Oral Presentation Freshwater Sciences 2023

Maximizing the value of large-scale datasets to predict nutrient yields at the land to water interface (#422)

Andres F Suarez Castro 1 , Simon Linke 2 , Dale M Robertson 3 , David A Saad 3 , Bernhard Lehner 4 , David P Hamilton 1
  1. Australian Rivers Institute, Griffith University, Brisbane, QLD, Australia
  2. CSIRO, Brisbane, Queensland, Australia
  3. U. S. Geological Survey, Upper Midwest Water Science Center, Madison, Wisconsin, USA
  4. Department of Geography, McGill University, Montreal, Quebec, Canada

Predicting sources of nutrient runoff is fundamental to apply management actions that build resilience and protect water ecosystems. However, despite billions of dollars spent in water quality monitoring programs, decision makers across the globe are confronted with inadequate information to know where to allocate management actions that benefit the land to water interface. Here, we present an adaptable framework to maximize the value of data mapped at large spatial scales (> 1 km and global extents) used to predict nutrient yields at local scales (e.g., at the sub-basin, watershed and stream levels). This framework was applied in three regions with different levels of data availability: the Upper Midwest of the United States, New Zealand and North-western South America. For each region, we quantified and compared predictions of nitrogen delivery between models developed using novel large-scale datasets (e.g. HydroATLAS, LakeATLAS, HydroWASTE, MapSPAM) and models based on local data. In addition, we identified how model estimates varied with the spatial scale of analysis. Our results show that nutrient balance models based on large scale datasets had a high model estimation performance (R> 0.75) of total loads and yields. Both large scale and models based on local data provided similar coefficient estimates of diffuse (e.g. atmospheric deposition, farm fertilizers, manure), and point nutrient sources. However, spatial biases in the distribution of monitoring sites increased the uncertainty in total nutrient loads and yields predictions at local scales. We discuss how this framework helps to understand when coarse scale data improves nutrient estimates for local, catchment-scale application, particularly in data poor regions. We also present avenues of research to determine at what point the use of large-scale datasets cease to improve decision making for managing catchments at local scales.