Large variability of nitrate load estimated from sparse measurements by typical methods in Atlantic Canada

Kang Liang ^1,2* Yefang Jiang ³ Keith Fuller ⁴ Marcos Cordeiro ⁵ Xuesong Zhang ² Junyu Qi ⁶ Xiaoyuan Geng ⁷ Tao Liu ⁸ Qian Zhang ⁹ Mohammad Amir Azimi ¹⁰ Fanrui Meng ¹⁰

• ¹ Oak Ridge Institute for Science and Education (ORISE), Oak Ridge, Tennessee, United States
• ² Hydrology and Remote Sensing Laboratory, Beltsville Agricultural Research Center, Agricultural Research Service (USDA), Beltsville, Maryland, United States
• ³ Charlottetown Research and Development Centre, Agriculture and Agri-Food Canada (AAFC), Charlottetown, Prince Edward Island, Canada
• ⁴ Kentville Research and Development Centre, Agriculture and Agri-Food Canada (AAFC), Kentville, Nova Scotia, Canada
• ⁵ Department of Animal Science, Faculty of Agricultural and Food Sciences, University of Manitoba, Winnipeg, Manitoba, Canada
• ⁶ Earth System Science Interdisciplinary Center, College of Computer, Mathematical and Natural Sciences, University of Maryland, College Park, College Park, Maryland, United States
• ⁷ Canadian Soil Information Service (CanSIS), AER Ottawa, Agriculture and Agri-Food Canada (AAFC), Ottawa, Ontario, Canada
• ⁸ School of Forest Resource and Environmental Science, Michigan Technological University, Houghton, Michigan, United States
• ⁹ Center for Environmental Science, University of Maryland, College Park, Cambridge, Maryland, United States
• ¹⁰ Faculty of Forestry & Environmental Management, University of New Brunswick, Fredericton, New Brunswick, Canada

Nitrogen pollution in aquatic ecosystems, primarily from agricultural sources, presents significant environmental challenges. At the land management decision level, reducing nitrate leaching requires knowledge of nitrate loading over time and location, the complexity of which is amplified by limited data availability, especially in poorly gauged watersheds. This issue is particularly pronounced in cold and humid regions where water quality data are often collected during the growing season only. Large data gaps result in systematic errors when estimating nitrogen load based on traditional regression methods. In this study, we explore the feasibility of using processbased hydrologic model to estimate nitrate loads from sparse temporal water quality data in a coastal agricultural watershed in Atlantic Canada and compared its performance with three regression methods. We found that the absence of the available 16% non-growing season data during the 10-year study period can lead to significant biases (as high as 21%) in load estimation by regression methods. In contrast, nitrate load estimates obtained with the Soil and Water Assessment Tool (SWAT) were less sensitive to systematic data gaps. The results suggest that process-based models like SWAT can be a viable alternative for nitrate load estimation when limited data is available. As agri-environmental water quality issues become more pressing, it is crucial to use appropriate methods based on data quality and availability to avoid misleading results.