Abstract
Calculating the volumes of water discharged by streams is becoming increasingly important in water accounting and deciding how much water to allocate to competing uses. Water accounting is particularly important in Australia, as the driest inhabited continent and also in the face of potential impacts of a changing climate. Stream networks all over the world are littered with gauging stations, which take regular measurements of steam flow in order to help natural resource managers make decisions regarding water allocation. Estimating total discharge volumes is also of utmost importance when estimating pollutant loads from catchments.
Introduction
In hydrology, calculating the total volume of water passing through a stream (discharge) is typically achieved by integrating a hydrograph—a plot of flow over time. The standard approach utilizes the trapezoidal rule, which assumes a linear relationship between discrete data points. However, this deterministic method fails to account for three major sources of uncertainty: process uncertainty (unknown behavior between readings), rating curve uncertainty (the approximation of flow from water height), and inherent measurement error from sensors. In water-scarce regions like Australia, these errors can lead to significant miscalculations in water accounting and pollutant load estimations. This research proposes a robust Monte Carlo framework that provides a distribution of likely volumes rather than a single, potentially misleading point estimate.
Methodology
The proposed statistical pipeline moves beyond simple geometric integration by treating streamflow as a continuous-time stochastic process.
Nonparametric Stochastic Differential Equations (SDEs)
Instead of relying on fixed physical formulas, the model uses nonparametric SDEs to derive "drift" (average trend) and "volatility" (randomness) functions directly from historical data. A Weighted Nadaraya-Watson (WNW) estimator is employed to estimate the conditional distribution of flow given water height, incorporating time as a variable to account for shifts in the rating curve over the station's history.
Monte Carlo Simulation and Diffusion Bridges
The system generates an ensemble of 1,000 possible hydrographs. To fill gaps between known observations, the model utilizes "diffusion bridges"—simulated paths constrained to start and end at observed data points while following learned SDE dynamics in between. Each simulated hydrograph is then integrated to produce a probability distribution of the total discharge volume.
Case Study: Obi Obi Creek
The method was validated using daily flow data from Gardener’s Falls at Obi Obi Creek in Queensland, Australia, spanning from 2000 to 2010. The variables included log-flow, water height, seasonal timing, and daily rainfall grids.
Key Results
The Monte Carlo framework demonstrated strong consistency with traditional methods while providing essential uncertainty bounds. For the year 2009, the traditional trapezoidal estimate ($V_{trap} = 0.9343$) was found to fall well within the bounds of the generated distribution.
The results indicate that the variability in flow distributions is significant. By providing 5th and 95th percentiles, the model offers water managers a "risk budget," which is more robust for policy-making than a single estimate that may ignore inherent measurement and process errors.
Conclusion
By embracing stochasticity and nonparametric flexibility, this approach adapts to the unique dynamics of specific streams without requiring manual re-calibration. While computationally more intensive than traditional methods, the ability to quantify uncertainty and accurately infill missing data represents a significant advancement in hydrological modeling. Future work may explore the potential of these nonparametric SDEs as standalone rainfall-runoff models.
Summary updated