A simple ecohydrological model captures essentials of seasonal leaf dynamics in semi-arid tropical grasslands

• Published in: Biogeosciences Vol. 7; no. 3; pp. 907 - 920
• Main Authors: Choler, P., Sea, W., Briggs, P., Raupach, M., Leuning, R.
• Format: Journal Article
• Language: English
• Published: Copernicus Publications 08.03.2010
• Subjects: Australia, Queensland
• ISSN: 1726-4189; 1726-4170; 1726-4189
• DOI: 10.5194/bg-7-907-2010

Modelling leaf phenology in water-controlled ecosystems remains a difficult task because of high spatial and temporal variability in the interaction of plant growth and soil moisture. Here, we move beyond widely used linear models to examine the performance of low-dimensional, nonlinear ecohydrological models that couple the dynamics of plant cover and soil moisture. The study area encompasses 400 000 km2 of semi-arid perennial tropical grasslands, dominated by C4 grasses, in the Northern Territory and Queensland (Australia). We prepared 8-year time series (2001–2008) of climatic variables and estimates of fractional vegetation cover derived from MODIS Normalized Difference Vegetation Index (NDVI) for 400 randomly chosen sites, of which 25% were used for model calibration and 75% for model validation. We found that the mean absolute error of linear and nonlinear models did not markedly differ. However, nonlinear models presented key advantages: (1) they exhibited far less systematic error than their linear counterparts; (2) their error magnitude was consistent throughout a precipitation gradient while the performance of linear models deteriorated at the driest sites, and (3) they better captured the sharp transitions in leaf cover that are observed under high seasonality of precipitation. Our results showed that low-dimensional models including feedbacks between soil water balance and plant growth adequately predict leaf dynamics in semi-arid perennial grasslands. Because these models attempt to capture fundamental ecohydrological processes, they should be the favoured approach for prognostic models of phenology.