Recovering soil productivity attributes from experimental data: a statistical method and an application to soil productivity dynamics

We present a means to recover information about soil quality trends from data sets, such as those from long-term crop experiments, in which time series of direct measures of soil properties may be unavailable. The first objective of the paper is to develop a method to recover information about the evolution of soil quality attributes from a limited range of data. The second objective is empirical: focusing on the productivity component of soil quality, to apply a dynamic statistical estimation method to infer a measure of soil productivity and its evolution from a time series of data on yields, nutrient inputs and management techniques. The results of the empirical analysis confirm well-known input–productivity relationships, but also reveal new information about their dynamics. For example, more intensive cropping reduces soil productivity, but the dynamic effects of crop choice on productivity in a given period decline over time. N fertilizer can substitute for soil productivity in the short term, but in the long term, soil productivity decline due to intensive cultivation cannot be alleviated by higher N application rates. Simulations of soil productivity evolution across differing land management regimes reveal that continuous cropping of corn rapidly reduces soil productivity even at high N application rates, while rotational choices, especially the use of legumes, can lead to quite rapid soil productivity regeneration. Both our conceptualization of the soil productivity measure and our approach to its measurement have the advantage of making use of only limited longitudinal data. Our empirical findings convey some important implications for future research related to the sustainability of agricultural production worldwide.