First ERA5-Land dataset to be released this spring

Joaquín Muñoz Sabater


Following the recent release of ECMWF’s ERA5 climate reanalysis from 1979 onwards, the release of the first subset of ERA5-Land data covering the period 2001 to 2018 is planned for this spring. The dataset will be updated in a timely manner together with ERA5 updates. Like ERA5, ERA5-Land is being produced by ECMWF as part of implementing the EU-funded Copernicus Climate Change Service (C3S). This is the first time that a global land surface dataset describing the water and energy cycles and spanning nearly two decades will be available at a grid spacing of 9 km and hourly temporal frequency. The main features of this new dataset compared to previous reanalyses and ERA5 are shown in the table.

Meeting user requirements

ERA5-Land provides much finer-scale information than the land surface component of ERA5, which has a 31 km grid spacing. The figure shows the soil temperature on 15 March 2010 according to three ECMWF reanalyses: ERA-Interim, ERA5 and ERA5-Land. The enhanced resolution in ERA5-Land makes it possible to reveal much greater detail in the soil temperature structure and to resolve lower temperatures in the peaks of the Alpine region.

Soil temperatures in ERA-Interim, ERA5 and ERA5-Land. The charts show soil temperature of the top 7 cm of soil at 12 UTC on 15 March 2010 according to ERA-Interim (79 km grid spacing, left), ERA5 (31 km grid spacing, middle), and ERA5-Land (9 km grid spacing, right). The temperature values over the Mediterranean Sea in ERA-Interim and ERA5 are sea-surface temperature values.

In line with requirements for C3S datasets, ERA5-Land will provide uncertainty estimates for each variable. To begin with, this will be done through uncertainty estimates of ERA5-equivalent variables at a spatial resolution of 62 km, and at a later stage (subject to positive scientific feedback) through a dedicated 10‑member ensemble at 31 km with 3‑hourly output. ERA5‑Land is a single simulation forced by ERA5 low atmospheric meteorological fields, with additional lapse-rate correction. There is no coupling to the atmospheric model, and observations only influence the simulation indirectly through the forcing. This makes running ERA5‑Land computationally affordable. Timely reprocessing of some or all of the dataset is particularly important to support seasonal forecasting, which requires consistent land initial conditions. Such reprocessing is envisaged in case of major updates of the land model or the forcing data. In addition, in ERA5-Land the spin-up of surface variables with long memory, such as root-zone soil moisture, is significantly reduced compared to the land surface component of ERA5. This has been achieved by running a single stream for the period 2000 onwards and by using as initial conditions the last year of the ERA5 stream preceding the year 2000.







Jan 1979–present

Jan 1979–Dec 2010

Jan 1950–present

Jan 1950–present

Spatial resolution

~79 km grid spacing 
60 levels

~79 km grid spacing

~31 km grid spacing
137 levels

~9 km grid spacing

Model version


IFS Cycle 36r4


IFS Cycle 45r1

Assimilation system

IFS Cycle 31r2


IFS Cycle 41r2


Uncertainty estimate



Based on 10-member 4D-Var ensemble
at 63 km

To be based on 10-member atmospheric forcing
at 31 km

Output frequency

6-hourly analysis fields

6-hourly analysis fields

Hourly (three-hourly for the ensemble)

Hourly (three-hourly for the ensemble)

Preliminary scientific assessments of the first few years of production have been carried out for 2‑metre temperature, soil moisture, river runoff and lake fields. They show that ERA5-Land is of very good quality, adding value to ERA5 surface fields and providing users with a more accurate dataset for surface applications. The impact can be particularly important over complex terrain, where accurate orography is very important. ERA5-Land meets the growing requirement from land user communities to gain access to long-term higher-resolution datasets. In the context of the Copernicus programme, this requirement is of special relevance as water resources management, agricultural activities and drought prediction, among others, demand long-term datasets at a finer resolution than what climate reanalysis can currently provide.


Production of the dataset for the period 1979 to 2000 is currently under way, and it is planned to be released in the autumn of 2019, subject to satisfactory quality control checks. ERA5-Land will eventually be extended back to 1950 to match the period covered by ERA5. Production for this last subset is planned to start before the end of the year.