create a matrix with my conditions

Question

I need some help creating one code to extract some rows with certain conditions. I have a 3D matrix (latitude,longitude,time(15000 values,1 value a day)) called wind_speed. I have too, an Excel file, with 16 specific latitudes and longitudes that I need specifically study. I need to calculate the wind speed by year but I have difficulty with that extraction. All I want is 16 (one for each set of lat/lon) new 2D matrixes that give me the years in the rows with 365 or 366 days, one day by line.

I've already specified the first day of the sample with:

N0=datenum(1979,1,1);

and extract my 16 points from an Excel file with:

[num,txt,raw] = xlsread('Coordinates.xlsx');
latPoint=squeeze(num(:,2));
lonPoint=squeeze(num(:,1));

for ii=1:16
    ilon = find(longitude(:,1)==lonPoint(ii));
    ilat = find(latitude(:,1)==latPoint(ii));
    wind_speed_points(:,ii) = squeeze(wind_speed(ilon,ilat,:));
end

So far, so good. I've already run this and works to found in the matrix wind_speed_points the values correspondent to a specific year

iyear=0;
for i=1:nt; %%nt=lenght(time)
    year = str2double(datestr(N0+i-1,'yyyy'));
    if(year == 1979)
        iyear=iyear+1;
       for ii=1:16
        w_speed_new(iyear,ii) = squeeze(wind_speed_points(i,ii));
       end
    end
end

I will have a matrix 365x16 that represent 365 days for each set lon/lat extracted from the Excel. But is not what I want. All I want is 16 matrix (one for each set) with lets say 22 rows (1979,1980,1981,1982,1983...1999,2000) and 365 or 366 lines. To create a matrix with same size the final value of each vector with 365 lines could be zero.

"22 rows and 365 or 366 lines" not possible for a matrix, this is always nXm — Daniel

Floris Floris · Accepted Answer · 2013-11-22T04:08:39

I have a little bit of difficulty in understanding your question. When all is said and done I think you want to be able to find all the wind speeds for a particular (1 in 16) lat/lon pair, for a particular year. You are given the 16 positions, and you know the date and location for each of 15,000 data points.

First of all, the math doesn't seem to add up. 15000 values over 16 locations would be a little less than 3 years' worth of data, but you seem to say you have dates from 1979 to 2000. So I am going to guess a little bit - and the guess is that you don't have data for every day of every year and every location.

If that is true, then the best solution is to collect your data in a cell array with one index for the position and one index for the year - and with a variable number of entries corresponding to the data that is available. If I'm wrong about this assumption, then a cell2mat will allow you to convert this to a 3D matrix or if you like, 16 2D matrices). The basic code you were using seems to have been pretty close.

windArray = cell(22, 16); % one entry per year, per location
year = 1979:2000;
for position = 1:16
  thisLon = (longitude(:,1) == lonPoint(position);
  thisLat = (latitude(:, 1) == latPoint(position);
  for yi = 1:numel(year)
    thisYear = (datenum(year, 1, 1):datenum(year, 12, 31)) - datenum(1979, 1, 1) + 1;
    windArray{yi, position} = wind_speed(thisLon(thisYear) & thisLat(thisYear));
  end
end

How does this work? The thisLon and thisLat vectors are boolean expressions that will be true when the lon/lat is matched. The thisYear vector has an index of every day of the year, with Jan 1, 1979 being the first date. Indexing the thisLon and thisLat vectors with the date, we find the wind_speed corresponding to just that point in space and time that you care about. The fact that you have different numbers of points for each entry in the windArray doesn't matter…

How do you use it? When you want the average wind speed at location 12, for year 5, you just do

averageWindSpeed = mean(windArray{5, 12}(:));

I hope this helps!

create a matrix with my conditions

2 Answers