Why has catchment evaporation increased in the past 40

Hydrol Earth Syst Sci 22 5143ndash5158 2018httpsdoiorg105194hess-22-5143-2018copy Author(s) 2018 This work is distributed underthe Creative Commons Attribution 40 License

Why has catchment evaporation increased in thepast 40 years A data-based study in AustriaDoris Duethmann and Guumlnter BloumlschlInstitute for Hydraulic and Water Resources Engineering Vienna University of TechnologyKarlsplatz 13223 1040 Vienna Austria

Correspondence Doris Duethmann (duethmannhydrotuwienacat)

Received 13 March 2018 ndash Discussion started 28 March 2018Revised 11 July 2018 ndash Accepted 11 August 2018 ndash Published 4 October 2018

Abstract Regional evaporation has increased in many partsof the world in the last decades but the drivers of theseincreases are widely debated Part of the difficulty lies inthe scarcity of high-quality long-term data on evaporationIn this paper we analyze changes in catchment evapora-tion estimated from the water balances of 156 catchmentsin Austria over the period 1977ndash2014 and attribute them tochanges in atmospheric demand and available energy veg-etation and precipitation as possible drivers Trend anal-yses suggest that evaporation has significantly increasedin 60 of the catchments (p le 005) with an average in-crease of 29plusmn 14 mm yrminus1 decademinus1 (plusmn standard deviation)or 49plusmn 23 decademinus1 Pan evaporation based on 24 sta-tions has on average increased by 29plusmn5 mm yrminus1 decademinus1

or 60plusmn 10 decademinus1 Reference evaporation over the156 catchments estimated by the PenmanndashMonteith equa-tion has increased by 18plusmn 5 mm yrminus1 decademinus1 or 28plusmn07 decademinus1 Of these 21 are due to increased globalradiation and 05 due to increased air temperature accord-ing to the PenmanndashMonteith equation A satellite-based veg-etation index (NDVI) has increased by 002plusmn001 decademinus1

or 31plusmn 11 decademinus1 Estimates of reference evaporationaccounting for changes in stomata resistance due to changesin the NDVI indicate that the increase in vegetation activ-ity has led to a similar increase in reference evaporationas changes in the climate parameters A regression betweentrends in evaporation and precipitation yields a sensitivity ofa 022plusmn005 mm yrminus2 increase in evaporation to a 1 mm yrminus2

increase in precipitation A synthesis of the data analysessuggests that 43plusmn15 of the observed increase in catchmentevaporation may be directly attributed to increased atmo-

spheric demand and available energy 34plusmn14 to increasedvegetation activity and 24plusmn5 to increases in precipitation

1 Introduction

Evaporation (E) which includes transpiration throughplants is an important process in the water energy andcarbon cycles and directly controls agricultural productiv-ity and water availability for human purposes In the contextof global climate change regional E has increased in manyparts of the world in the last decades (Huntington 2006)However due to the difficulty of measuring E especially atlarge spatial scales the drivers of changing E are still de-bated

Decadal changes in catchment E may be inferred fromthe catchment water balance as storage changes are usuallysmall over decadal scales Surprisingly few studies have in-vestigated trends in water-balance-based evaporation (Ewb)Those studies that exist generally found increases in Ewbin the 20th century Examples include large river basins inthe US (Milly and Dunne 2001 Walter et al 2004 Krameret al 2015) the Tibetan Plateau (Zhang et al 2007) andcatchments in Switzerland (Spreafico et al 2007) A study of109 basins around the world in the period 1961ndash1999 foundonly few significant trends but a tendency towards positivetrends in North and South America and Europe and a ten-dency towards negative trends in Africa and Siberia (Ukkolaand Prentice 2013)

Observed trends in catchment Ewb can be complementedby point measurements although it is invariably difficultto link point and catchment scales Lysimeter data from

Published by Copernicus Publications on behalf of the European Geosciences Union

5144 D Duethmann and G Bloumlschl Why has catchment evaporation increased in the past 40 years

Rietholzbach Switzerland over 1976ndash2007 showed a de-creasing trend in E in the first half of the period and an in-creasing trend in the second half (Teuling et al 2009) Ob-servations based on eddy covariance are usually too short fortrend analyses (Wang and Dickinson 2012) but they havebeen used to train models that use satellite and climate data atlonger timescales and larger space scales (Jung et al 2010Wang et al 2010 Zhang et al 2010) While these studiesgenerally agree on positive trends since the 1980s they dif-fer in the magnitude of the estimated trends (Dong and Dai2017) It has been noted however that the results of suchmodels need to be treated with care as they are sometimesinconsistent with trends from the water balance particularlyfor wet basins (Zhang et al 2012 Liu et al 2016)

Potential drivers for changes in E are changes in avail-able energy and atmospheric evaporative demand which isdriven by variations in wind vapor deficit and air tem-perature Available energy and the evaporative demand ofthe atmosphere can be estimated based on climatic driversor measured with evaporation pans In many parts of theworld (including North America China India) annual panevaporation (Epan) has decreased in the second half of the20th century with rates of 10ndash40 mm yrminus1 decademinus1 despiteincreases in air temperature (Peterson et al 1995 Roder-ick et al 2009 McVicar et al 2012) In some instancesthis decrease in Epan has been explained by decreases in netradiation andor wind speed (Roderick and Farquhar 2002Roderick et al 2007) In other instances decreasing Epanhas been interpreted as a consequence of increasing actualevaporation (Brutsaert and Parlange 1998 Brutsaert 2013)In Europe most studies found increasing trends of Epaneg in Ireland (1963ndash2005) (Stanhill and Moumlller 2008)England (1957ndash2004 and 1986ndash2010) (Stanhill and Moumlller2008 Clark 2013) Greece (1983ndash1999) (Papaioannou et al2011) and the Czech Republic (1968ndash2010) (Trnka et al2015) In this paper we use the term atmospheric conditionsto summarize the drivers available energy and atmosphericdemand

Other potential drivers are changes in land cover and vege-tation (Piao et al 2007) Rising atmospheric CO2 concentra-tions increase plant growth (Piao et al 2007) and influencestomata closure (Gedney et al 2006) Finally changes in wa-ter availability resulting from changes in precipitation maycontribute to changing E For example Jung et al (2010)attributed the hiatus of the increasing trend in global terres-trial E during 1998ndash2008 to the limited moisture supply inthe Southern Hemisphere

Several global scale studies attributed modeled changesin E to their drivers The land surface models of Douvilleet al (2013) could only explain variations in E over 1950ndash2005 if natural forcings enhanced greenhouse gas concen-trations and aerosols were considered Based on an ensem-ble of land surface models that considered variations in cli-mate land cover atmospheric CO2 concentration and ni-trogen deposition Mao et al (2015) found that E trends

over 1982ndash2013 were dominantly driven by variations inclimate in particular precipitation Miralles et al (2014)showed that variations in E in the tropics were strongly in-fluenced by variations in precipitation driven by El NintildeondashSouthern Oscillations In contrast using a modified PenmanndashMonteith equation with detrended input variables Zhang etal (2015) concluded that the increase in global terrestrial Eover 1982ndash2013 was largely driven by vegetation greeningwhile changes in global radiation wind speed air vapor pres-sure air temperature and atmospheric CO2 concentrationhad only minor effects A full consensus regarding the driversof the increasing E that has been observed does not seemto exist The cited studies above looked at changes in mod-eledE at the global scale based on globally available meteo-rological data Complementary to these global studies thereis a need for data-based studies focusing on smaller regionswith high-quality data

The aim of this study is to (a) identify changes in catch-ment evaporation in the past 40 years and (b) identify thedrivers of these changes We use high-quality data sets ofdischarge precipitation and other climate variables from156 catchments in Austria during the period 1977ndash2014We analyze regional averages over these catchments in or-der to increase the robustness of the analysis The potentialdrivers of changes in catchment evaporation examined are (i)the atmospheric conditions quantified by reference evapora-tion (E0) and Epan (ii) vegetation quantified by a satellite-based vegetation index and (iii) precipitation as a proxy forthe available water

2 Data and methods

21 Water-balance data and water-balance estimates

211 Discharge data and catchment attributes

For the analysis of changes in the water-balance-based evap-oration (Ewb) we identified all catchments in Austria wheredaily discharge data in the period 1977ndash2014 (hydrolog-ical years November to October) with a maximum of 2years missing were available The beginning of the analy-sis period was set to 1977 because most discharge series inAustria start in the mid-1970s Catchments with substantialanthropogenic influences from dams or water withdrawals(Viglione et al 2013) catchments containing glaciers anda few high-mountain catchments where observed dischargeexceeded observed precipitation were excluded This selec-tion resulted in a total of 156 catchments (Fig 1) ranging insize from 23 to 6214 km2 (average 316 km2) Land cover wasderived from the Corine 2000 data (European EnvironmentAgency 2016) The land cover is largely dominated by forestand grassland (Table 1) Median catchment elevations werecalculated from the SRTM digital elevation model (Jarvis etal 2008) They range from 287 to 1920 m (average 910 m)

Hydrol Earth Syst Sci 22 5143ndash5158 2018 wwwhydrol-earth-syst-scinet2251432018

D Duethmann and G Bloumlschl Why has catchment evaporation increased in the past 40 years 5145

Table 1 Characteristics of the 156 study catchments in Austria

Median (lower quartileupper quartile)

Area (km2) 198 (95368)Median elevation (m) 825 (5711218)Coniferous forest () 31 (1347)Mixed and broadleaf forest () 12 (334)Natural grassland () 2 (016)Pasture grassland () 12 (619)Arable (incl heterogeneous agricultural areas) () 5 (029)Ratio of reference evaporation to precipitation 050 (038065)

