Verification and validation

In assessing whether a Zonation conservation prioritization analysis is producing informative results, we borrow the concepts of verification and validation from the field of software engineering. Validation is the process of assessing whether our spatial prioritization model answers to our problem definition, i.e. how well the prioritization addresses a real-world need (Figure 12). Verification on the other hand is concerned with checking whether the implementation of the Zonation analyses corresponds to the model of spatial prioritization and how error-free the implementation is (Figure 12). In other words, validation answers to the question “Are we building the right system?” and verification to “Are we building the system right?” (Easterbrook, 2010).

You should do verification after every stage of the prioritization project (see also Figure 13). You should check, for example, whether pre-processing produced sensible and technically correct inputs for Zonation. Did Zonation produce sensible results, or were there perhaps some errors in the process? If the interpretation of the results does not make sense, is the problem in the input data or in the Zonation analysis configuration? Much of this type of verification is manual checks of various kinds, but it is possible to automate some of the verifications. For example, automating the data pre- and post-processing programmatically (see Chapter 4 and Section 6.1) reduces the risk of manual errors and makes verification easier. For the most common things to keep an eye on in terms of verification, see Box 2 and Box 3 (“common pitfalls”).

Figure 12. The concepts of validation and verification in relation to different stages of spatial conservation prioritization (see also Figure 13).

The fact that the prioritization analysis is technically correct does not mean that the results are useful. Whereas verification is typically a relatively objective and technical matter of checking that the prioritization analysis does what it is supposed, validation involves broader and more subjective assessment. The results of an informative spatial prioritization should of course correspond to the observed occurrence of biodiversity in nature. For a relatively simple analysis, which does not include any complicating factors such as connectivity or costs, the accuracy of the identified priorities can sometimes be evaluated against independent validation datasets and expert judgement. Introducing abstract components (such as connectivity) into the analysis makes validation harder. This is another reason why the sequential development of Zonation variants is a useful strategy; validating just the final production variant(s) can be difficult because of multiple interacting factors, whereas the validity of simpler development variants is typically easier. Perhaps the most common question you will hear from anyone looking at the results in more detail is “Why is this location a high-priority?” or its complement “Why isn’t this location a high-priority?” Using the landscape identification post-processing in Zonation (see Section 6.1.1) is useful here, because it gives insight into the underlying data and helps you to understand what is driving the priorities in the solution. What was inside an area that increased its priority?

Judging whether the model of spatial prioritization truly captures the occurrence of biodiversity features, our preferences and other relevant considerations is not an easy thing to do. Nevertheless, validation is crucial in evaluating the degree to which the results are useful and applicable. In the longer run, you will also need validation to improve your model of spatial prioritization and its implementation adaptively and iteratively. Even when the prioritization is not perfect e.g. due to incomplete data, one can still ask if it is informative in some sense and a reasonable complement to expert judgment.