Do temporal changes in soil density affect cosmic ray neutron sensing: a multi-site comparison

Abstract

Cosmic ray neutron sensor (CRNS) technology is increasingly promoted in agricultural environments for monitoring of volumetric soil water content (SWC). Among its advantages over traditional measurements is the provision of field-scale representative SWC timeseries without obstruction of management activities. To derive an integrated SWC average at a site, CRNS is typically calibrated using observations of gravimetric SWC, soil organic matter and bulk density (BD). Those variables may in the best case be derived from a large number of soil samples, collected ideally on multiple occasions and under a range of hydrometeorological conditions. Most CRNS applications use an average site-specific value of bulk density derived for a site from ≥1 field calibration and it is considered static over time. While this is a safe assumption for many environments, in agricultural settings, management activities (e.g. tillage) may introduce substantial changes in BD over time. This may affect the accuracy of the CRNS-based SWC estimates, which in turn could affect management decisions (e.g. on irrigation) or modelling efforts, relying on the CRNS estimates as inputs. Therefore, the objective of this work is to quantify the relevance and potential impact of temporal changes in BD both on the estimation of CRNS SWC (e.g. of interest for land managers) and on the neutron count rate. We focused on agricultural environments where temporal changes in BD are deemed relevant. For that we used data from four agricultural sites (three in Germany and one in Scotland, UK) spanning different soil types, level of agricultural management and hydrometeorological conditions. The selected sites are equipped with a stationary CRNS and BD was sampled on 3 or more occasions for sensor calibration. We quantify the differences in estimates of SWC by using the range of average BD values at a site and compare these differences to other sources of uncertainty (e.g. the integration time of neutron counts). We additionally consider existing theories on the interaction of neutrons and soil bulk density to evaluate the direct impact of BD changes on the neutron count rate with dedicated simulations. Finally, we make recommendations on when BD variability and thus its sampling over time may become important for the derivation of CRNS outputs.