A global map of mean annual runoff based on

discharge observations from large catchments

Hylke E. Beck ([email protected])
January 15, 2016

Abstract
This report describes the development of a global map of mean annual runoff
(MAR) based on discharge observations from 1651 large catchments (10 003 to
4 691 000 km2 ) around the globe. The map was produced based on the assumption
that the mean annual volumetric discharge difference between a station and its
upstream neighbor(s) represents the MAR generated in the interstation region. The
produced map represents a unique spatial estimate of MAR at a global scale useful
for, among other things, the spatial evaluation of macro-scale hydrological models.
Compared to an equivalent map developed at the University of New Hampshire,
the new map (i) is based on substantially more catchments (1651 versus 663),
(ii) has a higher resolution (∼0.04◦ versus 0.5◦ ), (iii) covers a larger portion of the
hydrologically-active, ice-free land surface (58.9 % versus 56.5 %), and (iv) agrees
considerably better with a completely independent MAR map based on small
catchments.

1 Introduction
Discharge is one of the major components of the hydrologic balance that, together
with precipitation, can be directly measured with relative ease. Accordingly, discharge
observations are frequently used to ensure that hydrological models accurately partition
precipitation into runoff, evaporation, and storage. This report describes the development
of a global, observation-based map of mean annual runoff (MAR) which can be used,
among other things, to spatially evaluate macro-scale hydrological models. This new
map can be considered an updated version of the one produced by the University of
New Hampshire (UNH; Fekete et al., 2002).

2 Data and methodology

The discharge observations used to produce the MAR map originate from four sources.
First, we downloaded observed discharge for 9322 USA stations part of the Geospatial
Attributes of Gages for Evaluating Streamflow (GAGES)-II database (Falcone et al.,
2010) from the U.S. Geological Survey (USGS) National Water Information System

1
(NWIS; http://waterdata.usgs.gov/nwis). Second, mean annual discharge val-
ues for 3843 stations were taken from the Global Runoff Data Centre (GRDC) website
(http://grdc.bafg.de; GRDC, 2011). Third, observed discharge for 841 stations
around the globe were requested from the GRDC. Fourth and last, we used observed
discharge for 321 Australian stations from the compilation of Peel et al. (2000). Corre-
sponding catchment boundaries were provided by the data suppliers. This resulted in an
initial dataset comprising 14 327 stations, of which only those with record length > 5 yr
(not necessarily consecutive) and catchment area > 10 000 km2 were retained. From
the remaining 2155 stations, those with interstation region (defined as the catchment of
the station excluding the subcatchments of its upstream neighbors) < 5000 km2 were
excluded to avoid artifacts in the MAR map due to systematic biases in the discharge
observations. The final dataset consisted of 1651 stations with corresponding catchment
areas ranging in size from 10 003 to 4 691 000 km2 (median 31 468 km2 ) and interstation
areas ranging in size from 5026 to 1 736 000 km2 (median 16 745 km2 ). Fig. 1 shows
the locations of the stations and the corresponding interstation regions.
The global MAR map (∼0.04◦ resolution) was produced based on the assumption
that the mean annual volumetric discharge difference between a station and its upstream
neighbor(s) represents the MAR generated in the interstation region. The same approach
was used by Arnell (1995) and Fekete et al. (2002) to produce MAR maps for Europe
and the globe, respectively. Specifically, MAR was computed following:
k
 
  Q x − ∑ Q i
365.25 × 24 × 3600  i=1 
MARx = × k
, (1)
1000  
Ax − ∑ Ai
i=1

where MARx is the mean annual (specific) runoff (mm yr−1 ) for the interstation region
of station x, Q is the observed mean annual volumetric discharge (m3 s−1 ), A is the
catchment area (km2 ), and i = 1, 2, ..., k denote the upstream stations neighboring station
x. To reduce potential biases due to seasonal gaps in the observed discharge record, for
each station Q was computed only if at least five annual mean values were available.
Annual mean values were computed only if at least ten monthly mean values were
available, while monthly mean values were computed only if at least twenty days with
flow data were available.

3 Results and discussion

Fig. 2a shows the global MAR map produced in this study. The map should be useful
for various large-scale hydrological applications, notably the spatial evaluation of large-
scale hydrological models. However, there are some caveats worth mentioning. First,
since the discharge observations are affected by water withdrawals (Döll et al., 2003)
and transmission losses (Lange, 2005), the MAR estimates may be underestimated and
may even become negative when the observed discharge decreases downstream. Second,
channel diversions will cause neighboring interstation regions to respectively overesti-
mate and underestimate MAR. Third, the map is less reliable in highly heterogeneous
regions due to the use of spatially-uniform values for the interstation regions.

2
 (a) Stations

(b) Interstation regions in arbitrary colors

Figure 1: Global maps showing (a) the stations and (b) the corresponding interstation
regions used to derive the global MAR map (n = 1651).

3
 (a) MAR map from this study (mm yr−1 )

(b) MAR map from the UNH (mm yr−1 )

(c) MAR map from the GSCD (mm yr−1 )

Figure 2: Global, observation-based MAR maps from (a) this study, (b) the UNH, and
(c) the GSCD.

