<p><strong>Abstract.</strong> Atmospheric oxidation of isoprene emission from land plants affects radiative forcing of global climate change. There is an urgent need to understand the factors that control isoprene emission variability on large spatiotemporal scales but such direct observations of isoprene emission do not exist. Two readily available global-scale long-term observation-based data sets hold information about surface isoprene activity: gross primary productivity (GPP) and

more »

... ic formaldehyde column variability (HCHOv). We analyze multi-year seasonal linear correlations between observed GPP and HCHOv. The observed GPP–HCHOv correlation patterns are used to evaluate a global Earth system model that embeds three alternative leaf-level isoprene emission algorithms. GPP and HCHOv are decoupled in the summertime in the southeast US (<i>r</i>=&amp;minus;0.03). In the Amazon, GPP and HCHOv are weakly correlated in March-April-May (MAM), correlated in June-July-August (JJA) and weakly anticorrelated in September-October-November (SON). Isoprene emission algorithms that include soil moisture dependence demonstrate greater skill in reproducing the observed interannual seasonal GPP–HCHOv correlations in the southeast US and the Amazon. In isoprene emission models that include soil moisture dependence, isoprene emission is correlated with photosynthesis and anticorrelated with HCHOv. In an isoprene emission model without soil moisture dependence, isoprene emission is anticorrelated with photosynthesis and correlated with HCHOv. Long-term monitoring of isoprene emission, soil moisture and meteorology is required in water-limited ecosystems to improve understanding of the factors controlling isoprene emission and its representation in global Earth system models.</p>