Are strong fire-vegetation feedbacks needed to explain the spatial distribution of tropical tree cover?

Aim The spatial pattern of tropical fire-induced tree mortality is partly determined by climate, but feedbacks of tree cover on fire are also important. We re-examine some recent observations proposed as evidence for very strong tree-cover feedbacks on fire, sufficient to allow savanna and forest to be alternative stable states over large areas of the tropics. Two pieces of previously reported observational evidence are examined: (1) the trimodal statistical distribution of tropical tree fraction, and (2) the fact that different tree fractional cover is found at different locations in the tropics with similar rainfall. Location Global tropics. Methods For point (1) above we analyse the statistical distribution of tree fraction predicted by the logistic equation of tree growth and self-competition, with spatially varying mortality rates. For (2), the relationship between mean annual rainfall and mean net primary productivity (NPP) in a climate model is examined. Results (1) A trimodal distribution of tree cover does not necessarily require tree-cover feedback on fire. It can arise from a combination of two factors: nonlinearities in vegetation dynamics and climate-driven spatial variation in mortality (the intermediate fire-productivity hypothesis). (2) Different locations in the tropics can have identical rainfall but significantly different NPP, even with no feedback of tree cover on fire. Main conclusions Our results show that strong tree-cover feedback on fire is not necessary to explain observations (1) and (2). However, it is still possible that strong fire-vegetation feedback is the primary explanation - our results do not rule this out. We simply demonstrate the possibility of an alternative hypothesis (of strong climate control). In reality, it is likely that both tree-cover feedback and climate contribute. It is challenging to separate these two effects cleanly. More work is needed to quantify their separate effects. We show that plots of mortality versus productivity are useful tools for understanding spatial variations in tree cover.