4
 Fig. 2b shows the University of New Hampshire (UNH) GRDC Composite Runoff
Fields dataset (version 1.0; http://www.grdc.sr.unh.edu; Fekete et al., 2002)
MAR map which is similar to the one produced here. Although monthly data and data
for ungauged regions are provided along with the UNH MAR map, these are model-
based and therefore not of interest here. Their map has a 0.5◦ resolution and was based
on discharge observations from 663 catchments (> 10 000 km2 ) across the globe. Our
map has a slightly higher spatial coverage—considering only the hydrologically-active,
ice-free land surface1 of the earth (33.5 × 1012 km2 ), our map has a coverage of 58.9 %
versus 56.5 % for the UNH map. The markedly lower MAR value for the UNH map in
the western part of the Amazon Basin is probably due to the use of incorrect observed
discharge data for GRDC station 3623100.
Fig. 2c shows the Global Streamflow Characteristics Dataset (GSCD; version 1.9;
http://water.jrc.ec.europa.eu/GSCD/; Beck et al., 2013; Beck et al., 2015)
observation-based MAR map (0.125◦ resolution) which was used to evaluate the newly
produced and UNH maps (Figs. 2a and 2b, respectively). The GSCD MAR map was
based on discharge observations from 4072 catchments smaller than 10 000 km2 , hence
is completely independent from both the map produced here and the UNH map. To allow
quantitative comparisons, the three maps were resampled to a common 1◦ geographical
grid and square-root transformed to give more weight to arid regions. The comparisons
revealed that the newly derived map agreed considerably better with the GSCD map
than did the UNH map—considering only the regions where all three maps have data,
the mean absolute difference was 2.26 for the newly derived map versus 3.21 for the
UNH map. These values are unitless due to the applied root square transformation.

4 File details
The dataset consists of the MAR map (global runoff map.tif), the interstation
region identifier map (ID1.tif), and the stations shapefile (stations.shp). The
stations shapefile has fields ID1 (interstation region identifier), ID2 (data provider
identifier), Area1 (catchment area in km2 ), and Area2 (interstation region area in km2 ).
The maps are provided in a ∼0.04◦ (∼5 km at the equator) global geographic grid (WGS
84) in GeoTIFF format.

5 License agreement
By using the dataset you agree to cite the dataset’s Digital Object Identifier (10.5281/zen-
odo.44782) in publications that make use of the map. The dataset is licensed under
the Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.
To view a copy of this license, visit http://creativecommons.org/licenses/
by-nc-sa/3.0/deed.en_US.
1 The ice-covered portion of the land surface was identified by climate type EF (polar, frost; using the
Köppen-Geiger climate type map of Peel et al., 2007) and the hydrologically-active portion by aridity index
(AI) less than five. AI is defined as the ratio of mean annual potential evaporation (PET) to precipitation (P).
WorldClim data (Hijmans et al., 2005) were used for P and Consultative Group on International Agricultural
Research (CGIAR) data (Trabucco et al., 2008) for PET.

5
Acknowledgements
The GRDC and USGS are thanked for providing the flow data.

References
Arnell, N. W. (1995). Grid mapping of river discharge. Journal of Hydrology 167(1–4),
39–56.
Beck, H. E., A. I. J. M. van Dijk, and A. de Roo (2015). Global maps of streamflow
characteristics based on observations from several thousand catchments. Journal of
Hydrometeorology 16(4), 1478–1501.

Beck, H. E., A. I. J. M. van Dijk, D. G. Miralles, R. A. M. de Jeu, L. A. Bruijnzeel,

T. R. McVicar, and J. Schellekens (2013). Global patterns in baseflow index and
recession based on streamflow observations from 3394 catchments. Water Resources
Research 49(12), 7843–7863.
Döll, P., F. Kaspar, and B. Lehner (2003). A global hydrological model for deriving water
availability indicators: model tuning and validation. Journal of Hydrology 270(1),
105–134.
Falcone, J. A., D. M. Carlisle, D. M. Wolock, and M. R. Meador (2010). GAGES:
A stream gage database for evaluating natural and altered flow conditions in the
conterminous United States. Ecology 91(2), 621.

Fekete, B. M., C. J. Vörösmarty, and W. Grabs (2002). High-resolution fields of global

runoff combining observed river discharge and simulated water balances. Global
Biogeochemical Cycles 16(3), 15–1.
GRDC (2011). Long-term mean monthly discharges and annual characteristics of GRDC
station. Technical report, Global Runoff Data Centre (GRDC), Federal Institute of
Hydrology (BfG), Koblenz, Germany.

Hijmans, R. J., S. E. Cameron, J. L. Parra, P. G. Jones, and A. Jarvis (2005). Very high
resolution interpolated climate surfaces for global land areas. International Journal
of Climatology 25(15), 1965–1978.
Lange, J. (2005). Dynamics of transmission losses in a large arid stream channel.
Journal of Hydrology 306(1–4), 112–126.
Peel, M. C., F. H. S. Chiew, A. W. Western, and T. A. McMahon (2000). Extension
of unimpaired monthly streamflow data and regionalisation of parameter values to
estimate streamflow in ungauged catchments. Report prepared for the Australian Na-
tional Land and Water Resources Audit. Centre for Environmental Applied Hydrology,
University of Melbourne, Australia.

6
Peel, M. C., B. L. Finlayson, and T. A. McMahon (2007). Updated world map of the
Köppen-Geiger climate classification. Hydrology and Earth System Sciences 11,
1633–1644.
Trabucco, A., R. J. Zomer, D. A. Bossio, O. van Straaten, and L. V. Verchot (2008).
Climate change mitigation through afforestation/reforestation: A global analysis
of hydrologic impacts with four case studies. Agriculture, Ecosystems & Environ-
ment 126(1), 81–97.