Figure 1 Distribution of the study catchments in Austria and theirassociation with one of four regions

All catchments have a ratio of reference evaporation to pre-cipitation smaller than 1 and are thus classified as energy-limited or humid according to Budyko (1974) The catch-ments were assigned to regions with homogeneous variationsin climate (Fig 1) that were derived from a multi-variable(temperature precipitation sunshine duration air pressure)principal component analysis (Matulla et al 2003 Matulla2005 Auer et al 2007)

212 Catchment average meteorological data

Air temperature and precipitation were obtained from thegridded SPARTACUS data set (Hiebl and Frei 2016 2018)This data set has temporal and spatial resolutions of 24 hand 1 km respectively and was designed to be suitablefor trend analyses The interpolation method of minimumand maximum air temperature accounts for nonlinearities inthe thermal profile and uses a constant station network of150 stations Precipitation is based on a two-step interpola-tion scheme in which 1249 stations (including 119 totalizerprecipitation gauges) were used for obtaining a daily back-ground climatology for 1977ndash2006 and a constant numberof 523 stations was used for interpolating ratios between

the daily precipitation and the background climatology Toaccount for the systematic underestimation from gauge un-dercatch we corrected the gridded precipitation data setfor gauge undercatch using the following equation (Richter1995)

Pcorr = Porig+ b middotPeorig (1)

where Pcorr is undercatch corrected precipitation Porig un-corrected precipitation and b e are coefficients that dependon precipitation type and wind exposure We estimated theprecipitation type as snow for mean air temperatures belowminus1 C as rain for mean air temperatures above 3 C andas mixed precipitation between minus1 and 3 C (ATV-DVWK2001) The coefficients of Richter (1995) for moderatelysheltered locations were applied to all grid points

Measurements of relative humidity at 0700 and 1400 LTand global radiation were provided by the Austrian CentralInstitute for Meteorology and Geodynamics (ZAMG) Sta-tions with more than 5 (15 for global radiation) miss-ing data during 1977ndash2014 (hydrological years November toOctober) were excluded which resulted in 125 and 6 stationsfor relative humidity and global radiation respectively Datagaps were filled using linear regression to the station with thehighest correlation The data were interpolated onto a 1 km2

grid using local ordinary least-squares regression with ele-vation The local neighborhood was set to a default radius of100 km for relative humidity and 200 km for global radiationThis was adjusted to include a minimum of 10 (global radia-tion 4) and a maximum of 40 stations The grid values wereaggregated to the catchment average series

Trends in wind data were not included in the analysissince station observations of wind speeds are known to beprone to inhomogeneities (Boumlhm 2008) annual anomalies ofwind speed data from 85 stations in Austria appear to be un-related to each other (Fig S1a in the Supplement) and tem-poral trends over 1977ndash2014 do not show any spatial pattern(Fig S2a) (see Supplement S1) Furthermore interpolatingwind data in space results in high uncertainties We thereforeused uniform monthly wind speeds averaged over all yearsfrom all stations in Austria The potential effect of changesin wind speed on evaporation is analyzed in Supplement S2

wwwhydrol-earth-syst-scinet2251432018 Hydrol Earth Syst Sci 22 5143ndash5158 2018

213 Estimating evaporation from the waterbalance (Ewb)

The catchment water balance can be written asdSsnow

dSsoil

dt= P minusQminusE (2)

where Ssnow is snow Sice ice Ssw surface water Ssoil soilwater Sgw groundwater storage P precipitation and Q dis-charge In order to be able to estimate E from the water bal-ance some assumptions on storage changes need to be madeOver periods of several years we may assume that changesin surface water soil and snow storage are small Studieson groundwater-level changes in Austria do not show large-scale groundwater changes over the study period (Blaschkeet al 2011 Neunteufel et al 2017) Trends in annual meangroundwater levels at 2114 sites in Austria over 1976ndash2006showed a heterogeneous picture with decreasing trends (p le005) at 18 of the sites increasing trends at 12 of thesites and insignificant trends at 70 of the sites (Blaschkeet al 2011) We therefore assume that changes in ground-water storage (and changes in any groundwater fluxes) aresmall Estimates of absolute values of Ewb (trends in Ewbare not affected) furthermore depend on the assumption thatgroundwater fluxes across the catchment boundaries can belargely neglected which is supported by prior rainfall-runoffstudies in the catchments that suggest that the water balancecan be closed (Parajka et al 2005) Catchments with glaciershave been excluded from the analysis For the timescale ofdecades catchment evaporation can therefore be estimatedas Ewb = P minusQ Since annual data of Ewb might be influ-enced by storage effects we applied a Gaussian filter with astandard deviation of 2 years for the graphical presentationof variations of Ewb over the study period

22 Pan evaporation data

Daily pan evaporation (Epan) data using the GGI-3000 evap-orimeter were provided by the Central Hydrographical Bu-reau (HZB) in Vienna and by ZAMG in Vienna Furthermonthly Epan data were obtained from the MeteorologicalYearbook (ZAMG 1977ndash1990) Coordinates elevations aswell as means and standard deviations of the warm-seasonEpan of the stations are listed in the Supplement (Table S1)Missing values in the ZAMG data were replaced by estimatesfrom an empirical Dalton-type formula with locally derivedcoefficients if wind speed and saturation deficit values wereavailable (Neuwirth 1978) Missing values in the HZB datawere replaced by estimates from the climate factor method(Hydrographischen Zentralbuumlro 1996) with locally derivedcoefficients if wind speed air temperature and relative hu-midity were available Remaining negative values and val-ues larger than 15 mm dayminus1 were considered erroneous andflagged as missing data Gaps with a maximum gap size offour days were linearly interpolated Totals over the summerhalf year (May to October) were calculated if more than 90

of the daily values or all monthly values were available Dueto the uneven record lengths we analyzed trends for three pe-riods 1979ndash2005 (13 stations) 1983ndash2014 (8 stations) and1993ndash2014 (16 stations) which includes only series with alength of at least 20 years and 5 years or less missing fromthe record Since trend analyses at individual stations did notshow regional differences normalized data of all series witha length of at least 20 years during 1977ndash2014 (24 stations)were pooled to a common series The data were normalizedby subtracting the mean over the overlapping period 1993ndash2005 excluding 1995 and 1998 which had many missing val-ues at several stations Summer Epan was upscaled to the fullyear by the average ratio of annual and summer referenceevaporation (133) for comparison

23 Estimation of reference evaporation

231 Reference evaporation

In order to examine the effects of changes in atmosphericconditions we estimated reference evaporation (E0) by thePenmanndashMonteith equation for well-watered short grass veg-etation (Allen et al 1998) This is represented by

E0 = 0408 middot1 middot (RnminusG)+ γ middot

185 400(T+273)middotra

middot (esminus ea)

1+ γ middot(

1+ rsra

where Rn is the net radiation at the crop surface(MJ mminus2 dayminus1) G is the soil heat flux density(MJ mminus2 dayminus1) T is the mean air temperature at 2 mheight (C) ra is the aerodynamic resistance (s mminus1) rs isthe surface resistance (s mminus1) es is the saturation vaporpressure (kPa) ea is the actual vapor pressure (kPa) 1 isthe slope of the vapor pressure curve (kPa Cminus1) and γ isthe psychrometric constant (kPa Cminus1) According to thereference conditions of a vegetated surface with a heightof 012 m rs = 70 s mminus1 and ra = 208u2 where u2 is thewind speed at 2 m height (m sminus1) that was derived fromthe wind speed at 10 m height based on a logarithmic windspeed profile (Allen et al 1998) The ground heat flux wasneglected The vapor pressure deficit esminus ea was calculatedas the average of the vapor pressure deficit at the minimumair temperature (using relative humidity at 0700 LT) andat the maximum air temperature (using relative humidityat 1400 LT) Rn was calculated from global radiation (RsMJ mminus2 dayminus1) albedo (α set to 023) and net longwaveradiation (Rnl MJ mminus2 dayminus1) represented by the equation

Rn = α middotRs+Rnl (4)

where Rnl was estimated according to Allen et al (1998)based on minimum and maximum air temperature clear-skysolar radiation measured Rs and the mean daily vapor pres-sure E0 was calculated on a daily basis on a 1 km2 grid andaggregated to catchment average annual values (hydrologicalyears November to October)

The average contributions of the input variables net radia-tion air temperature and vapor pressure deficit to the trendof E0 were evaluated using estimates of E0 with one or sev-eral of the input variables held fixed to a particular year Theyear 1994 was selected since annual mean values of E0 andits input variables were close to the mean value over the studyperiod For this analysis the daily series of catchment evap-oration were estimated from input data aggregated to catch-ment average daily values in order to reduce the computingtime

The contribution ϕiE0 of variable i to the trend of E0in catchment k was calculated as (see eg Galbraith et al2010)

ϕiE0(k)=τi(k)minus τc(k)

τE0(k) (5)

where τc(k) is the trend of the control (where all input vari-ables are kept to those of 1994) τi(k) is the trend ofE0 calcu-lated with only variable i varying over the study period (andall other inputs as for the control) and τE0(k) is the trend ofE0 with all input variables varying The two-way interactioneffect ϕitimesjE0 of the variables i and j in catchment k wascalculated as

ϕitimesjE0(k)=τitimesj (k)minus τc(k)

τE0(k)minusϕiE0(k)minusϕjE0(k) (6)

where τitimesj (k) is the trend of E0 calculated with variable iand j varying over the study period (and all other inputsas for the control) For average effects and their variabil-ity we calculated averages and spatial standard deviationsof ϕiE0(k) and ϕitimesjE0(k) over all catchments

