Not every climate change-related question can be answered in a way that leads to a clear choice (Averyt et al. 2013). For the information producers, being clear about what is possible and what is not is an ethical responsibility (CSPWG 2015) and can help identify knowledge gaps and direct research in application-relevant ways (NRC 2012a). For information users, just knowing when approaches reach their limits is useful information as it allows managers and planners to develop appropriate practices for making decisions in a changing world (NRC 2009). Most water management and planning design decisions depend on the tails of the distribution (floods and droughts) which are inherently difficult and require planning for the unknowns by designing system redundancy and adding safety factors (Stakhiv 2011).

In many cases, it is hard to know before doing some analysis. There are, however, several situations where it is best to proceed with caution.

(1) Sparse observations: many locations have a limited historical baseline from which to build an understanding of how the system will be influenced by an altered climate. While global climate model output is by definition global, many applications require finer spatial resolution. If, however, there is nothing to ground truth models to, downscaling approaches and hydrologic modeling can be misleading and give a false sense of precision. In these cases, other approaches may be more meaningful.

(2) Spatial scales are too small: in some locations topography, coastal winds, fog, cool-air pooling or other local effects significantly affect local hydrology and are not sufficiently captured in existing datasets (Reclamation 2016; Curtis et al. 2014). If this is the case, it is important to determine whether this affects the impact(s) of interest. Additionally, when information is provided for specific locations it should be considered within its larger context to check whether the precision and spatial variability are appropriate.

(3) Temporal scales are too small: extreme events (on daily or sub-daily timesteps) are often difficult to observe and capture because of limitations of observation networks and biases in models and downscaling methods. In these cases, it might be better to do climate-informed perturbations of the system using stochastic hydrology (described in section 4.9).