Preliminary analysis of spatially extended ecosystems modelled by activator-inhibitor type reaction-diffusion systems, such as dry savannas (Klausmeier 1999, Eigentler and Sherratt 2019, Tzuk et al 2020), show that that tipping may or may not be preceded by a pattern forming Turing bifurcation (Figs. 1, 2). The conditions for which this may or may not happen can be explicitly expressed in terms of model parameters (Rietkerk et al 2021). However, this only gives insight in the onset of spatial patterns, while these only become relevant beyond onset (‘far from equilibrium’). An explicit scenario by which this may happen, and thus by which tipping may be evaded, has been unravelled within the literature on dry savannas, and consists of a cascade of shifts of regular vegetation patterns to longer wavelengths (Siteur et al 2014, Bastiaansen et al 2018, Bastiaansen et al 2020, Rietkerk et al 2021; Fig. 2B). The crucial question now is whether this scenario may also play a role in other types of systems and/or whether there are alternative scenarios through which small amplitude Turing patterns may evolve into large scale interacting localized structures that enable the system to evade tipping. Due to the singular nature of the models we will be able to mathematically ‘control’ the basic localized structures, such as stripes or spots, and their interactions (Bastiaansen and Doelman 2019, Bastiaansen et al 2019, Jaïbi et al 2020). However, we should note that this is only true for model systems consisting of two state variables, considering only two divergent spatial scales of interaction; mathematical insight is lacking for complex systems or ecosystems governed by three or more state variables and scales of spatial interaction. To make the crucial connection between these localized patterns and those that appear from the Turing bifurcation, we will investigate the Busse balloon and its boundary, by a combination of bifurcation analyses, simulations and numerical continuation (Sherratt 2012, Van der Stelt et al 2013). Here, a fundamental challenge is to develop theory that expands from one-dimensional Busse balloons to two-dimensional patterns, including (spatial mixtures of) different morphologies, and to patterns emerging from destabilization of spatial boundaries. This extension to two-dimensional patterns is especially relevant for the observational approach for savanna and tundra that we will use to validate and challenge the new theory