Multidimensional Trend: Time unit used in nedcdf files

jonasardo

Using ArcGIS Pro 2.9.1 on windows 10 I imported a netcdf file to a Multidimensional Raster Layer. The data is a time series with one data set per year for 20 years (2001-2020). The original time dimension in the netcdf file is days  When I calculate the trend using ArcGIS Pro (Multidimensional Trend) I get very small values and I suspect that Multidimensional Trend uses days instead of year as time unit. How can I correct that?

Details in attached PDF, data attached.

Any suggestion appreciated

/Jonas Ardö

MOD17A3HGF.006_500m_aid0001.zip

Multidimensional_trend.pdf

jonasardo

Solved using "trend" in CDO (https://code.mpimet.mpg.de/projects/cdo). Gives reasonable value in line with "manual" calculations. Even if multidimensional tools are nice and very computational effective (which CDO is as well) I find it tricky to keep track of time settings etc. And why can't TREND take normal multiband rasters files as input? The ARCMAP tool Curvefit (https://www.umesc.usgs.gov/management/dss/curve_fit.html) is excellent but slow for big data sets, the data set below ( SUBDATASET_1_DESC=[20x19070x18091] Npp_500m (16-bit integer) ) took a about 3 minutes to calculate the trend (gain, offset) with either CDO  "Trend" or ARCGIS Pro "Multidimensional Trend", using the *.nc file as direct input. Interesting pattern in east Congo?

/Jonas Ardö