Comparison to other high resolution climate products

CHELSA, as most other climate products is a ‘model’ that aims at representing reality as precise as possible. Nevertheless, it is just a model, and has its strength and weaknesses. Here, we want to provide a simple comparison to some other products such as PRISM or WorldClim, to illustrate some of the differences between these models.

The comparison of the predictions with existing products and station data indicates an improvement in the spatio-temporal performance of the precipitation data based on cloud cover-informed downscaling. For this newly developed semi-mechanistic downscaling approach for daily precipitation, high resolution (30 arc sec) satelite-derived cloud frequency was incorporated.

For more information see:

Karger, D.N., Wilson, A.M., Mahony, C., Zimmermann, N.E., Jetz, W. (2021): Global daily 1km land surface precipitation based on cloud cover-informed downscaling. Scientific Data. doi.org/10.1038/s41597-021-01084-6

Taylor plots for comparison between CHELSA_EarthEnv (CHELSA), PRISM, CHIRPS, MSWEP, and WorldClim with GHCN-D observations for the continental United States from 2003-2016.
Bias in annual mean precipitation estimates for five different precipitation datasets compared to observations from GHCN-D and changes in absolute bias between a coarse resolution (0.25°) and the native resolution of the different datasets.
Difference in absolute bias between precipitation at 0.25° coarse resolution and precipitation at high resolution for the comparison of five different datasets to observations from GHCN-D stations based on annual means.
Comparison between mean annual precipitation rates (left) estimated from four different products with those of PRISM used as a reference data. All products compared to PRISM capture the mesoscale precipitation patterns quite well, however, differences mainly exist in the eastern Rocky Mountains, where CHIRPS, MSWEP and WorldClim are dryer compared to PRISM (right).
Case study intercomparison of CHELSA-EarthEnv and other gridded precipitation products for the Coast Range of British Columbia.

Differences in annual precipitation between CHELSA and WorldClim.

Difference between CHELSA V2.1 (left) and WorldClim V2.1 (right) over the western United States. This area is especially dense in climate stations used by WorldClim, has step precipitation gradients in the west, and a large variety of topography. Precipitation is generally lower in WorldClim compared to CHELSA.

Differences in annual precipitation between CHELSA and Prism.

Difference between CHELSA V2.1 (left) and PRISM (right) over the western United States. PRISM uses a large amount of quality controlled climate stations and for that reason can be considered one of the most precise product available. In the Colorado Mountains precipitation amounts are quite similar between both products. In this regions a large amount of precipitation falls as snow, which PRISM incorporates using snow gauges, and CHELSA incorporates using a gauge undercatch correction.

Differences between CHELSA and WorldClim at a small scale

Difference between CHELSA V2.1 (left) and WorldClim V2.1 (right) in western Canada. In this region the differences between both models are stricking. The amount of precipitation estimated in WorldClim is much lower compared to CHELSA. Additionally, precipitation gradients are reversed in WorldClim with some regions showing higher amounts of precipitation in the lowlands than in the mountains.

Differences between CHELSA and PRISM at a small scale

Difference between CHELSA V2.1 (left) and PRISM (right) in western Canada. Both models are in relative good aggrement with each other and are able to reproduce the complex precipitation gradients in this area.

Comparison to cloud cover in data sparse regions

Comparison between CHELSA V2.1 (left) and MODIS annual cloud cover (right) along the Himalayas. CHELSA predicts precipitation in areas where clouds are more frequent and less precipitation in areas where clouds are rare. While clouds can either mean rain, or no rain, no clouds usually mean no rain. In data sparse regions such as the tropics, the non occurrence of clouds can therefore be used as an indicator how well a model captures orographic rainfall.
Comparison between WorldClim V2.1 (left) and MODIS annual cloud cover (right) along the Himalayas. Notably, certain valleys which have no clouds during the year have high precipitation.