Skip to main content
Have a personal or library account? Click to login
Assimilation of lake water surface temperature observations using an extended Kalman filter Cover

Assimilation of lake water surface temperature observations using an extended Kalman filter

Open Access
|Dec 2014

Figures & Tables

Fig. 1

Typical annual cycle of the mixed layer depth in a boreal (dimictic) 21 m deep lake as reproduced by FLake. The mixed regime takes place in autumn and early spring (for a short period), and the stratified regime takes place in summer. Diurnal oscillations, with the convective and wind-mixing regimes, are also represented.

{ label needed for table-wrap[@id='UT0001'] }

Table 1. The impact I (for the definition see Section 4.2) of the assimilation of SYKE observations of LWST for the open water period (EKF-S experiment, summer of 2011) for 27 lakes whose geographical coordinates (deg) and mean depth D are given

Name (longitude, latitude)D (m)I (%)Name (longitude, latitude)D (m)I (%)
Kuivajärvi (23.9, 60.8)2.294.8Rehja-Nuasjärvi (28.0, 64.2)8.595.5Tuusulanjärvi (25.1, 60.4)3.294.3Vaskivesi (23.8, 62.1)7.097.1Pääjärvi 1 (24.5, 62.9)3.896.6Haukivesi (28.4, 62.1)9.194.9Pesiöjärvi (28.7, 64.9)3.995.4Kallavesi (27.7, 62.8)9.796.3Kyyvesi (27.1, 62.0)4.496.5Pielinen (29.6, 63.3)10.194.6Jääsjärvi (26.1, 61.6)4.696.2Konnevesi (26.6, 62.6)10.695.4Nilakka (26.5, 63.1)4.996.6Saimaa (28.1, 61.3)10.894.5Pyhäjärvi (22.3, 61.0)5.596.4Ala-Rieveli (26.2, 61.3)11.292.4Längelmävesi (24.4, 61.5)6.894.4Päijänne (25.5, 61.6)14.193.7Ounasjärvi (23.6, 68.4)6.697.3Inarijärvi (27.9, 69.1)14.397.1Lappajärvi (23.7, 63.1)6.993.4Näsijärvi (23.8, 61.6)14.794.0Oulujärvi (27.0, 64.5)7.095.0Pääjärvi 2 (25.1, 61.1)14.896.7Unari (25.7, 67.1)7.094.0Kilpisjärvi (20.8, 69.0)19.596.8Kevojärvi (27.0, 69.8)7.098.0

Table 2. Design of the experiments

Experiment nameData assimilation ObservationsCycling periodNumber of lakes
FR (free run)NoNoNo27EKF_SYesSYKE24 h27EKF_MYesMerged24 h (48 h for cross-validation)4
Fig. 2

Time evolution of the mixed layer temperature in °C for Lake Inarijärvi (the mean depth is 14 m) for the summer period from May 2011 to November 2011. The FR, EKF-S and EKF-M results are shown by the blue, green and cyan lines, respectively. The SYKE and merged LWST observations are represented by the red dots and pink crosses, respectively.

Fig. 3

Time evolution of the ice and snow thickness in metres for Lake Inarijärvi (the mean depth is 14 m) for the winter–spring period from November 2010 to June 2011. The FR, EKF-S and EKF-M results for ice thickness are shown by the blue, green and cyan lines, respectively. The FR results for snow thickness are shown in pink. The EKF-S and EKF-M results for the snow thickness coincide with the FR results and are not shown.

Table 3. Ice break-up dates in the FR, EKF-S and EKF-M experiments and the first spring observed LWST for the EKF-S and EKF-M experiments (note that the ice break-up dates are contemporaneous with the spring start of observations for EKF-S and EKF-M)

InarijärviSaimaaLappajärviTuusunlajärvi
FRDate06/0605/1005/0905/07EKF-SDate05/2305/1105/1305/06LWST7.6°C12.6°C12.0°C8.9°CEKF-MDate05/1605/0505/0404/26LWST4.6°C7.8°C2.7°C2.8°C
Fig. 4

Time evolution of (a) the mean water temperature and (b) the bottom temperature in °C for Lake Inarijärvi (the mean depth is 14 m) for the summer period from May 2011 to November 2011. The FR, EKF-S and EKF-M results are shown by the blue, green and cyan lines, respectively.

Fig. 5

Time evolution of (a) the mean water temperature and (b) the bottom temperature in °C for Lake Saimaa (the mean depth is 11 m) for the summer period from April 2011 to November 2011. The FR, EKF-S and EKF-M results are shown by the blue, green and cyan lines, respectively.

Fig. 6

Time evolution of (a) the mixed layer depth in metres and (b) the shape factor for Lake Inarijärvi (the mean depth is 14 m) for the summer period from May 2011 to November 2011. The FR, EKF-S and EKF-M results are shown by the blue, green and cyan lines, respectively. Note that the shape factor changes rapidly and during mixing period it equals to a bogus value of 0.5.

Fig. 7

Time evolution of increments, background and analysis values of (a) the mixed layer depth in metres and (b) the mean water temperature in °C for Lake Saimaa (the mean depth is 11 m) for the period of June–July 2011. The EKF-S results are shown by the green line; the analysis increments, the background and analysis values are represented by the pink, green and blue crosses, respectively.

Fig. 8

Time evolution of the mixed layer temperature in °C for Lake Saimaa (the mean depth is 11 m) for the period of June–July 2011. The FR and EKF-S results are shown by the blue and green lines, respectively. The LWST observations are represented by the pink dots.