232 Effect of changes in vegetation activity onreference evaporation

In order to examine changes in vegetation we used the Nor-malized Difference Vegetation Index (NDVI) based on satel-lite data Changes in vegetation activity as observed by theNDVI represent an integrated signal of changes in phenol-ogy leaf area index vegetation fraction vegetation typeand land cover Observed 15-day maximum value compositeNDVI data at a resolution of 8 km from the Advanced VeryHigh Resolution Radiometer (AVHRR) for 1982ndash2014 wereobtained from Tucker et al (2005) The NDVI data were ag-gregated to catchment averages and linearly interpolated todaily catchment average series

In the PenmanndashMonteith equation an increase of the vege-tation activity reduces rs which increases the potential evap-oration We calculated the reference evaporation consider-ing a variable rs (E0v) using Eq (3) applying a variable rsderived from the satellite data instead of a constant rs of70 s mminus1 The vegetation effect was estimated by calculat-ing (i) E0v from the original NDVI series and (ii) E0c from adetrended NDVI series

We applied two approaches for estimating rs from theNDVI to consider the uncertainty of these estimates Inthe first approach rs was estimated from the leaf area in-dex (LAI) and the fraction of photosynthetically active radia-tion (FPAR) From Sellers et al (1996) FPAR was estimatedas

FPAR=(Sminus Smin)

(Smaxminus Smin)middot (FPARmaxminusFPARmin)+FPARmin (7)

where S is a transformed NDVI value(1+NDVI)(1minusNDVI) and Smin and Smax are the 5 and 98 quantiles of S for a given land cover class TheLAI was estimated from the FPAR (Sellers et al 1996) as

LAI= LAImax middotlog(1minusFPAR)

log(1minusFPARmax) (8)

where LAImax is the maximum LAI of a land cover classIn Eqs (7) and (8) we applied the following coeffi-cients for grassland NDVImin = 0039 NDVImax = 0674FPARmin = 0001 FPARmax = 095 and LAImax = 5 (Sell-ers et al 1996) rs was estimated as rs = rl middot (LAI middot 05)minus1assuming a leaf stomata resistance rl of 100 s mminus1 for well-watered grass (Allen et al 1998)

In the second approach we used the relationship be-tween rs and the NDVI of Zhang et al (2010) which is rep-resented as

rs(NDVI)=(

1b1+ b2 middot exp(minusb3 middotNDVI)

)minus1

where b1 = 175 s mminus1 b2 = 2000 s mminus1 b3 = 6 and b4 =

minus1(b1+ b2) (coefficients for grass)The average contributions of changes in atmospheric con-

ditions and of changes in vegetation to the trend in E0v(ϕatmE0v and ϕvegE0v ) averaged over all catchments and thetwo approaches for estimating rs from the NDVI were esti-mated as follows

ϕatmE0v =1

nsumk=1

2suml=1

τE0c(k l)

τE0v(k l)

ϕvegE0v =1

nsumk=1

2suml=1

τE0v(k l)minus τE0c(k l)

τE0v(k l) (10)

where k is the catchment index n the total number of catch-ments l refers to one of the two approaches for estimating rsfrom the NDVI τE0c(k l) is the trend inE0c and τE0v(k l) isthe trend in E0v of catchment k when using approach l

24 Trend analyses regression analyses andattribution of the trend in Ewb

Trends were estimated by the Senrsquos slope estimator (Sen1968) Trend significance was assessed by the nonparametricMannndashKendall test (Mann 1945 Kendall 1975) The trend-free prewhitening technique was applied to remove lag-one

serial correlation (Yue et al 2002) Uncertainties in the trendmagnitude were estimated using a bootstrapping approachFor this purpose 1000 samples of size N were drawn withreplacement from the record of length N years The Senrsquosslope was calculated from each of the 1000 samples and thestandard deviation was determined The trends and the stan-dard deviations were first calculated for each catchment andthen averaged over the catchments to derive average trendsand their uncertainties over a number of catchments

The trends in Ewb in the individual catchments were re-lated to the respective trends in E0 the NDVI and meanannual precipitation by regression analysis in order to un-ravel the relation between changes in Ewb to changes in at-mospheric conditions changes in vegetation and changes inprecipitation

The contributions of the different drivers to the increaseinEwb were estimated as follows The sensitivity of the trendin Ewb to trends in precipitation was estimated based on theslope of the linear regression between trends in Ewb andtrends in precipitation Since Ewb is estimated from precipi-tation and discharge trends in Ewb are not independent fromtrends in precipitation and the regression relationship mayoverestimate the effect of trends in precipitation on trendsin Ewb The magnitude of this overestimation was estimatedby Monte Carlo simulations with correlated annual precipi-tation and discharge series generated according to the statis-tics of the data (see Supplement S3) The sensitivity of thetrend in Ewb to trends in precipitation (sprecEwb ) was esti-mated as the slope of the linear regression between trendsin Ewb and trends in precipitation corrected by the overesti-mation effect estimated by the Monte Carlo simulations Theaverage contribution of changes in precipitation to the trendin Ewb (ϕprecEwb ) was estimated as

ϕprecEwb =sprecEwb middot τprec

where τprec and τEwb are average trends over all catchmentsin precipitation and Ewb The contributions of changes in at-mospheric conditions and vegetation could not be estimatedin a similar way since the trends in E0 or in the NDVI werenot related to the trends in Ewb (see Sect 321 and 323)This is probably due to a relatively low spatial variability inchanges in E0 and changes in the NDVI While the spatialvariability of changes in precipitation is relatively high thespatial variability of changes in available energy which is animportant driver for changes in E0 and in the NDVI is low

A different approach was therefore used for estimating theaverage contributions of changes in atmospheric conditionsand vegetation to the trend in Ewb (ϕatmEwb and ϕvegEwb )Assuming that the remainder of the trend in Ewb is causedby changes in atmospheric conditions and vegetation theircontributions were estimated according to the ratio of theireffects on E0v represented by

ϕatmEwb =(1minusϕprecEwb

)middotϕatmE0v

ϕvegEwb =(1minusϕprecEwb

)middotϕvegE0v (12)

where ϕatmE0v and ϕvegE0v are the contributions of changesin atmospheric conditions and in vegetation to the trendin E0v (see Eq 10) averaged over all catchments and bothparameterizations for rs

Uncertainties in the attribution estimate are based on thestandard deviation of the regression slope of the trend in pre-cipitation against the trend in Ewb the standard deviationsof ϕatmE0v and ϕvegE0v over all catchments as well as thetwo parameterizations for rs

3 Results

31 Changes in evaporation estimated from the waterbalance (trend detection)

Catchment Ewb trends increased significantly (p le 005) in93 out of the 156 catchments (60 ) during 1977ndash2014 Onecatchment shows a significant decreasing trend On averageover all catchments the annual Ewb increased with a rateof 29plusmn 14 mm yrminus1 or 49 plusmn 23 per decade (plusmn standarddeviation of the trend the standard deviation refers to theaverage uncertainties of the trend estimates) The increasewas largest during 1980ndash1995 and flattened out later (Fig 2aand d) The increase in Ewb is more consistent over spaceand time than the changes in precipitation and discharge (Ta-ble 2 Figs 2b c and S6) which would be expected and addscredence to the estimates

Annual precipitation trends increased significantly (p le005) in 64 out of the 156 catchments (41 ) with an averageincrease of 32plusmn 23 mm yrminus1 or 24plusmn 17 per decade Twocatchments show significant decreasing trends Increaseswere particularly large in the eastern Alpine region of thestudy domain and generally occurred in summer (Figs S7and S8)

Discharge trends increased significantly (p le 005) in16 out of the 156 catchments (10 ) and decreased signifi-cantly in 9 catchments (6 ) resulting in an average trend of2plusmn 23 mm yrminus1 or 02plusmn 31 per decade Catchments withincreasing discharge are located in the eastern Alpine regionwhere precipitation increased the most Catchments with sig-nificant decreasing trends are located in the west of Austriawhere precipitation did not change much (Fig S6c) Interest-ingly the decadal fluctuations of the discharge series withinthe study period are very similar to those of the precipitationseries (Fig 2b and c)

The analyses in this study are based on undercatch cor-rected precipitation using coefficients for moderately pro-tected locations In order to analyze the sensitivity of thecorrection assumption on the estimated trends in Ewb weestimated trends in Ewb using uncorrected precipitation andundercatch corrected precipitation with correction parame-ters for unprotected locations Without undercatch correc-tion estimates of average catchment precipitation would be

Table 2 Means and trends of catchment evaporation estimated from the water balance (Ewb) precipitation and discharge Numbers givenare the spatial averages over regions (and the entire study area) of the mean and the standard deviation

Ewb Precipitation Discharge

Mean Trend Mean Trend Mean Trend(mm yrminus1) (mm yrminus1 (mm yrminus1) (mm yrminus1 (mm yrminus1) (mm yrminus1

decademinus1) decademinus1) decademinus1)

North 625 31plusmn 13 1235 38plusmn 22 610 7plusmn 21Southeast 668 27plusmn 13 1060 17plusmn 20 392 minus8plusmn 16South 556 27plusmn 15 1411 45plusmn 25 855 16plusmn 24West 546 27plusmn 19 1931 minus13plusmn 28 1385 minus43plusmn 31

All 604 29plusmn 14 1339 32plusmn 23 735 2plusmn 23

Figure 2 Anomalies of (a d) catchment evaporation estimated from the water balance (Ewb) (b e) precipitation and (c f) dischargeover 1977ndash2014 (a)ndash(c) Mean anomalies by region Data smoothed using a Gaussian filter with a standard deviation of 2 years (d)ndash(f) Meananomalies over all catchments The thin blue line shows the mean over all catchments the grey shaded area the variability between catchments(plusmn1 SD ndash standard deviation) the bold black line the smoothed mean and the dashed red line the linear trend

