the model was applied to determine the full set of land unit-land use type combinations (or land use configuration) that should be established in a river catchment in order to optimize regional land performance. Regional land performance was expressed as the integration over the study area of the level achieved by a number of conflicting on-site attributes. An additional requirement set for the solutions produced by this model was that the achievement levels of the considered attributes are as uniform as possible, avoiding solutions that perform well for certain attributes and very poorly for the others. To enhance the flexibility of the approach described above, in a second stage, the Mathematical Programming model was further elaborated to allow land use changes to occur at fixed time intervals. This enhancement introduces the temporal dimension in the concept of land use configuration and extends it to the notion of land use trajectory configuration. Herewith, a land use trajectory is defined as a particular sequencing of several pre-defined land use types, defining the land use change that may or may not occur at the end of every interval within a considered time span. The result of the Mathematical Model when applied to the studied problem was a land use trajectory configuration that provides information about how land use should evolve throughout a period of time in order to obtain, at the end of this period, an optimal and balanced, temporally cumulated regional land performance. The application of these two Mathematical Programming models to the Southern Andes study area proved that a combination of techniques borrowed from multi-criteria decision analysis and Operations Research can result in valid methods for determining static or dynamic land use configurations that favour land performance enhancement while keeping a balance among the achievement levels of the considered attributes. The methods proposed in this dissertation must be seen as new or enhanced tools for land use planning in general and afforestation planning in particular. The aim of these methods or of any decision support tool is not to replace but rather to support the sensible judgement by stakeholders where it regards complex spatio-temporal decision problems.status: publishe