How to seasonalise daily weather classification and precipitation data

Hi,

I'm working on a climate data set for Svalbard from August 1957 to May 2015. I have daily rainall data (mm) and an associated weather classification number (1-11) which corresponds to a weather type for that day.

I'm trying to seasonalise this, for e.g. I want to see how much Rainfall fell on days associated with weather type 1 in spring but I'm having no luck.

Can anyone help?

I've managed to create whole time periods of weather type 1 precipitation for e.g my reference period 1971-00, but I need to seasonalise it first.

Data is attached, with KNG01date_d being the day, and KNG01rain_d being the rainfall (both start 1957.09).

In the attached data, column 1 represents weather type (1-11) and column 2 is the daily rainfall for that data.

For example. My script for finding total rain from weather type 1 between 1971 to 2000:

WC1_1971_00_KNG01_rain = find(KNG01rain_d(4871:15828,1)==1);
Sum_Rain7100_KNG01 = sum(KNG01rain_d(WC1_1971_00_KNG01_rain, :));
WC1_7100 = Sum_Rain7100_KNG01(2);

My question is, how can I seasonalise this to have Spring Summer Autumn Winter and then create weather types 1-11 with corresponding precipiation from that?

Sorry for the long winded and perhaps difficult to understand explanation

Thanks to anyone who can help!

[dpb]

So which months of the year are considered to be which seasons?

See findgroups and splitapply and friends.

I'd suggest using a timeseries object to hold the data or table

Answer 1

[Guillaume]

Up on the right of the page, there is a field for you to tell your version of matlab. Since it's not filled, I'm assuming a modern enough version that my answer below will work:

The simplest way to do what you want is to store your data in a timetable.

raindata = array2timetable(KNG01rain_d, 'StartTime', datetime(1957, 8,1, 'Format', 'dd-MMM-yyyy'), ...
                          'TimeStep', days(1), ...
                          'VariableNames', {'weathertype', 'rainfall'});

Then we can easily add information to the table, such as season:

raindata.season = quarter(raindata.Time);

And if we want to know the sum of rainfall for each season and weather time:

rainfall_per_season_and_weather = groupsummary(raindata, {'season', 'weathertype'}, 'sum')

There are plenty more functions to perform calculations on groups, such as rowfun, splitapply, etc., in modern matlab.

If you want to perform the above just for a certain period:

groupsummary(raindata(isbetween(raindata.Time, datetime(1971,1,1), datetime(1999, 12, 31)), :), ... 
             {'season', 'weathertype'}, 'sum')

edit: Now that you've attached your date matrix (I thought you left it up to us to work out the date range), you can also build the timetable with:

raindata = array2timetable(KNG01rain_d, ...
                           'RowTimes', datetime(KNG01date_d, 'ConvertFrom', 'datenum', 'Format', 'dd-MMM-yyyy'), ...
                           'VariableNames', {'weathertype', 'rainfall'});

According to this, your data starts in september, not august.

Also, I forgot to say, the find in your original code is completely unnecessary

WC1_1971_00_KNG01_rain = KNG01rain_d(4871:15828,1)==1;  %no need for find use the logical array directly
Sum_Rain7100_KNG01 = sum(KNG01rain_d(WC1_1971_00_KNG01_rain, :));

[Ed Harrison]

Amazing, Thank you so much :)!

Answer 2

[Andrei Bobrov]

Where KNG01date_d?

date1 = datetime(KNG01date_d,'ConvertFrom','datenum');
TT = array2timetable(KNG01rain_d,'RowTimes',date1,...
                                          'v',{' weather_type','rainfall'});
TT.season = discretize(month(TT.Time),[1,3:3:12,13],'categorical',...
                            {'Winter','Spring','Summer','Autumn','Winter'});
TT_out = varfun(@sum,TT,'GroupingVariables',{'season','weather_type'});

or

TT_out2 = groupsummary(TT,{'season','weather_type'},'sum');

[Ed Harrison]

Apologies, attached here!

[dpb]

Please attach to original Q?...

[Ed Harrison]

Done :)