9 lower and the resulting estimates of Ewb would be 20 lower than with undercatch correction for moderately pro-tected locations (Table 3) The percentage of catchments withsignificant increasing trends in Ewb would increase from60 to 65 and the average trend in Ewb would increasefrom 29 to 31 mm yrminus1 or from 49 to 64 per decadeUsing undercatch correction for wind-exposed stations hasan effect of similar magnitude but of opposite directionThese results show that while undercatch correction of pre-cipitation has a strong effect on average Ewb it only mod-erately affects its trends Please note that in all figures andtables of this paper with the exception of Table 3 the precip-itation undercatch has been corrected

TheEwb estimates were used to estimate ratios of actualEto Emax the maximum possible evaporation under the actualvegetation when soils are wet A ratio close to unity wouldsuggest precipitation not to be a likely driver of increasesin Ewb Since the land cover in the study catchments is dom-inated by forest (average fraction over all study catchmentsof 052) Emax is likely much higher than E0 (eg Teul-ing 2018) Analyses from non-weighable lysimeters sug-gest Emax to be 20 ndash30 higher than E0 for sites withpine forests at typical stand ages of 80ndash100 years comparedto sites with grass (ATV-DVWK 2001) We estimated Emaxfor each catchment as Emax = E0 middot

sum(li middot fi) where li is the

fraction of land cover i and fi is the ratio of EmaxE0 for

Table 3 Effect of undercatch correction on estimates of average precipitation (P ) Ewb and their trends (averages over all study catchmentsin 1977ndash2014 significant trends for p le 005)

Undercatch correction Average P Percentage of Average Average Percentage of Average(mm yrminus1) catchments increase in P Ewb catchments increase in Ewb

with sign (mm yrminus1 (mm yrminus1) with sign (mm yrminus1

increases in decademinus1) increases in decademinus1)P () Ewb ()

Parameters for wind-exposed locations 1414 35 271 679 54 258Parameters for moderately protected locationslowast 1339 41 316 604 60 293No undercatch correction 1221 45 337 486 65 310

lowast Parameters for moderately protected locations are used for all other analyses in this paper

land cover i which was approximated as 12 for forests and1 for all other land cover types This results in median (up-perlower quantile) values for EwbEmax of 084 (077091)suggesting that in most catchments Ewb is significantlylower than Emax It has to be noted that these estimates areuncertain due to uncertainties in the factor f and the absoluteestimates of Ewb

32 Drivers of the increases in evaporation (attribution)

321 Changes in atmospheric conditions ndash referenceevaporation

Changes in the atmospheric conditions were examined byanalyzing E0 and Epan Averaged over all catchments an-nual E0 increased by 18plusmn 5 mm yrminus1 or 28plusmn 07 perdecade during 1977ndash2014 (Fig 3a) Spatial variations in theincrease in E0 are small and there is no significant correla-tion between trends in Ewb and trends in E0 (r2

= 002 p =009) (Fig 8a) Partial correlations between trends in Ewband E0 when trends in annual precipitation and the NDVIwere accounted for are not significant either

Over our study period global radiation on average in-creased by 51plusmn 09 W mminus2 or 38plusmn 07 per decade(Fig 3b) Mean air temperature (calculated as the aver-age of minimum and maximum air temperatures) increasedby 045plusmn 009 C decademinus1 (Fig 3c) while vapor pres-sure deficit showed variations with higher values during theearly 1990s but no trend over 1977ndash2014 (Fig 3d) The anal-ysis of the E0 estimates with one or several of the input vari-ables held fixed to a particular year showed that the increasein E0 was largely driven by increasing net radiation whichcontributed 76plusmn 8 to the trend in E0 and increases in airtemperature which contributed 19plusmn 6 to the trend in E0(Fig 4)

A trend analysis of wind speed observations shows largescatter between the stations Averaged over all stations windspeeds decreased by minus30plusmn 25 per decade (see Supple-ment S3) According to a simple analysis the average trendinE0 reduces to 24plusmn07 per decade when allowing for de-creasing wind speeds as compared to 28plusmn07 per decadewhen assuming no trends in wind speed Thus negative

Figure 3 Anomalies of (a) E0 (b) global radiation (c) meanair temperature (d) vapor pressure deficit over 1977ndash2014 and allstudy catchments The thin blue line shows the mean over all catch-ments the grey shaded area the variability between catchments(plusmn1 SD) the bold black line the filtered mean (Gaussian filter witha standard deviation of 2 years) and the dashed red line the lineartrend

trends in wind speed probably slightly reduced the effectsof the positive trends in global radiation and air temperatureon E0

322 Changes in atmospheric conditions ndash panevaporation

Pan evaporation significantly (p le 005) increased at 5 of13 stations over 1979ndash2005 at 4 of 8 stations over 1983ndash2014 and at 5 of 16 stations over 1993ndash2014 there are nosignificant negative trends (Fig S9 Table S1) and the re-

Figure 4 Mean contributions of variations in net radiation (rad)air temperature (T ) and vapor pressure deficit (vpd) and their two-way and three-way interaction effects on the trend in E0 Bars showmeans over all catchments and error bars show the standard de-viation of the variation between catchments Percent is relative totrends in E0

gional differences in the trends are small The average nor-malized Epan series shows a highly significant (p le 001) in-crease of 29plusmn5 mm yrminus1 or 60plusmn10 decademinus1 over 1977ndash2014 (Fig 5)

323 Changes in vegetation activity

Catchment average NDVI shows a clear seasonal cycle withhigh values in summer and low values in winter (Figs 6aand S10) Mean trends in the NDVI over all catchments for15-day composites are positive nearly over the course of theentire year and are particularly strong during MarchApriland NovemberDecember (Fig 6b) Timing of the trends iscorrelated to median catchment elevation During OctoberndashMarch positive trends are mostly observed in catchmentswith low median elevation while during MayndashJune this re-verses and stronger positive trends are observed in high ele-vation catchments (Fig S11) The average and standard de-viation over all catchments of the trend in the average annualNDVI is 002plusmn 001 decademinus1 or 31plusmn 11 decademinus1

To estimate the effect of these vegetation changes on Ewe calculated E0v with rs estimated based on the originalobserved NDVI data and E0c with rs based on detrendedNDVI data E0v which reflects changes in vegetation activ-ity and atmospheric conditions showed a stronger increaseover the study period than E0c which reflects changes in at-mospheric conditions only (Fig 7a and b) Estimated as aver-age and standard deviation over all catchments and over bothapproaches for estimating rs the contribution of changes inatmospheric conditions to the trend inE0v was 56plusmn15 andthe contribution of changes in vegetation to the trend in E0vwas 44plusmn 15 (Fig 7c)

Changes in annual cumulative NDVI are not correlated tochanges in Ewb (r2

= 001 p = 023) (Fig 8b) This wasalso the case for partial correlations between trends in Ewb

Figure 5 Pan evaporation anomalies of 25 stations with a minimumof 20 years available data during 1977ndash2014 The thin blue lineshows the mean the grey area plusmn1 SD and the thick blue line thefiltered mean (Gaussian filter with a standard deviation of 2 years)The red (orange green) lines show means for subsets of stationsin 1979ndash2005 (1983ndash2014 1993ndash2014 maximum of 5 years miss-ing during the period)

and the NDVI when trends in annual precipitation or trendsin E0 were taken into account

324 Changes in precipitation

We used annual precipitation as a proxy for the water avail-able for E Trends in annual precipitation are described inSect 31 Higher increases in Ewb are observed in catch-ments with higher increases in annual precipitation (r2

024 p lt 00001 Fig 8c) The trends in precipitation arenot related to trends in the NDVI (r =minus001 p = 087)which suggests that the relationship between trends in Ewband trends in precipitation does not include indirect effectsof changes in precipitation on trends in Ewb through changesin vegetation activity Trends in precipitation are also notrelated to trends in E0 (r =minus012 p = 014) A linear re-gression of the trend in Ewb against the trend in annual pre-cipitation results in a slope of 030plusmn 004 This relationshipmay however overestimate the influence of changes in pre-cipitation on changes in Ewb since Ewb is derived from thewater balance and trends in Ewb are thus not independentfrom trends in precipitation The magnitude of this over-estimation effect was estimated using Monte Carlo simula-tions as 008plusmn 003 (see Supplement S3) and the estimatefor the sensitivity of trends in Ewb to trends in precipita-tion was therefore corrected to 022plusmn 005 This suggeststhat a 1 mm yrminus2 increase in precipitation is associated witha 022plusmn 005 mm yrminus2 increase in Ewb Thus with an av-erage precipitation trend of 32 mm yrminus1 decademinus1 on aver-age 69plusmn 16 mm yrminus1 decademinus1 (uncertainty derived fromthe standard deviation of the trend slope and the correctionvalue) of the Ewb trend may be related to the increase in pre-cipitation

Figure 6 Changes in NDVI (a) Seasonal cycle of NDVI averaged over 1982ndash1986 and 2010ndash2014 Solid lines show averages over allcatchments and shaded areas show plusmn1 SD (b) Seasonal cycle of trends in catchment average NDVI values over 1982ndash2014 The solid lineshows the mean NDVI trend over all catchments the grey shaded area the spatial variability of the trends between catchments (plusmn1 SD) andthe blue line the fraction of significant positive trends (p le 005)

Figure 7 Effect of variations in atmospheric conditions and vegetation on E0v (a b) Anomalies of E0v with rs estimated from originaland detrended NDVI data with (a) rs estimated based on Sellers et al (1996) and (b) rs estimated based on Zhang et al (2010) The thinline shows the mean over all catchments the grey shaded area shows the variability between catchments (plusmn1 SD) and the thick blue lineshows the filtered mean (Gaussian filter with a standard deviation of 2 years) The number in brackets gives the trend estimate over 1982ndash2014 (c) Contributions of variations in atmospheric conditions and variations in vegetation to the trend in E0v Bars show means over allcatchments and over both approaches for estimating rs and error bars show the variability over all catchments and over both approaches forestimating rs (plusmn1 SD) Percent is relative to trends in E0v

