How to count the number of pixels having values in 3Dtime series data?

Hello everyone,

I have 3D time series data (180 x 360 x 120), where 180 is latitude, 360 is longitude and 120 is number of months.

For every month, I want to count how many pixels are having values because in every month data is not available for all the 180x360 pixels.

For example: In january there are 100 pixles having data, feb = 115, march = 1224, april = 447, may = 995

Accordingy my 2D output will be : Out = [100 115 1224 447 995];

Thanks

Answer 1

[Pratyush Swain]

Hello Vedanta,

Given that you have a 3D time series data and you want to create a 2D output containing count of valid pixel values each month, you can refer to following example implementation:

% Assuming data of 3D timeseries is of dimension: 180 x 360 x 12 --> 12 is referring to the number of months here 

% Example timeseries object with random data %
ts = timeseries(rand(180, 360, 12))

  timeseries

  Common Properties:
            Name: 'unnamed'
            Time: [12x1 double]
        TimeInfo: tsdata.timemetadata
            Data: [180x360x12 double]
        DataInfo: tsdata.datametadata

% Accessing Data property %
data=ts.Data;

% Introducing some NaNs as an example of missing data %
data(data < 0.5) = NaN; 

% Preallocate the output array for performance
Out = zeros(1, size(data, 3))

Out = 1x12

0 0 0 0 0 0 0 0 0 0 0 0

% Loop through each month and count non-NaN pixels
for month = 1:size(data, 3)
    % Count the number of non-NaN elements for the current month
    Out(month) = sum(~isnan(data(:, :, month)), 'all');
end

% Display the result
disp(Out);

       32512       32488       32525       32545       32406       32425       32428       32251       32317       32313       32290       32531

For more information please refer to following resources:

1- https://www.mathworks.com/help/matlab/ref/timeseries.html

2- https://www.mathworks.com/help/matlab/ref/sum.html

Hope this helps.

[Vedanta]

@Pratyush Swain Thank you