Improving SWE Estimation by Fusion of Snow Models with Topographic and Remotely Sensed Data

Abstract

This paper presents a new concept to derive the snow water equivalent (SWE) based on the joint use of snow model (AMUNDSEN) simulation, ground data, and auxiliary products derived from remote sensing. The main objective is to characterize the spatial-temporal distribution of the model-derived SWE deviation with respect to the real SWE values derived from ground measurements. This deviation is due to the intrinsic uncertainty of any theoretical model, related to the approximations in the analytical formulation. The method, based on the k-NN algorithm, computes the deviation for some labeled samples, i.e., samples for which ground measurements are available, in order to characterize and model the deviations associated to unlabeled samples (no ground measurements available), by assuming that the deviations of samples vary depending on the location within the feature space. Obtained results indicate an improved performance with respect to AMUNDSEN model, by decreasing the RMSE and the MAE with ground data, on average, from 154 to 75 mm and from 99 to 45 mm, respectively. Furthermore, the slope of regression line between estimated SWE and ground reference samples reaches 0.9 from 0.6 of AMUNDSEN simulations, by reducing the data spread and the number of outliers.

Measurement of Snow Water Equivalent Using Drone-Mounted Ultra-Wide-Band Radar

“…The area coverage of this UAV-mounted radar system is small compared to satellite remote sensing products [24]. However, the high spatial resolution generated by the UAVmounted radar has the opportunity to detect local variability, resulting in more accurate SWE measurements.…”

“…This research includes novel image threshold methods and clustering [18], parabolic fitting [19], apex detection by fitting an analytical hyperbola function to the profile edges detected with a Canny filter [20], template matching algorithms [21], and a neural network approach [22]. SWE can also be estimated on a large scale using neural networks to perform a nonlinear mapping between datasets of manual measurements to predict SWE [23], and by combining snow model simulation, manual measurements, and auxiliary products derived from remote sensing in a k-nearest neighbors (k-NN) algorithm [24].…”

Testing remote sensing estimates of snow water equivalent in the framework of the European Drought Observatory

“…20 The strong interconnection among these factors makes remote estimates of SWE challenging, as exemplified by the numerous algorithms proposed in the literature, 21 ranging from simple empirical relationships to complex modeling of the physical processes. However, despite the abovementioned advantages that satellite estimations have over ground measurements, their coarse spatial resolution 22 limits the direct usage of passive microwave SWE products over complex and rough terrain in mountain regions, 23 where they may be characterized by high uncertainty. 24,25 Moreover, due to the underlying physics of the retrieval process, passive microwave estimates of SWE tend to saturate at high SD levels 26 and are often unreliable in the case of wet snow.…”

Including Parameter Uncertainty in an Intercomparison of Physically-Based Snow Models

“…Multi-physics snow models in which individual representations of snow-physical processes can be switched between different options have been found to be valuable tools for various applications. Such model systems enable uncertainty quantification (Günther et al, 2019) and help to generate ensembles for forecasting and assimilation systems (Lafaysse et al, 2017) or other data driven fusion approaches (De Gregorio et al, 2019). Multi-physics snow models paved the way for uncertainty quantification in physically-based snow models, as systematic and simultaneous investigations of various uncertainty sources including the model structure, parameter choices and forcing errors became possible.…”