33 Synthesis of attribution

We may now estimate the contributions of the differentdrivers to the increase in Ewb The regression of the trendin Ewb against the trend in annual precipitation (Fig 8c)suggests that on average 69plusmn16 mm yrminus1 decademinus1 of theEwb trend of 293 mm yrminus1 decademinus1 may be related to theincrease in precipitation (Sect 324) The relative contri-butions of atmospheric conditions and vegetation were as-sumed to conform to their relative effects onE0v (Sect 323Fig 7c) Thus the remaining 224plusmn 16 mm yrminus1 decademinus1

are split at a ratio of 056plusmn 015 to 044plusmn 015 into beingdue to atmospheric conditions and vegetation respectivelyThis results in a contribution of changes in atmospheric con-ditions of 125plusmn 42 mm yrminus1 decademinus1 and in a contribu-tion of changes in vegetation of 98plusmn40 mm yrminus1 decademinus1

In summary the data suggest that changes in atmosphericconditions vegetation activity and precipitation have con-tributed 43plusmn15 34plusmn14 and 24plusmn5 respectively tothe average increase in Ewb in the study catchments (Fig 9)

4 Discussion

41 Changes in catchment evaporation in Austria overthe past four decades

Based on the analysis of the water balances we found an av-erage Ewb increase of 29plusmn14 mm yrminus1 decademinus1 ie a totalincrease of 108plusmn52 mm yrminus1 or 180plusmn85 over 37 yearsThis increase is consistent between the different regionswhich points to the importance of drivers with spatially con-sistent changes across the study region ie global radiation

Figure 8 Scatter plots of trends in (a) E0 (b) NDVI and (c) annual precipitation against the trend in Ewb

Figure 9 Average contributions to the average trend of catchmentevaporation estimated from the water balance (Ewb) from changesin atmospheric conditions vegetation activity and precipitation Er-ror bars relate to the standard deviation of the estimate Percent re-lates to the total average trend of Ewb of 29 mm yrminus1 decademinus1

or air temperature rather than changes in precipitation Theincrease is strongest in the beginning of the study period andseems to have stopped around 2000 A similar pattern is ob-served for E0 and E0v (Figs 3a and 7a b) where the de-creasing tendency in global radiation (Fig 3b) appears to bethe main cause for the decreasing tendency after 2000

The Ewb estimates could potentially be influenced bychanges in groundwater storage Changes in groundwaterstorage were assumed small over timescales of decades Thisassumption is supported by the absence of general trendsin discharge together with the incoherent picture of trendsin groundwater levels (Blaschke et al 2011 Neunteufel etal 2017) which makes it unlikely that groundwater storagechanges have a big influence on the Ewb estimates

Increases in Ewb of a similar magnitude have alsobeen observed in Switzerland where Ewb increased bysim 20 mm yrminus1 decademinus1 over 1977ndash2007 (Spreafico et al2007 from Fig 1 on plate 66) Lower rates of increaseinEwb were observed in other regions and periods Estimatesbased on the water balance were 10plusmn 5 mm yrminus1 decademinus1

in several large catchments across the conterminous US dur-ing 1950ndash2000 (Walter et al 2004) 6plusmn4 mm yrminus1 decademinus1

in catchments in the eastern US during 1901ndash2009 (Krameret al 2015) and 7 mm yrminus1 decademinus1 in 16 catchments onthe Tibetan Plateau during 1966ndash2000 (Zhang et al 2007)The larger increases of E in our study may be related to thelarge increases in air temperature and global radiation in thestudy region over the period considered

42 Drivers of the observed changes in catchmentevaporation

All three drivers investigated ndash changes in atmospheric con-ditions changes in vegetation activity and changes in precip-itation ndash are found to be important for the observed increasein catchment evaporation The drivers are closely interlinked(Fig 10) For example an increase in air temperature notonly contributes directly to changes in E0 and thus E but italso contributes indirectly through increases in the vegetationactivity Attributing changes in E to the drivers can thereforebe done in different ways While in this study vegetationeffects on E include land use changes and indirect effectsvia climate and atmospheric CO2 on vegetation other stud-ies have treated the indirect effects separately Using a globalbiosphere model Piao et al (2007) attributed a global changein E of +03 mm yrminus1 decademinus1 during 1901ndash1999 to cli-mate (+07 mm yrminus1 decademinus1 including indirect effects ofclimate on vegetation atmospheric conditions and precipita-tion) and land use changes (minus08 mm yrminus1 decademinus1 mainlydeforestation) Changes in atmospheric CO2 had a furtherpositive effect of +04 mm yrminus1 decademinus1 through increas-ing the LAI and stomata resistance (Piao et al 2007) Basedon global land surface model simulations Mao et al (2015)found that climate effects (including indirect effects of cli-mate on vegetation atmospheric conditions and precipita-tion) on E were larger than the effects of land use atmo-spheric CO2 and nitrogen deposition during 1982ndash2013with precipitation being the most important climate variableThe latter finding is in line with several other studies (Jung etal 2010 Miralles et al 2014) although Zhang et al (2015)found vegetation change (including indirect effects of cli-mate on vegetation) to be more important

Figure 10 Drivers of changes in evaporation E including feedback effects between them

In this study increases in E driven by changes in at-mospheric conditions have been identified by increasesin E0 and Epan The main driver for the increase in E0in our study is an increase in global radiation of 51plusmn09 W mminus2 decademinus1 on average with a further contribu-tion from an increase in mean air temperature of 045plusmn009 C decademinus1 Even though the scatter is large the datasuggest that the predominantly decreasing wind speeds havelikely reduced the increase in E0 While attribution stud-ies summarized in McVicar et al (2012 Table 7) foundtrends in wind speed to be the most frequent dominantdriver for changes in E0 and Epan (followed by air tem-perature) trends in wind speed play a minor role in ourstudy region Teuling et al (2009) and Wang et al (2010)reported a strong influence of global radiation on changesin E in Europe Global radiation in Europe generally de-creased during the 1960sndash1980s (ldquoglobal dimmingrdquo) and in-creased from the 1980s (ldquoglobal brighteningrdquo) (Norris andWild 2007 Wild 2009 Sanchez-Lorenzo et al 2015)The magnitude of the increase in the present study (51plusmn09 W mminus2 decademinus1) is slightly higher than those reportedin other studies (20plusmn 12 W mminus2 decademinus1 in Central Eu-rope during 1971ndash2012 found by Sanchez-Lorenzo et al2015 and 37 W mminus2 decademinus1 in Austria and Switzerlandduring 1985ndash2005 found by Wild et al 2009) The increasesin global radiation over Europe since the mid-1980s havemainly been attributed to the reduction in aerosols (Norrisand Wild 2007) The strong sensitivity of E0 to global radi-ation suggests that projected air temperature increases do notnecessarily imply increased evaporation in the future

The observed increase in Epan in our study is consistentwith other European studies In Greece non-significant in-creases in Epan over 1983ndash1999 were observed in a pooledseries from 14 stations (Papaioannou et al 2011) Three

out of eight sites in Ireland showed a significant increasein Epan over 1963ndash2005 and one site showed a significantdecrease (Stanhill and Moumlller 2008) In England significantincreases in Epan were observed at two stations over 1957ndash2004 and 1986ndash2010 (Stanhill and Moumlller 2008 Clark2013) and in the Czech Republic AprilndashJune Epan signif-icantly increased at three out of five stations during 1968ndash2010 (Trnka et al 2015)

The NDVI data show a marked increase in vegetation ac-tivity in Austria similar to many other studies in the North-ern Hemisphere (Myneni et al 1997 Slayback et al 2003Liu et al 2015) Generally increases in air temperature andprecipitation and CO2 fertilization have been identified asdrivers (Piao et al 2006 Los 2013) In the study area trendsin precipitation and trends in the NDVI show no significantrelationship (r =minus001) suggesting that increases in pre-cipitation were not important for the changes in vegetationactivity We found strong increases in the NDVI in springand autumn indicating a lengthening of the active growingseason which has been noted by Myneni et al (1997) forexample Increases in the NDVI may be further enhanced byland cover changes In Austria the forest area increased from43 to 47 over 1977ndash2010 at the expense of cropland andextensive grassland (Krausmann et al 2003 Gingrich et al2015) as agricultural land in remote areas with low produc-tivity was abandoned due to economic pressure (Tasser andTappeiner 2002 Rutherford et al 2008)

According to our study the effect of increased vegetationactivity on E0v is of a similar magnitude to the effect ofchanges in the atmospheric conditions A strong influenceof changes in vegetation activity on E was also suggestedfor the eastern US based on correlations between the NDVIand Ewb (Kramer et al 2015) In two forested catchmentsin the southern Appalachians the indirect effects of climateon vegetation dynamics were found to be much more im-portant drivers for long-term increases in E (plus potentialstorage changes) than direct climate impacts (Hwang et al2018) Using simulations from a global coupled biospherendashatmosphere model Bounoua et al (2000) found an E in-crease of 43 mm yrminus1 for an NDVI increase of 008 whichis slightly higher than the average NDVI increase observedin this study of 006 Our study however does not accountfor the effect of increasing atmospheric CO2 concentrationson increasing stomata resistance which means that the effectof vegetation might be overestimated Stomata closure dueto increased atmospheric CO2 may have reduced global Esince 1960 by 16 to 20 mm yrminus1 decademinus1 (Gedney et al2006 Piao et al 2007)

