Blind spots in our understanding of the tropical nitrogen cycle
Understanding environmental controls on tropical nitrogen (N) cycling is critical for projecting flux rates and modeling ecosystem responses to rising CO 2 . Here, we ask whether geographic bias in field sampling may shape our understanding of N cycling by failing to capture the breadth of environmental heterogeneity within the tropics. We mapped empirical sampling locations of two major ecosystem N inputs and outputs—biological nitrogen fixation and denitrification—paired with a wide range of variables thought to influence N cycling. Both single and multi-variate analyses revealed that sampling density is greater in productive, wet, fertile, warmer tropical ecosystems with low seasonality—areas expected to have higher process rates—but biased away from spatially extensive biomes which may have high process rate variability and thus higher uncertainty. Even within tropical broadleaf forests, sampling is biased toward sites with unusually high soil fertility (N and phosphorus), potentially skewing biome-wide averages. Although some geographically undersampled regions share environmental characteristics with well-studied sites, others are unique, compromising inference from existing data. Finally, we find that while sampled field locations capture the range of N cycle process means/variances predicted by an ensemble of terrestrial biosphere models, this may be the result of circularity in model parameters and driving data rather than true coverage, including overrepresentation of forest plant functional types that may be parameterized from a small subset of sites. Data coverage expansion into areas with novel environmental conditions and higher uncertainty could be used to assess and improve model parameterization, as well as test general hypotheses.