The lat­est ver­sion is V2 (released July 23, 2017)

Mul­ti-Source Weight­ed-Ensem­ble Pre­cip­i­ta­tion (MSWEP) ret­ro­spec­tive is a new ful­ly glob­al pre­cip­i­ta­tion dataset (1979–2016) with a high 3-hourly tem­po­ral and 0.1° spa­tial res­o­lu­tion. The dataset is unique in that it takes advan­tage of a wide range of data sources, includ­ing gauges, satel­lites, and atmos­pher­ic reanaly­sis mod­els, to obtain the best pos­si­ble pre­cip­i­ta­tion esti­mates at glob­al scale.

The dataset has been val­i­dat­ed at glob­al scale using pre­cip­i­ta­tion obser­va­tions from ~60,000 gauges and hydro­log­i­cal mod­el­ing for ~9000 catch­ments (Beck et al., 2017b). MSWEP was found to per­form bet­ter over­all com­pared with oth­er pre­cip­i­ta­tion datasets such as CPC Uni­fied, CHIRPS, CMORPH-CRT, GPCP-1DD, GSMaP, PER­SIANN-CCS, PER­SIANN-CDR, WFDEI-CRU, and TMPA 3B42. MSWEP was select­ed as the main pre­cip­i­ta­tion forc­ing for the state-of-the-art evap­o­ra­tion mod­el GLEAM and for Tier-2 of the EU-FP7 project eartH­2Ob­serve.

For each catch­ment, the pre­cip­i­ta­tion dataset that pro­vides the best stream­flow sim­u­la­tions. Adapt­ed from Beck et al. (2017b).

For more infor­ma­tion about MSWEP V1, see the fol­low­ing open-access paper:

The paper describ­ing and val­i­dat­ing the lat­est ver­sion of the dataset (V2) is cur­rent­ly in prepa­ra­tion. The most impor­tant changes in V2 include: (i) the use of cumu­la­tive dis­tri­b­u­tion func­tion match­ing to cor­rect the spu­ri­ous driz­zle and atten­u­at­ed peaks evi­dent in V1; (ii) increas­ing the spa­tial res­o­lu­tion from 0.25° to 0.1°; (iii) the inclu­sion of ocean areas; (iv) the addi­tion of pre­cip­i­ta­tion data from the NCEP-CFSR reanaly­sis; (v) the addi­tion of infrared-based pre­cip­i­ta­tion esti­mates for the pre-TRMM era to sup­ple­ment the reanaly­sis and gauge data; (vi) the addi­tion of 0.1° dai­ly inter­po­lat­ed gauge data, to replace the coarse 0.5° CPC Uni­fied and GPCC datasets; (vii) the use of a dai­ly gauge cor­rec­tion scheme that accounts for region­al dif­fer­ences in the 24-hour UTC bound­ary of gauge reports, to min­i­mize tim­ing mis­match­es when apply­ing the dai­ly gauge cor­rec­tions; and (viii) exten­sion of the data record to 2016.

The lat­est MSWEP tech­ni­cal doc­u­men­ta­tion, includ­ing the ver­sion his­to­ry and exam­ples on how to read the data with MAT­LAB and Python, can be viewed here.


By using MSWEP in any pub­li­ca­tion you agree to cite Beck et al. (2017a). Please read the tech­ni­cal doc­u­men­ta­tion care­ful­ly before attempt­ing to use the data. Enter your name, affil­i­a­tion, and email address to reveal the URL for down­load­ing the data.

Your Name

Your affil­i­a­tion

Your Email



MSWEP may not be used for com­mer­cial or busi­ness appli­ca­tions. MSWEP is being devel­oped by Hylke Beck (Prince­ton Uni­ver­si­ty, Prince­ton, NJ, USA). The pre­cip­i­ta­tion dataset devel­op­ers are grate­ful­ly acknowl­edged for pro­duc­ing and mak­ing avail­able their datasets. The work was sup­port­ed through IPA sup­port for the first author from the U.S. Army Corps of Engi­neers’ Inter­na­tion­al Cen­ter for Inte­grat­ed Water Resources Man­age­ment (ICI­WaRM), under the aus­pices of UNESCO. By using MSWEP in any pub­li­ca­tion you agree to cite Beck et al. (2017a).