Even though the study area can generally be classifiedas humid we found a strong influence of changes in pre-cipitation on changes in E E is generally energy-limitedin the study region (Sect 211) However as indicated byEwbEmax ratios smaller than unity E may frequently belimited by available moisture particularly interception andEfrom the soil and non-vegetated areas Thus an increase inprecipitation can lead to an increase in E One reason forthe strong sensitivity of changes in E to the increases in an-nual precipitation is the seasonality of the observed changesin precipitation Increases in precipitation were concentratedin the summer season and changes in summer precipitationare expected to contribute more strongly to changes in E asthis is the period when E is highest whereas changes in win-ter precipitation more likely result in changes in discharge Itshould be noted that the strength of the relationship betweenchanges in E and changes in precipitation would be overesti-mated if there were artifacts in the trends of the precipitationdata due to inconsistent record lengths for example In thisstudy consistent record lengths were used Several studiesaround the world in particular in the tropics identified pre-cipitation as an important factor for changes inE (Jung et al2010 Miralles et al 2014 Mao et al 2015)

5 Conclusions and implications

Over the past four decades (1977ndash2014) catchment evap-oration increased on average over 156 study catchments inAustria by 29plusmn 14 mm yrminus1 decademinus1 This increase was at-tributed to changes in atmospheric demand and available en-ergy (suggested to account for 125plusmn42 mm yrminus1 decademinus1)changes in vegetation (98plusmn 40 mm yrminus1 decademinus1) andchanges in precipitation (69plusmn 16 mm yrminus1 decademinus1)

E0 increased on average over all study catchments by18plusmn 5 mm yrminus1 decademinus1 The increase in E0 was largelydriven by the increase in global radiation with further contri-butions from increasing air temperature Rising atmosphericdemand and energy available for E was also revealed by in-creases in the available Epan data Satellite-derived NDVIdata for 1982ndash2014 indicate an increase in vegetation activ-ity This increase may have led to a similar increase of E0vas the increase due to the climate variables A positive corre-lation between increases in E and increases in precipitationfurthermore points to increases in water availability as a thirddriver for the increases in Ewb

Over the study period trends in annual discharge wereclose to zero and increases in E were balanced by increasesin precipitation If the increase in precipitation had beenlower and the increase in E had been similar as in this studynotable reductions in discharge would have been likely Alower increase in precipitation would likely have reduced theincrease in E as 24plusmn 5 of the E increase was directly at-tributed to the increase in precipitation

Estimates of future changes in E of climate impact assess-ments are often based on predicted air temperature and pre-cipitation changes This study clearly shows that despite thelarge air temperature increase in the recent decades globalradiation was much more important for changes in E0 thanair temperature Potential future changes in global radiationdue to changes in cloud cover or air pollution for exampleshould therefore explicitly be accounted for in climate im-pact studies on E Furthermore hydrologic models used insuch studies should consider the effects of possible changesin vegetation on E which result from a longer growing pe-riod for example

Data availability The discharge data can be accessed throughhttpsehydgvat (last access 26 September 2018) The mete-orological data from ZAMG are currently not freely availablerequests should be directed to klimazamgacat The Corineland cover map can be downloaded from httpswwweeaeuropaeudata-and-mapsdataclc-2000-vector-6 (last access 26 Septem-ber 2018) The SRTM DEM can be obtained from httpsrtmcsicgiarorg (last access 26 September 2018)

The Supplement related to this article is availableonline at httpsdoiorg105194hess-22-5143-2018-supplement

Author contributions DD conceived and designed the study per-formed the analyses and prepared the manuscript GB contributedto the study design interpretation of the results and writing of themanuscript

Competing interests The authors declare that they have no conflictof interest

Acknowledgements We gratefully acknowledge the financialsupport from the DFG (German Research Foundation) througha research scholarship to Doris Duethmann (DU 15951-1) Wewould like to thank the Austrian Hydrographic Services and theZAMG for providing the hydrographic and meteorological dataand Juraj Parajka for useful suggestions on processing the dataWe would also like to thank Axel Thomas Ryan Teuling and twoanonymous referees whose comments improved the quality of themanuscript

Edited by Harrie-Jan Hendricks FranssenReviewed by Ryan Teuling and two anonymous referees

References

Allen R G Pereira L S Raes D and Smith M Crop evap-otranspiration ndash Guidelines for computing crop water require-ments FAO Irrigation and drainage paper 56 FAO Rome 300D05109 1998

ATV-DVWK Verdunstung in Bezug zu Landnutzung Bewuchs undBoden GFA-Ges zur Foumlrderung d Abwassertechnik eV Hen-nef Germany 2001

Auer I Boumlhm R Jurkovic A Lipa W Orlik A PotzmannR Schoumlner W Ungersbock M Matulla C Briffa K JonesP Efthymiadis D Brunetti M Nanni T Maugeri M Mer-calli L Mestre O Moisselin J M Begert M Muller-Westermeier G Kveton V Bochnicek O Stastny P LapinM Szalai S Szentimrey T Cegnar T Dolinar M Gajic-Capka M Zaninovic K Majstorovic Z and Nieplova EHISTALP ndash historical instrumental climatological surface timeseries of the Greater Alpine Region Int J Climatol 27 17ndash46httpsdoiorg101002joc1377 2007

Blaschke A Merz R Parajka J Salinas J and Bloumlschl GAuswirkungen des Klimawandels auf das Wasserdargebot vonGrund- und Oberflaumlchenwasser Oumlsterreichische Wasser-und Ab-fallwirtschaft 63 31ndash41 2011

Boumlhm R Heisse Luft Reizwort Klimawandel Fakten AumlngsteGeschaumlfte Va Bene Wien Klosterneuburg 2008

Bounoua L Collatz G J Los S O Sellers P JDazlich D A Tucker C J and Randall D ASensitivity of climate to changes in NDVI J Cli-mate 13 2277ndash2292 httpsdoiorg1011751520-0442(2000)013lt2277soctcigt20co2 2000

Brutsaert W Use of pan evaporation to estimate terrestrial evap-oration trends The case of the Tibetan Plateau Water ResourRes 49 3054ndash3058 httpsdoiorg101002wrcr20247 2013

Brutsaert W and Parlange M B Hydrologic cycle ex-plains the evaporation paradox Nature 396 p 30httpsdoiorg10103823845 1998

Budyko M I Climate and Life Academic Press New York 1974Clark C Measurements of actual and pan evaporation in the upper

Brue catchment UK the first 25 years Weather 68 200ndash2082013

Dong B and Dai A G The uncertainties and causes of the re-cent changes in global evapotranspiration from 1982 to 2010Clim Dynam 49 279ndash296 httpsdoiorg101007s00382-016-3342-x 2017

Douville H Ribes A Decharme B Alkama R and SheffieldJ Anthropogenic influence on multidecadal changes in recon-structed global evapotranspiration Nat Clim Change 3 59ndash62httpsdoiorg101038nclimate1632 2013

European Environment Agency Corine Land Cover 2000 seamlessvector data Version 185 Kopenhagen Denmark 2016

Galbraith D Levy P E Sitch S Huntingford C Cox PWilliams M and Meir P Multiple mechanisms of Ama-zonian forest biomass losses in three dynamic global vegeta-tion models under climate change New Phytol 187 647ndash665httpsdoiorg101111j1469-8137201003350x 2010

Gedney N Cox P M Betts R A Boucher O HuntingfordC and Stott P A Detection of a direct carbon dioxide ef-fect in continental river runoff records Nature 439 835ndash838httpsdoiorg101038nature04504 2006

Gingrich S Niedertscheider M Kastner T Haberl H CosorG Krausmann F Kuemmerle T Muumlller D Reith-Musel Aand Jepsen M R Exploring long-term trends in land use changeand aboveground human appropriation of net primary productionin nine European countries Land Use Policy 47 426ndash438 2015

Hiebl J and Frei C Daily temperature grids for Austria since 1961ndash concept creation and applicability Theor Appl Clima-tol 124 161ndash178 httpsdoiorg101007s00704-015-1411-42016

Hiebl J and Frei C Daily precipitation grids for Austria since1961 ndash development and evaluation of a spatial dataset for hydro-climatic monitoring and modelling Theor Appl Climatol 132327ndash345 httpsdoiorg101007s00704-017-2093-x 2018

Huntington T G Evidence for intensification of the globalwater cycle Review and synthesis J Hydrol 319 83ndash95httpsdoiorg101016jjhydrol200507003 2006

Hwang T Martin K L Vose J M Wear D Miles B Kim Yand Band L E Nonstationary hydrologic behavior in forestedwatersheds is mediated by climate-induced changes in growingseason length and subsequent vegetation growth Water ResourRes 54 1ndash17 httpsdoiorg1010292017WR022279 2018

Hydrographischen Zentralbuumlro Anleitung zur Bearbeitung von me-teorologischen Parametern zur Erfassung des Wasserkreislaufesim Rahmen des Hydrographischen Dienstes in Oumlsterreich Vi-enna 1996

Jarvis A Reuter H I Nelson A and Guevara E Holefilledseamless SRTM data 4th Edn International Centre for TropicalAgriculture (CIAT) Cali Colombia 2008

Jung M Reichstein M Ciais P Seneviratne S I SheffieldJ Goulden M L Bonan G Cescatti A Chen J Qde Jeu R Dolman A J Eugster W Gerten D GianelleD Gobron N Heinke J Kimball J Law B E Mon-tagnani L Mu Q Z Mueller B Oleson K Papale DRichardson A D Roupsard O Running S Tomelleri EViovy N Weber U Williams C Wood E Zaehle S andZhang K Recent decline in the global land evapotranspira-tion trend due to limited moisture supply Nature 467 951ndash954httpsdoiorg101038nature09396 2010

