To improve dengue forecasting under the CHAP modelling framework, the DHIS2 Climate and Health Team Nepal is processing Sentinel-2 imagery at 10-m spatial resolution. We are generating monthly country-wide composites using median values. All image-processing tasks are being carried out in Google Earth Engine and the Climate Tool,
So far, we have encountered two main challenges. First, the large data volume poses storage limitations, as downloading these high-resolution datasets to local machines is required for value extraction. Second, due to the 10-day temporal resolution and persistent cloud cover especially during the monsoon season, it has been difficult to obtain cloud-free images for some months. We will therefore need to explore gap-filling strategies, such as temporal interpolation, at a later stage.
We will continue to share further progress and report any additional issues encountered. Once the high-resolution datasets are finalized, we will fit the models and compare the results with those obtained using coarser-resolution datasets, such as ERA5 and CHIRPS.
2 Likes
Using the Earth Observation Processing Framework (EOPF) to access Sentinel-2 imagery might solve some of the issues. The Zarr format makes it possible to read just the data you need without downloading entire datasets. It can be combined with their STAC API to query for data in both space and time. Read more about the tools in this blog post.
Please continue the exploration and keep sharing!
2 Likes
Okay, Thanks @Bjorn_Sandvik . We will try this.