extract without change my NaN value

Hi, I have a data of 420x32x20 (precip over timexlatxlon)

And then, I need to apply this formula into each grid:

%F=n/N×100

%F = drought frequency (%)

%n = number of drought months

%N= total number of months

Example:

Over the 420 months, in one grid, i need to find the how many month with drought value (-1 till -4) and then divide by all month which is 420, then multiple by 100.

%Find the drought value
for i=1:32
for j=1:20
SPI3_SM(:,i,j) = nnz(CHIRPS_SPI3_SM(:,i,j)<-1 & CHIRPS_SPI3_SM(:,i,j)>-3);
end
end

%The amount of drought value divide by period/months
for i=1:32
for j=1:20
Freq(:,i,j)=SPI3_SM(:,i,j)/420;
end
end

%Multi by 100 to ge the percent
for i=1:32
for j=1:20
New_freq(:,i,j)=Freq(:,i,j)*100;
end
end

But, when I use NNZ, it also change my NaN into zero. Is there any other way for me to do?

Answer 1

[DGM]

Consider:

nanmask=isnan(myarray); % keep track of nans
outputarray=functionthatdoesthings(myarray); % do a thing 
outputarray(nanmask)=NaN; % restore nans

Also, these loops are unnecessary. This thing

%The amount of drought value divide by period/months
for i=1:32
    for j=1:20
        Freq(:,i,j)=SPI3_SM(:,i,j)/420;
    end
end

%Multi by 100 to ge the percent
for i=1:32
    for j=1:20
        New_freq(:,i,j)=Freq(:,i,j)*100;
    end
end

is the same as just

Freq=SPI3_SM/420;
New_freq=Freq*100;

[Nurul Ain Basirah Zakaria]

>> load('CHIRPS_SPI3_MSIA.mat')

>> nanmask=isnan(CHIRPS_SPI3_MSIA);

>> outputarray=nnz(CHIRPS_SPI3_MSIA <-1 & CHIRPS_SPI3_MSIA >-3);

>> for i=1:40

for j=1:80

SPI3_MSIA(:,i,j) = nnz(CHIRPS_SPI3_MSIA(:,i,j)<-1 & CHIRPS_SPI3_MSIA(:,i,j)>-3);

end

end

>> SPI3_MSIA(nanmask)=NaN;

Attempt to grow array along ambiguous dimension.

It comes out like this sir.

[DGM]

The result from this:

nnz(CHIRPS_SPI3_MSIA(:,i,j)<-1 & CHIRPS_SPI3_MSIA(:,i,j)>-3);

is a scalar. You seem to be expecting it to be a vector, given the addressing on the LHS. The result is a 2D array, but you're making a 3D mask. If you want to apply the mask to the output, you'll have to figure out its significance conceptually and figure out how to deal with the excess information. Maybe you want to do

nanmask=squeeze(any(isnan(CHIRPS_SPI3_MSIA),1));

maybe you want

nanmask=squeeze(all(isnan(CHIRPS_SPI3_MSIA),1));

I don't know.

Answer 2

[Jan]

Let's simplify the code at first:

for i=1:32
    for j=1:20
        SPI3_SM(:,i,j) = nnz(CHIRPS_SPI3_SM(:,i,j)<-1 & CHIRPS_SPI3_SM(:,i,j)>-3);
    end
end

Freq     = SPI3_SM / 420;
New_freq = Freq * 100;

As far as I understand, the first loops are not useful also. Do you define SPI3_SM before the loop? What about:

SPI3_SM = sum(-3 < CHIRPS_SPI3_SM & CHIRPS_SPI3_SM < -1, 1);

What does this mean now: "it also change my NaN into zero"? Of course NaN is not in the range [-3, -1]. So what do you want to happen with NaNs?

Maybe:

SPI3_SM(any(isnan(CHIRPS_SPI3_SM), 1)) = NaN;

[Nurul Ain Basirah Zakaria]

i want it to stay as NaN. I mean,

it should be like this,