Kendall M G Rank correlation methods 4th Edn Charles Grif-fin London 196 pp 1975

Kramer R J Bounoua L Zhang P Wolfe R E Huntington TG Imhoff M L Thome K and Noyce G L Evapotranspira-tion trends over the eastern United States during the 20th centuryHydrology 2 93ndash111 2015

Krausmann F Haberl H Schulz N B Erb K-H Darge E andGaube V Land-use change and socio-economic metabolism inAustria ndash Part I driving forces of land-use change 1950ndash1995Land Use Policy 20 1ndash20 2003

Liu W B Wang L Zhou J Li Y Z Sun F B Fu G B Li XP and Sang Y F A worldwide evaluation of basin-scale evapo-transpiration estimates against the water balance method J Hy-drol 538 82ndash95 httpsdoiorg101016jjhydrol2016040062016

Liu Y Li Y Li S C and Motesharrei S Spatial and tem-poral patterns of global NDVI trends Correlations with cli-mate and human factors Remote Sensing 7 13233ndash13250httpsdoiorg103390rs71013233 2015

Los S O Analysis of trends in fused AVHRR and MODIS NDVIdata for 1982ndash2006 Indication for a CO2 fertilization effectin global vegetation Global Biogeochem Cy 27 318ndash330httpsdoiorg101002gbc20027 2013

Mann H Non-parametric test against trend Econometrica 13245ndash259 1945

Mao J F Fu W T Shi X Y Ricciuto D M Fisher J BDickinson R E Wei Y X Shem W Piao S L Wang KC Schwalm C R Tian H Q Mu M Q Arain A CiaisP Cook R Dai Y J Hayes D Hoffman F M Huang MY Huang S Huntzinger D N Ito A Jain A King A WLei H M Lu C Q Michalak A M Parazoo N Peng C HPeng S S Poulter B Schaefer K Jafarov E Thornton PE Wang W L Zeng N Zeng Z Z Zhao F Zhu Q A andZhu Z C Disentangling climatic and anthropogenic controls onglobal terrestrial evapotranspiration trends Environ Res Lett10 1ndash13 httpsdoiorg1010881748-9326109094008 2015

Matulla C Regional seasonal and predictor-optimized down-scaling to provide groups of local scale scenarios in the com-plex structured terrain of Austria Meteorol Z 14 31ndash45httpsdoiorg1011270941-294820050014-0031 2005

Matulla C Penlap E K Haas P and Formayer H Compara-tive analysis of spatial and seasonal variability Austrian precip-itation during the 20th century Int J Climatol 23 1577ndash1588httpsdoiorg101002joc960 2003

McVicar T R Roderick M L Donohue R J Li L TVan Niel T G Thomas A Grieser J Jhajharia DHimri Y Mahowald N M Mescherskaya A V KrugerA C Rehman S and Dinpashoh Y Global review andsynthesis of trends in observed terrestrial near-surface windspeeds Implications for evaporation J Hydrol 416 182ndash205httpsdoiorg101016jjhydrol201110024 2012

Milly P C D and Dunne K A Trends in evaporation and surfacecooling in the Mississippi River basin Geophys Res Lett 281219ndash1222 httpsdoiorg1010292000gl012321 2001

Miralles D G van den Berg M J Gash J H ParinussaR M de Jeu R A M Beck H E Holmes T R HJimenez C Verhoest N E C Dorigo W A Teuling A Jand Dolman A J El Nino-La Nina cycle and recent trendsin continental evaporation Nat Clim Change 4 122ndash126httpsdoiorg101038nclimate2068 2014

Myneni R B Keeling C D Tucker C J Asrar Gand Nemani R R Increased plant growth in the north-ern high latitudes from 1981 to 1991 Nature 386 698ndash702httpsdoiorg101038386698a0 1997

Neunteufel R Schmidt B-J and Perfler R Ressourcenverfuumlg-barkeit und Bedarfsplanung auf Basis geaumlnderter Rahmenbedin-gungen Oumlsterreichische Wasser- und Abfallwirtschaft 69 214ndash224 2017

Neuwirth F Die Bestimmung der Verdunstung freier Wasser-flaumlchen aus laumlngerfristigen Mittelwerten Arch Met GeophBiokl Ser B 25 337ndash344 1978

Norris J R and Wild M Trends in aerosol radiative effects overEurope inferred from observed cloud cover solar ldquodimmingrdquoand solar ldquobrighteningrdquo J Geophys Res-Atmos 112 D08214httpsdoiorg1010292006jd007794 2007

Papaioannou G Kitsara G and Athanasatos S Impact ofglobal dimming and brightening on reference evapotranspi-ration in Greece J Geophys Res-Atmos 116 D09107httpsdoiorg1010292010jd015525 2011

Parajka J Merz R and Bloumlschl G Regionale Wasserbilanzkom-ponenten fuumlr Oumlsterreich auf Tagesbasis Oumlsterreichische Wasser-und Abfallwirtschaft 57 43ndash56 2005

Peterson T C Golubev V S and Groisman P YEvaporation losing its strength Nature 377 687ndash688httpsdoiorg101038377687b0 1995

Piao S L Friedlingstein P Ciais P Zhou L M and Chen AP Effect of climate and CO2 changes on the greening of theNorthern Hemisphere over the past two decades Geophys ResLett 33 L23402 httpsdoiorg1010292006gl028205 2006

Piao S L Friedlingstein P Ciais P de Noblet-Ducoudre NLabat D and Zaehle S Changes in climate and land usehave a larger direct impact than rising CO2 on global riverrunoff trends P Natl Acad Sci USA 104 15242ndash15247httpsdoiorg101073pnas0707213104 2007

Richter D Ergebnisse methodischer Untersuchungen zurKorrektur des systematischen Messfehlers des Hellmann-Niederschlagsmessers Selbstverl des Dt WetterdienstesOffenbach 1995

Roderick M L and Farquhar G D The cause of decreased panevaporation over the past 50 years Science 298 1410ndash14112002

Roderick M L Rotstayn L D Farquhar G D and HobbinsM T On the attribution of changing pan evaporation GeophysRes Lett 34 L17403 httpsdoiorg1010292007gl0311662007

Roderick M L Hobbins M and Farquhar G Pan evaporationtrends and the terrestrial water balance I Principles and obser-vations Geogr Compass 3 746ndash760 2009

Rutherford G N Bebi P Edwards P J and Zimmermann NE Assessing land-use statistics to model land cover change ina mountainous landscape in the European Alps Ecol Model212 460ndash471 httpsdoiorg101016jecolmodel2007100502008

Sanchez-Lorenzo A Wild M Brunetti M Guijarro J AHakuba M Z Calbo J Mystakidis S and Bartok B Re-assessment and update of long-term trends in downward surfaceshortwave radiation over Europe (1939ndash2012) J Geophys Res-Atmos 120 9555ndash9569 httpsdoiorg1010022015jd0233212015

Sellers P J Los S O Tucker C J Justice C O Dazlich DA Collatz G J and Randall D A A revised land surface pa-rameterization (SiB2) for atmospheric GCMs ndash 2 The generationof global fields of terrestrial biophysical parameters from satel-lite data J Climate 9 706ndash737 httpsdoiorg1011751520-0442(1996)009lt0706arlspfgt20co2 1996

Sen P K Estimates of the regression coefficient based onKendallrsquos tau J Am Stat Assoc 63 1379-1389 1968

Slayback D A Pinzon J E Los S O and TuckerC J Northern hemisphere photosynthetic trends 1982ndash99Global Change Biol 9 1ndash15 httpsdoiorg101046j1365-2486200300507x 2003

Spreafico M Weingartner R and Leibundgut C HydrologischerAtlas der Schweiz (Kartenmaterial) Bundesamt fuumlr Landesto-pografie swisstopo Wabern Switzerland 2007

Stanhill Gand Moumlller M Evaporative climate changein the British Isles Int J Climatol 28 1127ndash1137httpsdoiorg101002joc1619 2008

Tasser E and Tappeiner U Impact of land usechanges on mountain vegetation Appl VegSci 5 173ndash184 httpsdoiorg1016581402-2001(2002)005[0173ioluco]20co2 2002

Teuling A J A forest evapotranspiration paradox investi-gated using lysimeter data Vadose Zone J 17 170031httpsdoiorg102136vzj2017010031 2018

Teuling A J Hirschi M Ohmura A Wild M Reichstein MCiais P Buchmann N Ammann C Montagnani L Richard-son A D Wohlfahrt G and Seneviratne S I A regionalperspective on trends in continental evaporation Geophys ResLett 36 L02404 httpsdoiorg1010292008gl036584 2009

Trnka M Brazdil R Balek J Semeradova D HlavinkaP Mozny M Stepanek P Dobrovolny P Zahradnicek PDubrovsky M Eitzinger J Fuchs B Svoboda M HayesM and Zalud Z Drivers of soil drying in the Czech Repub-lic between 1961 and 2012 Int J Climatol 35 2664ndash2675httpsdoiorg101002joc4167 2015

Tucker C J Pinzon J E Brown M E Slayback D A Pak EW Mahoney R Vermote E F and El Saleous N An ex-tended AVHRR 8-km NDVI dataset compatible with MODISand SPOT vegetation NDVI data Int J Remote Sens 26 4485ndash4498 httpsdoiorg10108001431160500168686 2005

Ukkola A M and Prentice I C A worldwide analysis of trendsin water-balance evapotranspiration Hydrol Earth Syst Sci 174177ndash4187 httpsdoiorg105194hess-17-4177-2013 2013

