I extracted values from several rasters, and in some cases, the value is null. However, when I export the attribute table to a csv or a txt file (from ArcMap), the null values are converted to 0. Is there any way to prevent this from happening?
These zeros are a real downer, as throw off my analysis, which is done in R.
[GIS] Null values coerced to zero (0) when exporting to csv or txt
arcgis-10.2arcmapcsvr
Related Solutions
If I understand you correctly, you want to keep the OID field. That should be simple, open attribute table, click on Add Field, name this new field for example OID_1, choose type (short integer or long integer for many features), and then use Field calculator.
The equation will be OID_1 = [OID]. After you click OK, it will copy OID values into OID_1. When the export is done, the field should be accessible.
You could try the SAGA GIS feature 'Export Grid to XYZ'. Despite the name this also supports multiband images (e.g. you could export a 'rendered GeoTiff' to get XYZRGBA, as well as getting XYZ from a 1-band DEM). You can also choose whether or not to drop NODATA rows.
I don't have any massive rasters to hand to test this, but you could try that to see if the memory usage and performance is more acceptable than gdal2xyz.
Failing that, if you need to write your own code, a python library to look at is RasterIO. This is a wrapper around GDAL/Numpy and provides a higher level interface. It also allows the windowed-reading you refer to.
Best Answer
So far, the best workaround I've found is to convert null values to to values, such as -999.99 using the field calculator. Then when I export export the attribute table I can identify these values (-999.99) as being "NULL" in my analysis.