@@ -25,18 +25,26 @@ The model wind data is first interpolated to the regular grid and then their val
Inverse distance weighted (IDW) interpolation assumes that the points close to the observation point have more influence on that and the influence decreases further away from the observation point, depending on a defined weighted function such as the inverse of the distance between the observation point and the model point. The interpolation is done for both wind components in the west-east and south-north directions, namely ${U}$ and ${V}$ components.
Figure \ref{uw}, showing ${U}$ wind component, is an example of model data used in this study. The latitude of model data is in a range $53.5$N-${67.725}$N and their respective longitude lies between ${7.0}$E-${27.958}$E. Note, however, that AROME grid covers a bit larger domain than that in Figure \ref{uw}.
Figure \ref{uw}, showing ${U}$ wind component, is an example of model data used in this study. The latitude of model data is in a range $53.5$N-${67.725}$N and their respective longitude lies between ${7.0}$E-${27.958}$E. Note, however, that AROME grid covers a bit larger domain than that in Figure \ref{uw}.
\section{Observational data}
Observational data comes from mora database. The information on mora database can be found \href{http://mora-apps/}{here}. We only use WMO stations which are located on the sea or close to the coast. Figure \ref{station} shows the location of the $64$ mora stations on the sea. Note that not all these stations measure the wind exactly at the height $10$ m above the ground. That is, we only use those stations having a height comparable to $10$ m. The wind data is retrieved from mora website in a json format which are then processed and converted to CSV format, making them convenient for comparing to the model data.
\section{Results}
In the comparison between the model and observation, we inspect the time series of the $U$ and ${V}$ wind components, as well as the scatter plots. We also calculate the correlation coefficient between the model and observation. We only show the results of few stations as there are many stations.
In the comparison between the model and observation, we inspect the time series of the $U$ and ${V}$ wind components, as well as the scatter plots. We also calculate the correlation coefficient between the model and observation. We only show the results of few stations as there are many stations. Note that, we only use the stations which are located at a comparable height to ${10}$ m.
It should be mentioned that a more accurate analysis requires a vertical interpolation of the observational wind data to the height of ${10}$ taking into account the air stability condition and the air temperature profile over the sea. Here, we assume that the air is in a neutral stability condition which is a reasonable assumption for the air over the sea. This is confirmed in several studies for the wind over the sea.
Figures \ref{uma}-\ref{umf} depict the time series of the ${U}$ wind at ${6}$ stations. From these, it is seen that the model data agrees well with the observations. The same is true for the ${V}$ wind as evident in the figures \ref{vma}-\ref{vmf}.
Another interesting analysis metric is the scatter plot in that we can see the cloud of data. Figures \ref{sma}-\ref{smf} illustrate the scatter plot of ${U}$ wind of the observation versus the model data. As seen, the cloud shape is kind of thin so that it resembles the line in the statistical sense. It is in agreement with the time series plots as discussed above. The same is true for the ${V}$ wind component as seen in figures \ref{smaa}-\ref{smff}.
Lastly, we look into the Correlation Coefficient for ${6}$ stations as tabulated in Table \ref{tabstation}. As seen, the Correlation Coefficient is very high, indicating the good performance of AROME model in predicting wind over the ocean when compared to the observation.
\section{Conclusion and future perspective}
In this study, we have compared the wind data from AROME atmospheric regional model to the observational data for the wind obtained from mora database. This is done only for the stations located on the Baltic sea and the north sea which are very crucial for the NEMO ocean circulation model which is run operationally in SMHI.
In this study, we have compared the wind data at ${10}$ m above the sea surface derived from AROME atmospheric regional model to the observational data for the wind obtained from mora database. This is done only for the stations located on the Baltic sea and the North sea which are very crucial for the NEMO ocean circulation model which is set up for the Baltic sea and North sea and run operationally in SMHI.
The model data is interpolated to the observational point using the inverse distance weight interpolation method. The validation period was between January 1, 2018 and April 1, 2018. The comparison between the model and observational data in terms of the time series, scatter plots and correlation coefficients reveals that AROME model does a good job in predicting the wind on the sea.
The model wind data is interpolated to the observational point using the inverse distance weight interpolation method. The observational data are not exactly located at ${10}$ m and hence we chose the data which are located at a comparable height to ${10}$ m. The validation period was between January 1, 2018 and April 1, 2018. The comparison between the model and observational data in terms of the time series, scatter plots and correlation coefficients reveals that AROME model does a good job in predicting the wind on the sea.
A follow up to this study is to do a comparison for other years than 2018 and also other periods of a year. The best possible scenario is to look at the data for at least few consecutive years. Also, it is interesting to find out how the interpolation method can affect the results. For instance, we can use the bilinear or the nearest point interpolation method. These warrant a motivation for the follow up study.
\clearpage
...
...
@@ -60,7 +68,7 @@ A follow up to this study is to do a comparison for other years than 2018 and al
