Seasonal changes in vegetation (phenology) affect surface roughness, albedo, atmospheric CO2 concentrations, biogeochemical cycling, and the hydrosphere. Recent work has shown that inter-annual variability in the global carbon cycle is strongly influenced by semi-arid ecosystems and that land surface models (LSMs) do a poor job of representing leaf phenology in semi-arid, drought deciduous regions. Due to feedbacks between vegetation and the fire cycle in these systems, poor predictions of leaf area can have a strong influence on carbon budgets and may explain some of the mismatch among land surface models’ representations of the terrestrial carbon cycle.
Most LSMs use a simple representation of seasonality whereby plants grow and drop leaves as a function of day length, temperature, and/or soil moisture, with no variation within plant functional types or across continents. Here we ask whether this assumption of static phenological strategy is robust by performing a meta-analysis of published research, combing the scientific literature for descriptions of leaf-on and leaf-off dates and drivers in semi-arid and savanna-type (SAST) ecosystems around the world. We combined data from 149 different studies and developed statistical models to predict leaf-on and leaf-off dates from the reviewed papers and ancillary data.
Results/Conclusions
Our review of the published literature resulted in data points (paper x location x plant functional type) from all of the continents except Antarctica, with 173 data points for green-up and 145 for brown-down, concentrated in Africa and South America. Interestingly, the distribution of methods was not even around the world – the majority of studies in Africa were based on remote sensing, while the majority of studies in South America were based on field observations. Many studies anecdotally attributed green-up and brown-down in these SAST ecosystems to soil moisture or rainfall, while the few studies that actually included experimental manipulations generally found mixed results (not all species responded in the same way to environmental changes). Our statistical analyses showed that latitude, then continent and plant functional type, were the strongest predictors of green-up and brown-down, not climatic effects. This result suggests that photoperiod and biogeographical differences should be taken in to account when LSMs attempt to predict phenology in SAST ecosystems.