I have a test raster that covers the world ((-180 90), (180 -90)). The SRID is WGS 84 (EPSG:4326). Values for cells in the north-western quadrant ((-180 90), (0 0)) are random integers in the range 1 to 10 (inclusive). Values for other cells are -99; this is the explicit NODATA value.
The test raster is stored in a PostGIS database (v.2.1). I'd like to compute summary statistics for a buffered point where the buffer is specified in either decimal degrees or in meters.
The Problem: My query using decimal degrees and the geometry type works as expected. However, the same query, using meters and the geography type, which I then cast back to the geometry type, does not. Can anyone suggest why this is the case?
Here's my query, where the buffer is specified in decimal degrees:
SELECT
filename,
(stats).*
FROM (
SELECT
filename,
ST_SummaryStats(
ST_Clip(
rast,
1,
ST_Buffer(
ST_SetSRID(
ST_Point(-90, 45),
4326
),
1
),
TRUE
),
1,
TRUE
) AS stats
FROM
rasters
WHERE
filename IN ('raster_grid_1_1_nodata.tif')
) AS stats_tbl;
As you can see, there are filename and rast columns that contain the name of the input file and the input raster, respectively. The result is as expected.
Here's my query, where the buffer is specified in meters:
SELECT
(summarystats).*
FROM (
SELECT
ST_SummaryStats(
ST_Clip(
rast,
1,
CAST(
ST_Buffer(
CAST(
ST_SetSRID(
ST_Point(-90, 45),
4326
) AS geography
),
1000
) AS geometry
),
TRUE
),
1,
TRUE
) AS summarystats
FROM
rasters
WHERE
filename IN ('raster_grid_1_1_nodata.tif')
) AS summarystats_tbl;
As you can see, I pass a geography to ST_Buffer so that the radius is specified in meters. The result, however, is not as expected: PostGIS reports that all pixels have the NODATA value and, consequently, does not compute summary statistics. Can anyone suggest why this is the case?
