Premium
Improving the estimation of Lake Kinneret's heat balance and surface fluxes using the Kalman Filter algorithm
Author(s) -
Nussboim Shulamit,
Rimmer Alon,
Lechinsky Yuri,
Gutman PerOlof,
Broday David
Publication year - 2017
Publication title -
limnology and oceanography: methods
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.898
H-Index - 72
ISSN - 1541-5856
DOI - 10.1002/lom3.10173
Subject(s) - kalman filter , ensemble kalman filter , environmental science , latent heat , meteorology , sensible heat , data assimilation , energy balance , algorithm , filter (signal processing) , noise (video) , water balance , remote sensing , computer science , statistics , mathematics , extended kalman filter , geology , geography , artificial intelligence , thermodynamics , physics , geotechnical engineering , image (mathematics) , computer vision
High frequency (10 min) meteorological measurements are usually the basis for surface fluxes calculations (net radiation, sensible and latent heat) over a lake surface. Data from simultaneous high frequency monitoring of the lake‐water temperature profile can be used as additional information for calculating these fluxes more accurately, if the large random fluctuations of such data could be overcome. This challenge can be achieved using an algorithm that filters out the natural noise of both the surface fluxes calculation (“model”) and the monitoring data (“measurement”), such as the Kalman Filter (KF). The KF uses statistics of the uncertainty in the dynamics of the heat balance model and the measurements, and improves the calculated heat storage estimate. The KF algorithm was applied for studying the energy balance at the surface of Lake Kinneret, Israel. We tested its operation using different algorithms, in light of seasonal variations associated with meteorological and lake temperature conditions. Typically, during the spring and summer the uncertainty of the heat storage data result in low Kalman Gain, K , whereas during calm lake conditions, in the autumn, the gain was high. Using only data already measured in “the past” and the current measurement, the KF is more suitable for cases where information regarding surface fluxes is required online than other filters (such as a moving average), which need data from “the future.”