Challenges and opportunities in land surface modelling of savanna ecosystems

摘要: Abstract. The savanna complex is a highly diverse global biome that occurs within the seasonally dry tropical to sub-tropical equatorial latitudes and are structurally functionally distinct from grasslands forests. Savannas open-canopy environments encompass broad demographic continuum, often characterised by changing dominance between C3-tree C4-grass vegetation, where frequent environmental disturbances such as fire modulates balance ephemeral perennial life forms. Climate change projected result in significant changes floristic structure, with increases woody biomass expected through CO2 fertilisation mesic savannas increased tree mortality rainfall interannual variability xeric savannas. interaction vegetation climate has traditionally challenged terrestrial biosphere models (TBMs), which aim simulate atmosphere land surface predict responses of forcing. In this review, we examine whether TBMs able adequately represent fluxes what implications potential deficiencies may have for projection scenarios rely on these models. We start highlighting defining characteristic traits behaviours savannas, how differ across continents information (or not) represented structural framework many TBMs. highlight three dynamic processes believe directly affect water use productivity system: phenology, root-water access dynamics. Following this, discuss current-generation they suitable simulating fluxes. Finally, give an overview eddy-covariance observations combination other data sources can be used model benchmarking intercomparison frameworks diagnose performance environment formulate road maps future development. Our investigation reveals systematically misrepresent effects (if considered at all) should critical areas Furthermore, must not static (i.e. prescribed behaviour) but capable responding conditions order emulate behaviour Without developments, however, will limited predictive capability making projections needed understand respond change.