Reconciling the disagreement between observed and simulated temperature responses to deforestation.

Land use changes have great potential to influence temperature extremes. However, contradictory summer daytime temperature responses to deforestation are reported between observations and climate models. Here we present a pertinent comparison between multiple satellite-based datasets and climate model deforestation experiments. Observationally-based methods rely on a space-for-time assumption, which compares neighboring locations with contrasting land covers as a proxy for land use changes over time without considering possible atmospheric feedbacks. Offline land simulations or subgrid-level analyses agree with observed warming effects only when the space-for-time assumption is replicated. However, deforestation-related cloud and radiation effects manifest in coupled climate simulations and observations at larger scales, which show that a reduction of hot extremes with deforestation - as simulated in a number of CMIP5 models - is possible. Our study provides a design and analysis methodology for land use change studies and highlights the importance of including land-atmosphere coupling, which can alter deforestation-induced temperature changes.