dominique briand

Hello, when I use the mabs() function (maximum of the absolute value) with the ncap2 command (I wrote a .nco script), the result is not what I expected, actualy it's the same as the max() function. Is this normal? print(@var_nm.min()); => Geomag_F, size = 1 NC_DOUBLE, value = -9.45874489844e-11 print(@var_nm.max()); => Geomag_F, size = 1 NC_DOUBLE, value = 6.54836185277e-11 print(*@var_nm.mabs()); => Geomag_F, size = 1 NC_DOUBLE, value = 6.54836185277e-11 ncap2 version 4.6.6

My variable are of different types, and have different fillValue. I think there is a link with my problem. I tried to set all elements that have the fillValue to 0, and then remove the fillValue from the "header". I did it with ncatted in two steps, ncatted -a _FillValue,,m,d,0 input output ncatted -a _FillValue,,d,, input output Is this correct? It seems to work fine, but my doubt is the d parameter of the first command (ncatted -a FillValue,,m,d,0 input output), it seems to me that I can replace...

Hello, my nco command is made to return, for a variable, the number of elements that are not equals to 0. ncap2 -s 'where(variable!=0)variable=1;print(variable.total());' input output It works for some parameters, but sometimes it doesn't! When I ncdump the output result, there are element that differ from 0 or 1... I noticed that the parameters are often of type : short, int byte or float. I don't know if it is linked... if I change the != by > or <, the result is the same! Any idea?

I can't fix my problem...but maybe it comes from the NaN values. One parameter just has NaN values, the result of my script gives me -32767, that corresponds to the _fillvalue... @all=get_vars_in(); sz=@all.size(); idx=0; print(sz); for(idx=0;idx<sz;idx++) { @var_nm=@all(idx); where(@var_nm!=0)@var_nm=1; print(int(*@var_nm).total()); }

Maybe I can explain better what I want to do, I try to detect the differences between the results of the same process over the same dataset but with different processing cluster (same software, different hardware). I first run a ncdiff, the results are sometimes really small (10^-13 for example) and I neglect them. But sometimes, the results are significative, and I noticed that isolated pixels are really different. My goal is to evealuate how many they are.

Actually I process a large number of different files, and want to work with every var to detect which var contains a non 0 value!

For example, two parameters that give me negative results byte Days_Of_Revisits(lat, lon) ; Days_Of_Revisits:_FillValue = 0b ; Days_Of_Revisits:long_name = "Days of revisits used for the retrieval" ; Days_Of_Revisits:units = "NA" ; int Mean_Acq_Time_Seconds(lat, lon) ; Mean_Acq_Time_Seconds:_FillValue = -2147483647 ; Mean_Acq_Time_Seconds:long_name = "Seconds of Mean time of acquisition for Soil Moisture value of DGG point" ; Mean_Acq_Time_Seconds:units = "s" ;

I just paste here the header of one of my files, the vars are not the same type, the fillValue are different... I read that the elements with _fillValue are not taken into account, so my *@var_nm.total() should avoid them... dimensions: inc = 15 ; lat = 584 ; lon = 1388 ; variables: double Azimuth_Angle(inc, lat, lon) ; Azimuth_Angle:_FillValue = -32768. ; Azimuth_Angle:long_name = "Angle class averaged Azimuth angle value corresponding to the measured BT value over current Earth fixed grid point....