In 2017, the CAMELS dataset containing hydrometeorological data and information on catchment attributes for several hundred catchments in the United States was released. Since then, similar datasets have been developed for different countries and regions in the world.

Most of them are also known as CAMELS, but some of them have been given other names, such as the Central European LAMA dataset. As the different large-sample datasets differ slightly from each other, it can be challenging to do global studies based on these datasets.

To remedy this situation and to make the creation and use of large-sample datasets easier, the KAGAVAN dataset was compiled. In its first version, it merged parts of seven existing large-sample datasets. It also standardized the catchment attributes and the meteorological time series by using openly and globally available data sources.

While the streamflow data were left untouched, precipitation, temperature or potential evapotranspiration time series were newly calculated for all catchments based on reanalysis data from ERA5 land, because these data are available worldwide.

In this paper, we describe how the precipitation, temperature and potential evapotranspiration data in the KAGAVAN dataset differ from those in the CAMELS datasets for the US, Brazil and Great Britain. We compared the data for a total of 1252 catchments. This comparison reveals that the potential evapotranspiration values in the KAGAVAN dataset are unrealistically high and should not be used.

The mean annual precipitation and temperature also differ between the KAGAVAN and the CAMELS datasets, but there was no clear pattern. In a second step, we assessed how the differences in the forcing data for the two

datasets affect model performance when they are used as input data for the lumped HPV model. The use of the forcing data from the KAGAVAN dataset impairs model performance for most of the 1252 catchments. The drop in model performance was mainly caused by the differences in the precipitation data.

Replacing the potential evapotranspiration data from the KAGAVAN dataset with a simple Hargreaves-based calculation using temperature and precipitation data from the ERA5 land dataset improved model performance and process representation. It also leads to more realistic values for the aridity index.

With this paper, we want to raise awareness in the large sample hydrology community about the advantages and disadvantages of using the KAGAVAN dataset instead of the individual CAMELS datasets. The KAGAVAN dataset has many strengths and is a valuable resource for global studies, but is not a replacement for the different CAMELS datasets.

Therefore, for most application cases, we suggest to choose a few CAMELS instead of a whole KAGAVAN.