Hello,
We collect daily precipitation data from weather stations. Some districts can have 0, 1 or many weather stations, and we need to aggregate the data in the hierarchy (district, province, national) and in time.
Without advanced GIS computing, one approximation would be to sum the precipitation in time but average spatially, but I’m having trouble configuring this in DHIS2.
The straightforward approach would be to select the appropriate Aggregation Type from the DataElement creation page, but there is no drop-down options for SUM (Average in organisation units hierarchy).
Can you recommend a configuration that would allow to achieve the desired behaviour? We use DHIS2 v40.9
Unfortunately, there isn’t a built-in aggregation type in DHIS2 that does sum in time but average across the organisation hierarchy, so there isn’t a metadata configuration that directly achieves what you’re looking for.
If you need to implement something now, here are a few options you can look into:
Option 1: Store precipitation with Sum as the aggregation type and create an indicator that divides the total precipitation by the number of contributing time series (for example by also collecting a value of 1 for each contributing time series). This gives you a simple average across contributing observations, which may be a reasonable first approximation depending on the density of the station network and the complexity of the local terrain.
Option 2: Do the aggregation outside DHIS2 and import the aggregated values back in. The Climate Tools project may provide some inspiration for working with organisation units and importing data values into DHIS2, although some custom scripting would still be needed.
One thing to keep in mind is that precipitation (like a number of other climate/weather variables) is spatially continuous and influenced by many factors. Extrapolating from point measurements at weather stations to an area is therefore not straightforward.
A common first approximation is the Thiessen polygon method, where each station represents a surrounding area, but even that doesn’t account for local geographic and atmospheric conditions (e.g. elevation, topography, distance to the coast, or prevailing winds) that influence precipitation.
More advanced interpolation and modelling is typically carried out by national meteorological services or derived from global datasets.
This is definitely a topic we’re keen to discuss further. Part of our team is already on annual leave. Most of us will be back in August, and that would be a good time to pick up the discussion again and explore possible approaches for this type of use case.