Fig. 9

Time evolution of (a) the mixed layer temperature and (b) the innovations (observed minus background values) and residuals (observed minus analysed values) in °C for Lake Saimaa (the mean depth is 11 m) for the period of June–July 2011. The EKF-S results are shown by the green line. The LWST observations, background and analysis values are represented by the pink, green and blue crosses, respectively. The innovations and residuals are represented by the pink and blue crosses, respectively.

{ label needed for table-wrap[@id='UT0004'] }

Table 4. The impact I (for the definition see Section 4.2), the bias, the root mean square error (RMSE) and the error standard deviation (ESTD) (°C) for the assimilation of the merged LWST observations for different lakes (the bias is calculated as simulated minus observed values, a positive bias means an overestimation of LWST)

FREKF-M

LakeI (%)BiasRMSEESTDBiasRMSEESTD
Inarijärvi97.9−2.035.024.590.120.960.96Saimaa96.8−1.073.673.52−0.041.111.11Lappajärvi97.50.192.872.870.461.511.44Tuusulanjärvi95.90.832.922.800.851.411.13

7. Appendix Limits of the components of the tangent linear model operator M, the linearised observation operator H, the background error covariance matrix B and the Kalman gain vector K for different mixing regimes

ComponentValueRemarks
Mixed regime
Matrix M
T¯tT¯00.8–0.9 K/KAlmost constant value, less for shallow lakesVector H
T¯T¯1.0 K/KConstant valueMatrix B
var(eT¯)4.5–4.8 K2Almost constant valueVector K (weights)
for T¯0.82–0.83Almost constant valueStratified regime
Matrix M
T¯tT¯00.5–1.0 K/KLarger values for deep lakes, smaller for shallow lakes; for shallow lakes the annual cycle has a minimum in midsummerT¯tη0(−1.2)–(−0.1) KLarger absolute values for shallow lakes, smaller for deep lakes; the annual cycle has a maximum in midsummerT¯tTb0.01–0.40 K/KLarger values for shallow lakes, smaller for deep lakes; the annual cycle has a maximum in midsummerT¯tCT(−4.5)–(−0.1) KLarger absolute values for shallow lakes, smaller for deep lakes; the annual cycle has a minimum in midsummerηtT¯0(−0.2)–0.2 K−1Almost zero values in midsummerηtη00.1–1.4Larger values for deep lakes, smaller for shallow lakes; the annual cycle has a maximum in autumnηtTb(−0.3)–0.5 K−1Larger absolute values for shallow lakes, smaller for deep lakes; the annual cycle has a minimum in autumnηtCT(−1.0)–1.0Almost zero values in midsummerTbtT¯0(−0.6)–0.6 K/KThe annual cycle with a minimum in springTbtη0(−2.0)–0.4 KClose to zero in midsummerTbtTb00.2–1.2 K/KFor shallow lakes an almost constant value of 1.0 K/K, for deep lakes small values in midsummerTbtCT0(−6.0)–3.0 KFor shallow lakes low values in midsummer, for deep lakes high values in springCTtT¯0(−0.03)–0.60 K−1Almost zero values in midsummerCTtη0(−0.01)–0.80Almost zero values in midsummerCTtTb(−1.4)–0.03 K−1Almost zero values in midsummerCTtCT(−0.1)–1.2For shallow lakes almost zero values in midsummer, for deep lakes an almost constant value of 1.0Vector H
TMLT¯1.0–2.6 K/KThe annual cycle has a maximum in midsummerTMLη0.0–20.0 KThe annual cycle has a maximum in midsummerTMLTb(−1.6)–(−0.1) K/KThe annual cycle has a minimum in midsummerTMLCT0.0–30.0 KThe annual cycle has a maximum in midsummer, the amplitude is stronger for deep lakesMatrix B
var(eT¯)4.0–16.0 K2The annual cycle has a maximum in midsummer, the amplitude is stronger for shallow lakescov(eT¯,eη)(−2.0)–0.2 KThe annual cycle has a minimum in late summer, the amplitude is stronger for shallow lakescov(eT¯,eTb)(−2.0)–16.0 K2The annual cycle has a maximum in late summer, the amplitude is stronger for shallow lakescov(eT¯,eCT)(−0.80)–(−0.01) KNo annual cycle, oscillationsvar(eη)0.01–0.90Almost constant small values, but outbreaks in autumncov(eη,eTb)(−5.0)–1.0 KThe annual cycle has a minimum in late summer, the amplitude is stronger for shallow lakescov(eη,eCT)(−0.05)–0.15Almost zero values in midsummervar(eTb)1.0–45.0 K2The annual cycle has a maximum in late summercov(eTb,eCT)(−2.5)–2.0 KNo annual cycle, oscillationsvar(eCT)0.01–0.18No annual cycle, oscillationsVector K (weighs)
for T¯0.20–0.83The annual cycle has a minimum in midsummerfor η(−0.06)–0.60 K−1No annual cycle, oscillationsfor T b (−0.6)–0.4No annual cycle, oscillationsfor C T (−0.05)–0.02 K−1The annual cycle has a maximum in midsummer
Language: English
Page range: 21510 - 21510
Submitted on: May 23, 2013
Accepted on: Sep 3, 2014
Published on: Dec 1, 2014
Published by: Stockholm University Press
In partnership with: Paradigm Publishing Services

© 2014 Ekaterina Kourzeneva, published by Stockholm University Press
This work is licensed under the Creative Commons Attribution 4.0 License.