Prospects and problems of simple linear models for estimating symbiotic N-2 fixation by crop and pasture legumes

Symbiotic dinitrogen (N-2) fixation of crop and pasture legumes is a critical component of agricultural systems, but its measurement is expensive and labour intensive. Simple models which can provide approximations based on crop or pasture dry matter production would be useful for agrononomists and those interested in regional nitrogen (N) cycle fluxes. We investigate meta analysis of published data on legume shoot dry matter production, shoot %N and legume %N fixed (%Ndfa) and look for relationships among these, as a possible way of providing useful approximations of N-2 fixation. We restricted our analysis to Australian studies where we have ready access to the primary data and where cultivars, management and climate are more constrained compared to a universal dataset. Regression analysis between shoot dry matter and amounts of shoot N-2 fixed were strong for all crop and pasture legumes with significant differences in slope and intercept values being obtained between pastures and crops, and between chickpea (Cicer arietinium) and all other crop and pasture legumes. Annual pasture legumes showed the strongest linear relationship between N-2 fixation and shoot dry matter and had the greatest slope (20.2-24.3 kg N-2 fixed/t), compared to 18.7 kg N-2 fixed/t for the perennial pasture legume lucerne (alfalfa, Medicago sativa), and between 10.7 to 23.0 kg N-2/t for crop legumes, depending upon species. It was recognised that the use of such shoot-based relationships would underestimate the total amounts of N-2 fixed since the contributions of fixed N present in, or derived from, roots and nodules are not included. Furthermore there needs to be careful consideration of the validity of an intercept term, which might reflect suppression of N-2 fixation at low dry matter and high soil mineral N availability, or possibly the use of non-linear regression. For chickpea crops grown in north-eastern Australia, multiple regression indicated that N-2 fixation was much more closely correlated with %Ndfa than dry matter production. Evidence presented also indicated that %Ndfa of other crops and lucerne in this region may similarly be influenced by soil mineral N. The regression approach presented provides a statistical basis to approximate N-2 fixation in the first instance. This work highlights some of the dangers of fitting single regressions to aggregated datasets and using these to approximate symbiotic N-2 fixation. The analysis indicates that where pasture legumes are grown in mixtures with non-legumes, and driven to high dependence on N-2 fixation, simple linear regressions may be quite useful, provided that possible differences between species are investigated as the slopes of the regressions between these can be quite different. For crop legumes, where low dependence on N-2 fixation can occur at higher mineral N availability, there is a need to carefully consider the intercept term, obtain estimates of mineral N availability, and/or resort to non-linear models. The gross generalisations presented in scatter plots cannot be reliably applied any more specifically, even within the datasets from which they were generated, and in some cases even within legume species between regions. They cannot substitute for direct measurement where any certainty is required under a particular set of defined conditions.