Networks are widely used across many fields to represent interactions, such as the web, social interactions, infrastructure, or biological connections. Algorithms and processes on these networks have many uses, for example to find the shortest path in a road network, or to understand the spreading of information through a social network. In order to better answer such questions, we should understand how algorithms and processes are affected by features of the networks. One useful tool for this can be found in random network models, serving as proxies for real-world networks. Their features can be controlled via model parameters. While these models are usually easier to understand and analyze than real-world networks, their usefulness depends on how realistic they are. Thus, we look at well-known network models and evaluate whether they are useful for predicting algorithm behavior on real-world networks - the external validity. We focus on many widely-used algorithms and processes, and establish that the models are good proxies for real-world networks. We observe that heterogeneity and locality are important features for explaining and predicting algorithm behavior. In order to use random network models as proxies for real-world networks or to investigate model fit, one should utilize the model configuration capabilities optimally. As the models can be configured via parameters that affect the network features, this entails choosing the best parameters for some target network features. However, the exact relation between the configuration and the resulting network features is not clear. We present a method for estimating the best-fitting parameter configuration for commonly used network models, given some target features of the resulting networks. Our iterative method based on stochastic approximation needs only few samples and works reliably across a range of parameter configurations. Furthermore, we take a closer look at the features and substructures of networks. In particular, based on methods from clustering and outlier detection, we consider the idea that the edge set actually comes from two edge sets with differing features. With this idea, we investigate the well-known bootstrap percolation process on this combination of networks, which had previously only been considered on single network models. We theoretically and empirically consider the process on two combined network models, with differing percolation thresholds for the two edge sets, and see an interesting emerging behavior of a slow and a rapid percolation phase. In addition, we consider ways to separate edge sets based on locality features. Assuming the edges are a combination of two network models, we show how the edges can be separated by a simple metric, even with imbalanced set sizes. Overall, we evaluate common random network models, improve methods to utilize them, and contribute towards the design and understanding of new, combined network models.

Detailed microstructural analysis of three basaltic sills of the Little Minch Sill Complex demonstrates that convection leaves a detectable signature in fully solidified bodies. The presence of dense clusters of equant grains of olivine and clinopyroxene in the central parts of sills can only be accounted for if they formed and were enlarged while suspended in convecting magma, with delayed settling to the sill floor. An associated stratigraphic invariance of plagioclase grain shape is consistent with growth while suspended in convecting magma. These microstructural indicators demonstrate that convection during solidification was vigorous and long-lived in the 135-m-thick picrodolerite-crinanite unit (PCU) of the composite Shiant Isles Main sill and vigorous and likely short-lived in the PCU of the composite Creagan Iar sill. In contrast, convection in the Meall Tuath sill was weak and short-lived: plagioclase grain shape in this sill varies with stratigraphic height, indicative of primarily in situ nucleation and growth at the magma-mush interface, while olivine and clinopyroxene were kept suspended in the overlying convecting magma. The magma in all three sills fractionated during solidification, permitting convection driven by the instability of an upper thermal boundary layer. The comparative vigour and longevity of convection in the Shiant Isles Main sill and the Creagan Iar sill was due to their emplacement above an earlier, still-hot, intrusion, resulting in highly asymmetric cooling.

Visualising Southern African Late Iron Age Settlements in the Digital Age studies the visualisation of Southern African Late Iron Age Settlements (LIAS) (c. 900–1800) across the late nineteenth, twentieth, and early twenty-first centuries (1871–2020), as found in a survey of the cultural production, circulation, reproduction, and theorisation of illustrations accompanying archaeological, anthropological, and historical Southern African LIAS research. A valuable contribution of LIAS research is its continuous demonstration of a pre-colonial hub of cosmopolitanisms on a scale never imagined in colonial histories of 'indigenous' communities – thought of as the ultimate 'other' of global modernity. This study focuses on the visualisation of four settlements, namely: Mapungubwe, Khami, Great Zimbabwe, and Bokoni. It is proposed that as with the authority of Eurocentric 'formative interpretations' of LIAS research currently under review, visualisations accompanying LIAS also need to be critically relooked at within appropriate visual cultural methodologies informed by postcolonial, decolonial and critical race theory. The study follows a two-fold methodological framework involving a textual analysis and an image-making process. On both accounts, the study focuses on the cultural politics of representation, asking: who and what is being made visible in the visualisation of settlements accompanying LIAS research; what forms of materiality and spatiality are pictured and performed; what is the affect such visualisations have on the people that experience them; and finally, what do they mean in the context in which they are made?

Hellweger et al. (Reports, 27 May 2022, pp. 1001) predict that phosphorus limitation will increase concentrations of cyanobacterial toxins in lakes. However, several molecular, physiological, and ecological mechanisms assumed in their models are poorly supported or contradicted by other studies. We conclude that their take-home message that phosphorus load reduction will make Lake Erie more toxic is seriously flawed.