CHG - Data - CenTrends

In response to increasing scientific and humanitarian need to place recent seasonal and multi-annual East African precipitation extremes in a deep historic context, UC Santa Barbara Climate Hazards Group (CHG) and Florida State University (FSU) pooled their station archives and expertise to produce a high quality gridded 'Centennial Trends' precipitation dataset.


The CenTrends moniker does not imply centennial linear trends in African rainfall, but rather a dataset supporting the analysis of seasonal and decadal excursions within a centennial context. The objective of CenTrends is to provide a reasonably complete and accurate set of gridded seasonal precipitation products for East Africa (-15°S-18°N, 28°E-54° E) from 1900 to 2014 (see Data Citation 1).

Extensive quality control of the data was carried out and seasonal anomalies interpolated using kriging. Though the CenTrends dataset has more than twice as many observations as the monthly Global Historical Climate Network (GHCN), station density is still low (1 station per 10,483 km2). The CenTrends acquisition process is still a work in progress: gaps remain in Ethiopia prior to 1961, in Uganda, Rwanda, Burundi, Eritrea and Somalia after 1999, in Rwanda, Burundi, Tanzania and Somalia prior to 1921, and in Eritrea prior to 1941. Efforts will continue to improve the FSU/CHG archive.

There are important differences in the overall coverage between countries. For example, Tanzania and Kenya have relatively dense observation networks stretching back to 1900, especially in the Serengeti region. Ethiopia, on the other hand, has very few available stations available before 1961. After 1961, however, an active meteorological agency and accessible data sharing policy provide good coverage through 2014. Coverage in Eritrea and Somalia is limited in general.

The CenTrends analysis uses the Climate Hazards Groups Precipitation climatology (CHPclim) as a background climatological mean state.

Code availability

The code used to produce the CenTrends data set was written by Chris Funk and is freely available at ftp://ftp.chg.ucsb.edu/pub/org/chg/products/CentennialTrends/R_code/.

Planned update schedule and naming protocol

We plan on updating the CenTrends dataset once each year, probably each spring. Data from globally available sources (GHCN, GSOD, and GTS) will be augmented, if possible, by data acquired from national meteorological agencies. The year will be appended to the version number (1) to represent these sequential updates (CenTrends v1 will be followed by v1.2015, v1.2016, …. Changes in version number will represent either a change in the algorithm or spatial domain. Prior versions of CenTrends will be archived, and kept publically available. We would hope that version 2.0 covers all of Africa.

Data Records (unlimited length)

Version 1 of the CenTrends data set provides four sets of seasonal East African precipitation totals: March-June, June-September, October-February, and October-September. An additional set of monthly data was also produced.

The domain selected extended extends from 28°E to 54°E, 15°S to 18°N. Each field has 261 columns and 331 rows. The upper-left corner of these data center on 18°N, 28°E.

Each season's CenTrends data are provided in a single netcdf format file, following standard CF4 (Climate and Forecast) metadata conventions. Another netcdf file contains the 12 months x 115 years of monthly gridded CenTrends fields. This data (Data Citation 1) is hosted on the CHG FTP site in ftp://ftp.chg.ucsb.edu/pub/org/chg/products/CentennialTrends and can also be accessed at http://datadryad.org. T

Each individual netcdf file contains five netcdf variables (Table 4): seasonal or monthly precipitation totals [mm], seasonal or monthly precipitation kriging standard errors [mm], time in days since 1900, latitude and longitude.

Data Citations