A site-specific indicator of nitrogen loads into surface waters from conventional and conservation agriculture practices: Bayesian network model. (December 2022)

• Record Type: Journal Article
• Authors: Radomyski, Artur; Ashauer, Roman
• Is Part Of: Ecological indicators. Volume 145(2023)
• Publication Date: 2022-12
• Subjects: BN Bayesian Network or Bayes' Net -- CL Counting-Learning -- CN Curve Number -- CPT Conditional Probability Table -- EM Expectation-Maximisation -- GWF Grey Water Footprint -- HSG Hydrologic Soil Group -- nGWF Nitrogen Grey Water Footprint -- USLE Universal Soil Loss Equation; Bayesian Network -- Conservation Agriculture -- Nitrogen -- Water footprint -- Agricultural runoff; Environmental monitoring -- Periodicals; Environmental management -- Periodicals; Environmental impact analysis -- Periodicals; Environmental risk assessment -- Periodicals; Sustainable development -- Periodicals; 333.71405
• DOI: 10.1016/j.ecolind.2022.109641
• Languages: English
• ISSNs: 1470-160X

Abstract

Highlights: No-till and small grain crops show best results for limiting N load and grey water especially in high runoff scenario. Surface runoff predictions are sensitive to weather inputs and soil loss predictions to topographic factors. The BN model can be used with few site-specific inputs but results more uncertain. The BN model could be a useful generic indicator for assessing trends in grey water generated by N pollution. Abstract: Agriculture is one of the main sources of diffuse pollution, such as fertilisers, plant protection products, solid particles or pathogens. The short- and long-term impact of agriculture on surface water quality is often driven by on-farm decisions as to what practices are used to manage weeds. Farming practices can affect numerous processes such as surface runoff and soil erosion which mediate release and transport of potential pollutants to edge-of-field water courses. Excessive nitrogen emissions are an ongoing problem of many farmlands, leading to pollution of surface water bodies and causing adverse outcomes such as eutrophication, resulting in deteriorating water quality. We developed a Bayesian Network (BN) model to predict nitrogen load into surface waters at the farm level using site specific characteristics such as landscape, soil, cropping system and, also, to compare different conventional and conservation agriculture practices. The BN was built from well-established and accepted models i.e., surface runoff by water is based on the Curve Number method (CN), the soil erosion rate is calculated with the Universal Soil Loss Equation (USLE), and nitrogen load is calculated based on a multilevel model, where the CN and the USLE outputs are used as inputs. All three sub-models produced satisfactory spread around 1:1 line. A classification technique was applied to evaluate predicted nitrogen loads against the reference data from the US EPA STEPL model with respect to nitrogen load threshold values. Out of all the data simulated with the BN 83 % agree with the reference model for the 20 kg/(ha × yr) nitrogen load threshold, and 97 % agree when the threshold is set to the middle of the prediction range, 100 kg/(ha × yr). The modelling exercise was performed on two pedo-climatic scenarios differing in their potential to generate surface runoff water and by considering the effect of combining three farming practices and two generic groups of crops on emitted nitrogen loads expressed as the Grey Water Footprint (GWF). Sensitivity analysis shows high importance of weather inputs for surface runoff, and topographic information for soil erosion, whereas agricultural treatments were ranked as less important. Under variable precipitation no-till practice would result in reducing emissions of water, soil and nitrogen compared to conventional farming. The results indicate that no-till practice would reduce nitrogen loads on sites with varying risk of runoff, and that a combination of no-till and small grain crops provides the best benefit in reducing nitrogen emissions. We demonstrate that a Baye's net is a useful, flexible tool for data and knowledge assimilation and a practical approach to test and compare effects of various agricultural interventions on pollutant emissions from a farming system.