95th ESA Annual Meeting (August 1 -- 6, 2010)

COS 20-2 - Analyzing the rain-use efficiency: New insights by meta-analysis and quantile regression

Tuesday, August 3, 2010: 8:20 AM
334, David L Lawrence Convention Center
Jan C. Ruppert and Anja Linst├Ądter, Botanical Institute, Range Ecology and Range Management Group, University of Cologne, Cologne, Germany

In drylands aboveground net primary production (ANPP) and rain-use efficiency (RUE; quotient of ANPP and the corresponding precipitation, Le Houérou 1984) are two of the most common ecological indicators for assessing ecosystem’s state (including degradation and desertification) and the supply of ecosystem services. For these purposes, they have some major advantages over other ecological indicators such as indicator species or plant functional types: (1) ANPP and RUE data is comparatively easy and cheap to collect. (2) The principal ability of ANPP and RUE to assess ecosystem state has frequently been confirmed. (3) ANPP and RUE allow cross-system/-scale comparisons. Despite their widespread and frequent application, both indicators face growing criticism: ANPP and RUE both aggregate complex information and have therefore been referred to as ‘lumped’ parameters. Furthermore after 25 years of research, there is still no consensus about the trend of ANPP and RUE along precipitation gradients, which makes it difficult to extrapolate these parameters in space and time. In our study we used meta-analysis to disentangle the influence of various ecological factors on ANPP and RUE. Their response as function of precipitation was analyzed with linear piecewise quantile regression (LPQR).


Our findings showed that ANPP, and therefore RUE, are significantly affected by precipitation and land-use, what supports their general ability to assess degradation processes, and their application in ecosystem-modelling. Furthermore modelled meta-analyses revealed that ANPP and RUE are affected by factors such as biome and soil type. With the tool of meta-analyses, we were able to separate these effects into significantly discriminable, quantitative effect sizes. Therefore the criticism of RUE being a lumped parameter and indicator, not being feasible to assess degradation, has lost its weight. Analyzing the trend of ANPP and RUE along precipitation gradients by LPQR, rather than by linear or exponential regression, revealed a more detailed picture of its relationship to precipitation. We found that both curves include several sequential linear intersects which form an unimodal trend. The arid and semi-arid sites considered in this study peak (both for ANPP and RUE) around precipitation values of 200 mm*y-1. Furthermore LPQR revealed why there have been different findings towards the shape of the trends: when analyzing the trends of ANPP and RUE it is crucial to account for the total length of the gradient and for the observed section of it.