Increasing accuracy of lake nutrient predictions in thousands of lakes by leveraging water clarity data

Abstract

Aquatic scientists require robust, accurate information about nutrient concentrations and indicators of algal biomass in unsampled lakes in order to understand and predict the effects of global climate and land‐use change. Historically, lake and landscape characteristics have been used as predictor variables in regression models to generate nutrient predictions, but often with significant uncertainty. An alternative approach to improve predictions is to leverage the observed relationship between water clarity and nutrients, which is possible because water clarity is more commonly measured than lake nutrients. We used a joint‐nutrient model that conditioned predictions of total phosphorus, nitrogen, and chlorophyll a on observed water clarity. Our results demonstrated substantial reductions (8–27%; median = 23%) in prediction error when conditioning on water clarity. These models will provide new opportunities for predicting nutrient concentrations of unsampled lakes across broad spatial scales with reduced uncertainty.

Urban Land-Cover Changes in Major Cities in China from 1990 to 2015

“…First, there is uncertainty in the detailed characterization of urban land-cover types using medium-resolution data due to the complexity of the urban landscape. The adoption of higher-resolution remote sensing data can improve the classification accuracy and facilitate the extraction of fine urban overlays [ 39 , 40 , 41 , 42 ]. In addition, limited by the resolution of Landsat data, the green space types (grassland, shrub, and woodland) were not further subdivided in this study.…”

Ecological prediction at macroscales using big data: Does sampling design matter?

“…Lake water clarity was included because it is available for nearly all lakes in our study sample (Fig. 2) and model predictions are more accurate when they are conditional on water clarity (Wagner and Schliep 2018, Wagner et al 2020). We also included five watershed‐specific characteristics for the area of land draining directly into the lake as well as the area that drains into upstream‐connected streams and lakes <10 ha (i.e., the inter‐lake watershed; Soranno et al 2017): watershed wetland cover; watershed complexity (a metric of watershed shape that measures the deviation of the watershed boundary from a circular shape); watershed to lake area ratio; watershed stream density; watershed forest cover; and watershed road density.…”

Improving the Estimation of Nitrogen and Phosphorus Concentrations in Lakes and Reservoirs Using a Stacked Approach

“…Understanding, predicting and mitigating the influences of changing land use and climate on lake eutrophication globally is an extremely important challenge facing aquatic research, and relies on the design of effective nutrient reduction programs (Jayakody et al., 2014). Inherent in this challenge is the need to accurately estimate potential water quality impacts in the future (Ballard et al., 2019; Wagner et al., 2020).…”

“…Estimations of future water quality demand reliable and precise information of nutrient concentrations in a large number of lakes located within the various contexts of natural and anthropogenic disturbances. Nevertheless, the accurate estimations of nutrient concentrations in the lacustrine ecosystems of a given region are challenging due to their complicated characteristics and the human and monetary costs of acquiring field observation data from a large amount of water bodies (Wagner et al., 2020). For example, there are 2,693 natural lakes with a surface area greater than 1.0 km 2 in China, and there are an additional 89,700 reservoirs that cover 26,870 km 2 (J. Huang et al., 2019, 2020; Meals et al., 2010; X. Yang & Lu, 2014).…”