Snow cover climatology over the Baltic States based on satellite data

Snow cover climatology over the Baltic
States based on satellite data
Justinas Kilpys
Climatology Division at Lithuanian Hydrometeorological Service, Vilnius, Lithuania
Introduction
Methodology
Snow is one of the essential climate variables.
For a long time in the Baltic States in-situ
measurements were the only data source for
snow monitoring. This study was an effort to
apply satellite data for snow cover monitoring
in this region. Time series of snow cover days
and snow water equivalent were generated
using MODIS and GlobSnow project data.
The MODIS daily snow cover product in climate modelling grid of 0.05 degree (MOD10C1)
was used to calculate annual and monthly snow cover days in the Baltic States for the period
2000–2013. The main limitation of MODIS data is that surface of the Earth is often obscured
by clouds. Simple backward and forward temporal gap filling technique (Foppa, Seiz, 2012)
was applied to fill the data gaps due to the cloud cover. The time series of snow water
equivalent (SWE) was generated using SWE product (25×25 km) from ESA DUE GlobSnow
project (Luojus et al., 2010). Satellite based number of snow cover days and SWE were
validated with the observations from ground stations.
Kuusiku
Dobele
Vilnius
Fig. 1 Elevation model of the Baltic States with
marked ground stations used for validation
Fig. 2 Mean annual number of snow cover days
based on MOD10C1 data, 2000–2013
Fig. 3 Mean maximum snow water equivalent
based on GlobSnow project data, 2000–2013
Validation
Validation with in-situ measurements from Vilnius (Lithuania), Dobele (Latvia) and Kuusiku (Estonia) meteorological stations showed that there is
a good agreement between ground observations and satellite data. The hit rate for MODIS based snow cover days (SCD) was 0.88–0.92
indicating high accuracy, although critical success index (CSI) was lower (0.68–0.72) indicating that in case of rare events snow covered pixels
were either overestimated or underestimated. Validation of GlobSnow SWE product with in-situ measurements showed that the overall accuracy
was lower than for SCD. The corr. coef. of SWE estimates varied from 0.64 at Dobele station to 0.91 at Kuusiku. At Dobele station GlobSnow
product showed underestimation error, while at other stations there was a tendency to overestimate SWE. The lower validation scores for satellite
based SCD and SWE were determined during the transition of seasons (November, March–April) when there are many thaws and snow cover is
ephemeral and patchy.
Fig. 5 Time series
of maximum snow
water equivalent at
Vilnius based on
GlobSnow and insitu data.
Fig. 6 Scatter plots
of SWE estimates
at Vilnius based on
GlobSnow and insitu data.
Fig. 4 Comparison of SCD derived from
MOD10C1 and in-situ measurements
Contact details
Justinas Kilpys
Climatology Division
Lithuanian Hydrometeorological Service
under the Ministry of Environment
Rudnios g. 6, LT-09300 Vilnius
Tel.: +370 648 06320
E-mail: [email protected]
Table 1 Validation scores for snow cover days (SCD)
derived from MOD10C1
Mean
In-situ
Corr. Hit
Station
abs.
FAR POD CSI SSclim
SCD
coef. rate
diff.
Kuusiku 102.3
22.1
0.75 0.88 0.25 0.88 0.68 -0.01
Dobele
76.1
8.3
0.94 0.92 0.22 0.86 0.70 0.82
Vilnius
93.1
8.9
0.89 0.92 0.15 0.83 0.72 0.79
Table 2 Summary of GlobSnow SWE product
performance over the Baltic States
Station
RMSE
(mm)
Bias (mm) Corr. Coef.
Sample
size
Kuusiku
17.9
3.7
0.91
79
Dobele
28.0
-4.1
0.64
35
Vilnius
11.6
4.9
0.80
85
Conclusions
This study showed that MOD10C1 snow product with 0.05 degree grid is adequate to
describe the temporal and spatial variation of snow cover days over the Baltic States. Data
gaps due to the cloud cover can be reduced using temporal gap filling, but this method is
applicable only for data re-processing. GlobSnow SWE product has a coarse resolution
(25×25 km) and thus it is not possible to accurately determine local features. However SWE
product is relevant for general overview of snow conditions in the region. Applying satellite
data for snow cover evaluation in the Baltic States is of interest on the national border regions
and trans-boundary river basins were measurement networks are scarce.