A unified explanation for the morphology of raised peatlands.

Raised peatlands, or bogs, are gently mounded landforms that are composed entirely of organic matter1-4 and store the most carbon per area of any terrestrial ecosystem5. The shapes of bogs are critically important because their domed morphology4,6,7 accounts for much of the carbon that bogs store and determines how they will respond to interventions8,9 to stop greenhouse gas emissions and fires after anthropogenic drainage10-13. However, a general theory to infer the morphology of bogs is still lacking4,6,7. Here we show that an equation based on the processes universal to bogs explains their morphology across biomes, from Alaska, through the tropics, to New Zealand. In contrast to earlier models of bog morphology that attempted to describe only long-term equilibrium shapes4,6,7 and were, therefore, inapplicable to most bogs14-16, our approach makes no such assumption and makes it possible to infer full shapes of bogs from a sample of elevations, such as a single elevation transect. Our findings provide a foundation for quantitative inference about the morphology, hydrology and carbon storage of bogs through Earth's history, as well as a basis for planning natural climate solutions by rewetting damaged bogs around the world.

Promoting Connectivity of Network-Like Structures by Enforcing Region Separation.

We propose a novel, connectivity-oriented loss function for training deep convolutional networks to reconstruct network-like structures, like roads and irrigation canals, from aerial images. The main idea behind our loss is to express the connectivity of roads, or canals, in terms of disconnections that they create between background regions of the image. In simple terms, a gap in the predicted road causes two background regions, that lie on the opposite sides of a ground truth road, to touch in prediction. Our loss function is designed to prevent such unwanted connections between background regions, and therefore close the gaps in predicted roads. It also prevents predicting false positive roads and canals by penalizing unwarranted disconnections of background regions. In order to capture even short, dead-ending road segments, we evaluate the loss in small image crops. We show, in experiments on two standard road benchmarks and a new data set of irrigation canals, that convnets trained with our loss function recover road connectivity so well that it suffices to skeletonize their output to produce state of the art maps. A distinct advantage of our approach is that the loss can be plugged in to any existing training setup without further modifications.