MSWEP (Mul­ti-Source Weight­ed-Ensem­ble Pre­cip­i­ta­tion) is a new glob­al ter­res­tri­al pre­cip­i­ta­tion (P) dataset (1979–2015) with a high 3-hourly tem­po­ral and 0.25° spa­tial res­o­lu­tion (Beck et al., 2016). The dataset is unique in that it takes advan­tage of a wide range of data sources, includ­ing gauge, satel­lite, and reanaly­sis data, to obtain the best pos­si­ble P esti­mates at glob­al scale. The fol­low­ing video shows 3-hourly P from MSWEP for the peri­od June to Decem­ber 2006:

The long-term mean of MSWEP is based on the ele­va­tion-cor­rect­ed CHP­clim dataset but replaced with more accu­rate region­al datasets where avail­able. A cor­rec­tion for gauge under-catch and oro­graph­ic effects is intro­duced by infer­ring catch­ment-aver­age P from stream­flow obser­va­tions at 13,762 sta­tions across the globe.


Bias cor­rec­tion fac­tors based on stream­flow obser­va­tions (avail­able for down­load in Geo­T­IFF for­mat here). The bias cor­rec­tion is only applied in regions with con­sid­er­able snow and/​or topo­graph­ic relief.

The tem­po­ral vari­abil­i­ty of MSWEP is deter­mined by weight­ed aver­ag­ing of P anom­alies from sev­en datasets; two based sole­ly on inter­po­la­tion of gauge obser­va­tions (CPC Uni­fied and GPCC), three on satel­lite remote sens­ing (CMOR­PH, GSMaP-MVK, and TMPA 3B42RT), and two on atmos­pher­ic mod­el reanaly­sis (ERA-Inter­im and JRA-55).


The rel­a­tive con­tri­bu­tions of the gauge, satel­lite, and reanaly­sis com­po­nents of MSWEP on an arbi­trar­i­ly cho­sen day (April 25, 2006).

MSWEP has been val­i­dat­ed at glob­al scale using hydro­log­i­cal mod­el­ing for approx­i­mate­ly 9000 catch­ments and using inde­pen­dent P data from 125 FLUXNET sta­tions. The dataset was found to per­form favor­ably com­pared to pop­u­lar gauge-adjust­ed datasets such as CPC Uni­fied, TMPA 3B42, GPCP-1DD, and WFDEI-CRU. MSWEP was select­ed as the main P forc­ing for the state-of-the-art evap­o­ra­tion mod­el GLEAM and for Tier-2 of the EU-FP7 project eartH­2Ob­serve. For more infor­ma­tion about MSWEP, see the fol­low­ing open-access paper:


MSWEP is avail­able in the wide­ly used netCDF for­mat. Please enter your name and email address to receive the FTP address for down­load­ing the data. By using MSWEP in any pub­li­ca­tion you agree to cite Beck et al. (2016). MSWEP is also avail­able via a THRED­DS data server. The map with bias cor­rec­tion fac­tors, which can be used to recon­struct the orig­i­nal, non-cor­rect­ed long-term aver­ages, is avail­able here (in Geo­T­IFF for­mat).

Your Name

Your Email

Ver­sion his­to­ry

Ver­sion 1.1  (2 August 2016) — No changes to the actu­al data, the only changes are in the netCDF for­mat­ting. First, we changed the order of the vari­ables from “lon, time, lat” to “time, lat, lon”, which should solve some of the prob­lems peo­ple have had with read­ing the data. Sec­ond, we fixed the time vari­able for the dai­ly data, which was mis­tak­en­ly off­set by 1 day.

Ver­sion 1.0  (30 May 2016) — Ini­tial release cor­re­spond­ing exact­ly to the descrip­tion in Beck et al. (2016).


P17-JRC_referenceMSWEP is being devel­oped by Hylke Beck (Euro­pean Com­mis­sion, Joint Research Cen­tre, Ispra, Italy) in col­lab­o­ra­tion with Albert van Dijk (ANU, Can­ber­ra, Aus­tralia), Ad de Roo (JRC, Ispra, Italy), Vin­cen­zo Lev­iz­zani (CNR-ISAC, Bolog­na, Italy), Jaap Schellekens (Deltares, Delft, The Nether­lands), Diego Miralles (VU Uni­ver­si­ty Ams­ter­dam, The Nether­lands), and Brecht Martens (Ghent Uni­ver­si­ty, Bel­gium). We grate­ful­ly acknowl­edge the P dataset devel­op­ers for pro­duc­ing and mak­ing their datasets avail­able. The Glob­al Runoff Data Cen­tre (GRDC) and the U.S. Geo­log­i­cal Sur­vey (USGS) are thanked for pro­vid­ing most of the observed stream­flow data.  We would also like to thank the FLUXNET com­mu­ni­ty for pro­vid­ing the eddy covari­ance data. This research received fund­ing from the Euro­pean Union Sev­en­th Frame­work Pro­gram­me (FP7/2007–2013) under grant agree­ment no. 603608, “Glob­al Earth Obser­va­tion for inte­grat­ed water resource assess­ment”: eartH­2Ob­serve.