arXiv subjects

Erik Chavez

Publications and source records attributed to Erik Chavez.

2 recordsLinked to original sources

arXiv2026

A stable hothouse triggered by a tipping mechanism

The climate system's nonlinear dynamics is influenced by various external forcings and internal feedbacks that can give rise to regional and even global tipping points that may lead to significant and potentially irreversible changes. Paleoclimatic records reveal that Earth's climate has shifted between distinct equlibria, including a "hothouse Earth" state with temperatures about 10 K higher than present. However, a specific mechanism for a sudden tipping to an alternate stable state, several degrees warmer than the present climate, has yet to be presented. We introduce a temperature-carbon-vegetation (TCV) model comprising an energy balance model of global temperature, coupled with global terrestrial and ocean CO2 dynamics, and with vegetation ecosystem change. Our model exhibits a new tipping mechanism that leads to a hothouse Earth under a high-emissions scenario. Its simulations align with both observations and IPCC-class global climate models prior to tipping. The two processes that produce global tipping are: (i) temperature-albedo feedback due to darkening of the terrestrial cryosphere by glacial microalgae; and (ii) limits to vegetation adaptation that lead to reduced carbon absorption.

physics.ao-ph↗

arXiv2023

ViTs for SITS: Vision Transformers for Satellite Image Time Series

In this paper we introduce the Temporo-Spatial Vision Transformer (TSViT), a fully-attentional model for general Satellite Image Time Series (SITS) processing based on the Vision Transformer (ViT). TSViT splits a SITS record into non-overlapping patches in space and time which are tokenized and subsequently processed by a factorized temporo-spatial encoder. We argue, that in contrast to natural images, a temporal-then-spatial factorization is more intuitive for SITS processing and present experimental evidence for this claim. Additionally, we enhance the model's discriminative power by introducing two novel mechanisms for acquisition-time-specific temporal positional encodings and multiple learnable class tokens. The effect of all novel design choices is evaluated through an extensive ablation study. Our proposed architecture achieves state-of-the-art performance, surpassing previous approaches by a significant margin in three publicly available SITS semantic segmentation and classification datasets. All model, training and evaluation codes are made publicly available to facilitate further research.

cs.CV↗