Integration method for various water quality monitoring data to estimate riverine constituent loads
Abstract
Water quality monitoring has been carried out in various strategies; e.g., systematic at a fixed interval, intensive in several storm events, discrete flow-proportional monitoring, or high-frequency automatic monitoring. Though the strategy itself is chosen for its purpose, we propose the integration method for those monitoring data in a generalized way to estimate the unbiased riverine constituent loads. Our generalized unbiased load estimation method called the IR (Importance resampling) method is based on the importance sampling (IS), an efficient Monte Carlo numerical integration method, and the rating curve method. IS method requires sampling with the probability proportional to size (PPS). The major idea of our generalized method was posterior resampling from existing sampling data using a goodness-of-fit statistic to extract PPS sample population. In addition to the IR resampling, our generalized method enables both point and interval estimation of riverine constituent loads. This method requires the high-frequency and continuous discharge data observed at an interval less than 1hr for the target period of load estimation and water quality monitoring data by an arbitrary sampling strategy. The resulting PPS sample size after resampling is determined by discharge data and the slope coefficient of a power-type rating curve. The effectiveness of this method was validated using high-frequency water quality (sodium, potassium, chloride, and suspended solids) and discharge data from a small forested catchment (12.14ha). The application of the IR method to the systematic monitoring data at a fixed interval for potassium load estimation resulted in about 10% in data availability. In other words, PPS sample size was 10% of the sample size of the systematic samples. The application for suspended sediment load estimation reveals that it was almost impossible to derive the load estimates owing to little PPS samples. The difference in the values of the slope coefficient of the rating curve was the main reason for this; they were 0.883 and 1.33 for potassium and suspended sediments. This method is applicable to an integration of various water quality monitoring data to estimate unbiased riverine constituent loads.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2017
- Bibcode:
- 2017AGUFM.H41F1506T
- Keywords:
-
- 0470 Nutrients and nutrient cycling;
- BIOGEOSCIENCES;
- 0496 Water quality;
- BIOGEOSCIENCES;
- 1871 Surface water quality;
- HYDROLOGY;
- 1880 Water management;
- HYDROLOGY