Changes in climate have wide-reaching implications for life on Earth. By looking at the past climate we can understand ongoing processes better. But global datasets covering the past climate have too low resolutions to be useful for small-scale investigations like crop yield modelling – until now!

Investigating the impacts of climate change is one of today’s most important tasks, and researchers from many fields are doing exactly that. They are for example climatologists, landscape ecologists, landscape managers, researchers modelling crops, and those modelling human health.

How can we investigate the past climate? 

They all use specific tools – especially one very popular tool: so-called reanalyses. These are climate datasets covering the past. Using mathematical formulas makes it possible to combine satellite data and measurements from all over the world into one big dataset. This allows us to look back in time and investigate different climate parameters over time for any place on Earth.

Is there a problem with using reanalyses?

There is not only one global reanalysis, but several different ones exist, and they differ based on what data was used to produce them. Many researchers use for example the ERA5 or the ERA5-Land reanalysis. What all reanalyses have in common is that their resolution is quite low. That means that one “pixel” or datapoint in the dataset covers an area of dozens to hundreds of square kilometers. For climate investigations for large regions this is okay, but many research fields require higher resolutions to produce relevant results. One datapoint having to cover information of a region of 900 km² does not allow for investigations on small scales.

The solution: a high-resolution downscaling of reanalysis

Several research groups have attempted to create high resolution reanalyses by taking a low resolution one and modifying it, with varying results. This is also what Richard Davy (CDP group) and former NERSC and UiB Master student Erik Kusch (now PhD candidate at Aarhus University, Denmark) did, and they presented their work in the recently published study “Reconciling high resolution climate datasets using KrigR”. Their approach is based on a code package they wrote themselves, called KrigR. They use a mathematical method called kriging; it not only makes it possible to downscale from low to high resolution, but it also gives you uncertainty in the high resolution data which has not been available before. Davy and Kusch were able to produce high resolution versions of the ERA5 and the ERA5-Land reanalyses for 1981-2010: Their products have a resolution of 0,9 km – So one datapoint in their versions covers only 0,81 km² instead of 81 or 900 km² as in the original reanalyses! They focused on surface air temperature and soil moisture content, but other parameters can be downscaled in the same way with KrigR.

A map of southern Norway and three different resolutions of reanalyses. ERA5 has a resolution of 30 km (meaning one datapoint covers a square of 900 km²), and ERA5-Land 9 km (equals a square of 81 km²). The new high resolution versions of ERA5 and ERA5-Land by Richard Davy and Erik Kusch are both at 0,9 km – this means one datapoint only covers 0,81 km²! Figure by Richard Davy

Who benefits from this work?

The difference between 900 km² and 0,81 km² is drastic, and it makes researchers in several fields very happy: Landscape ecologists can now study interactions between vegetation and the environment giving new insights into vegetation response to climate change; landscape managers can gain more insight into different processes affecting the landscape such as droughts; researchers that are modelling crops will be able to better predict crop yields benefitting the economy; and researchers investigating climate and human health can make use of the high frequency (hourly) data to better determine the effects of climate extremes on human health.

All in all, this work has great potential to have a significant impact on several research fields and ultimately on us, being dependent on understanding the climate and its effects on Earth, our home.

Reference:

Davy, R., & E. Kusch. Reconciling high resolution climate datasets using KrigR. Environmental Research Letters, 2021. doi: 10.1088/1748-9326/ac39bf