Viglione A Parajka J Rogger M Salinas J L LaahaG Sivapalan M and Bloumlschl G Comparative assessmentof predictions in ungauged basins ndash Part 3 Runoff signa-tures in Austria Hydrol Earth Syst Sci 17 2263ndash2279httpsdoiorg105194hess-17-2263-2013 2013

Walter M T Wilks D S Parlange J Y andSchneider R L Increasing evapotranspirationfrom the conterminous United States J Hydrom-eteorol 5 405ndash408 httpsdoiorg1011751525-7541(2004)005lt0405ieftcugt20co2 2004

Wang K C and Dickinson R E A review of global ter-restrial evapotranspiration observation modeling climatol-ogy and climatic variability Rev Geophys 50 RG2005httpsdoiorg1010292011rg000373 2012

Wang K C Dickinson R E Wild M and Liang S L Evidencefor decadal variation in global terrestrial evapotranspiration be-tween 1982 and 2002 2 Results J Geophys Res-Atmos 115D20113 httpsdoiorg1010292010jd013847 2010

Wild M Global dimming and brightening A re-view J Geophys Res-Atmos 114 D00D16httpsdoiorg1010292008jd011470 2009

Wild M Truessel B Ohmura A Long C N Konig-LangloG Dutton E G and Tsvetkov A Global dimming and bright-ening An update beyond 2000 J Geophys Res-Atmos 114D00D13 httpsdoiorg1010292008jd011382 2009

Yue S Pilon P Phinney B and Cavadias G The influence ofautocorrelation on the ability to detect trend in hydrological se-ries Hydrol Process 16 1807ndash1829 2002

ZAMG Jahrbuumlcher der Zentralanstalt fuumlr Meteorologie und Geo-dynamik Wien 1977ndash1990

Zhang K Kimball J S Nemani R R and Running S WA continuous satellite-derived global record of land surfaceevapotranspiration from 1983 to 2006 Water Resour Res 46W09522 httpsdoiorg1010292009wr008800 2010

Zhang K Kimball J S Nemani R R Running SW Hong Y Gourley J J and Yu Z B Vegetationgreening and climate change promote multidecadal rises ofglobal land evapotranspiration Scient Rep-UK 5 15956httpsdoiorg101038srep15956 2015

Zhang Y Q Liu C M Tang Y H and Yang Y H Trends in panevaporation and reference and actual evapotranspiration acrossthe Tibetan Plateau J Geophys Res-Atmos 112 D12110httpsdoiorg1010292006jd008161 2007

Zhang Y Q Leuning R Chiew F H S Wang E LZhang L Liu C M Sun F B Peel M C Shen YJ and Jung M Decadal trends in evaporation from globalenergy and water balances J Hydrometeorol 13 379ndash391httpsdoiorg101175jhm-d-11-0121 2012

Abstract

Introduction

Data and methods

Water-balance data and water-balance estimates

Discharge data and catchment attributes

Catchment average meteorological data

Estimating evaporation from the water balance (Ewb)

Pan evaporation data

Estimation of reference evaporation

Reference evaporation

Effect of changes in vegetation activity on reference evaporation

Trend analyses regression analyses and attribution of the trend in Ewb

Results

Changes in evaporation estimated from the water balance (trend detection)

Drivers of the increases in evaporation (attribution)

Changes in atmospheric conditions -- reference evaporation

Changes in atmospheric conditions -- pan evaporation

Changes in vegetation activity

Changes in precipitation

Synthesis of attribution

Discussion

Changes in catchment evaporation in Austria over the past four decades

Drivers of the observed changes in catchment evaporation

Conclusions and implications

Data availability

Author contributions

Competing interests

Acknowledgements

References

Rietholzbach Switzerland over 1976ndash2007 showed a de-creasing trend in E in the first half of the period and an in-creasing trend in the second half (Teuling et al 2009) Ob-servations based on eddy covariance are usually too short fortrend analyses (Wang and Dickinson 2012) but they havebeen used to train models that use satellite and climate data atlonger timescales and larger space scales (Jung et al 2010Wang et al 2010 Zhang et al 2010) While these studiesgenerally agree on positive trends since the 1980s they dif-fer in the magnitude of the estimated trends (Dong and Dai2017) It has been noted however that the results of suchmodels need to be treated with care as they are sometimesinconsistent with trends from the water balance particularlyfor wet basins (Zhang et al 2012 Liu et al 2016)

Potential drivers for changes in E are changes in avail-able energy and atmospheric evaporative demand which isdriven by variations in wind vapor deficit and air tem-perature Available energy and the evaporative demand ofthe atmosphere can be estimated based on climatic driversor measured with evaporation pans In many parts of theworld (including North America China India) annual panevaporation (Epan) has decreased in the second half of the20th century with rates of 10ndash40 mm yrminus1 decademinus1 despiteincreases in air temperature (Peterson et al 1995 Roder-ick et al 2009 McVicar et al 2012) In some instancesthis decrease in Epan has been explained by decreases in netradiation andor wind speed (Roderick and Farquhar 2002Roderick et al 2007) In other instances decreasing Epanhas been interpreted as a consequence of increasing actualevaporation (Brutsaert and Parlange 1998 Brutsaert 2013)In Europe most studies found increasing trends of Epaneg in Ireland (1963ndash2005) (Stanhill and Moumlller 2008)England (1957ndash2004 and 1986ndash2010) (Stanhill and Moumlller2008 Clark 2013) Greece (1983ndash1999) (Papaioannou et al2011) and the Czech Republic (1968ndash2010) (Trnka et al2015) In this paper we use the term atmospheric conditionsto summarize the drivers available energy and atmosphericdemand

Other potential drivers are changes in land cover and vege-tation (Piao et al 2007) Rising atmospheric CO2 concentra-tions increase plant growth (Piao et al 2007) and influencestomata closure (Gedney et al 2006) Finally changes in wa-ter availability resulting from changes in precipitation maycontribute to changing E For example Jung et al (2010)attributed the hiatus of the increasing trend in global terres-trial E during 1998ndash2008 to the limited moisture supply inthe Southern Hemisphere

Several global scale studies attributed modeled changesin E to their drivers The land surface models of Douvilleet al (2013) could only explain variations in E over 1950ndash2005 if natural forcings enhanced greenhouse gas concen-trations and aerosols were considered Based on an ensem-ble of land surface models that considered variations in cli-mate land cover atmospheric CO2 concentration and ni-trogen deposition Mao et al (2015) found that E trends

over 1982ndash2013 were dominantly driven by variations inclimate in particular precipitation Miralles et al (2014)showed that variations in E in the tropics were strongly in-fluenced by variations in precipitation driven by El NintildeondashSouthern Oscillations In contrast using a modified PenmanndashMonteith equation with detrended input variables Zhang etal (2015) concluded that the increase in global terrestrial Eover 1982ndash2013 was largely driven by vegetation greeningwhile changes in global radiation wind speed air vapor pres-sure air temperature and atmospheric CO2 concentrationhad only minor effects A full consensus regarding the driversof the increasing E that has been observed does not seemto exist The cited studies above looked at changes in mod-eledE at the global scale based on globally available meteo-rological data Complementary to these global studies thereis a need for data-based studies focusing on smaller regionswith high-quality data

The aim of this study is to (a) identify changes in catch-ment evaporation in the past 40 years and (b) identify thedrivers of these changes We use high-quality data sets ofdischarge precipitation and other climate variables from156 catchments in Austria during the period 1977ndash2014We analyze regional averages over these catchments in or-der to increase the robustness of the analysis The potentialdrivers of changes in catchment evaporation examined are (i)the atmospheric conditions quantified by reference evapora-tion (E0) and Epan (ii) vegetation quantified by a satellite-based vegetation index and (iii) precipitation as a proxy forthe available water

2 Data and methods

21 Water-balance data and water-balance estimates

211 Discharge data and catchment attributes

For the analysis of changes in the water-balance-based evap-oration (Ewb) we identified all catchments in Austria wheredaily discharge data in the period 1977ndash2014 (hydrolog-ical years November to October) with a maximum of 2years missing were available The beginning of the analy-sis period was set to 1977 because most discharge series inAustria start in the mid-1970s Catchments with substantialanthropogenic influences from dams or water withdrawals(Viglione et al 2013) catchments containing glaciers anda few high-mountain catchments where observed dischargeexceeded observed precipitation were excluded This selec-tion resulted in a total of 156 catchments (Fig 1) ranging insize from 23 to 6214 km2 (average 316 km2) Land cover wasderived from the Corine 2000 data (European EnvironmentAgency 2016) The land cover is largely dominated by forestand grassland (Table 1) Median catchment elevations werecalculated from the SRTM digital elevation model (Jarvis etal 2008) They range from 287 to 1920 m (average 910 m)

dSsoil

middot (esminus ea)

1+ γ middot(

1+ rsra

τE0(k) (5)

FPAR=(Sminus Smin)

rs(NDVI)=(

)minus1

ϕatmE0v =1

nsumk=1

2suml=1

τE0c(k l)

τE0v(k l)

ϕvegE0v =1

nsumk=1

2suml=1

τE0v(k l) (10)

)middotϕatmE0v

3 Results

4 Discussion

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

dSsoil

middot (esminus ea)

1+ γ middot(

1+ rsra

τE0(k) (5)

FPAR=(Sminus Smin)

rs(NDVI)=(

)minus1

ϕatmE0v =1

nsumk=1

2suml=1

τE0c(k l)

τE0v(k l)

ϕvegE0v =1

nsumk=1

2suml=1

τE0v(k l) (10)

)middotϕatmE0v

3 Results

4 Discussion

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

dSsoil

middot (esminus ea)

1+ γ middot(

1+ rsra

τE0(k) (5)

FPAR=(Sminus Smin)

rs(NDVI)=(

)minus1

ϕatmE0v =1

nsumk=1

2suml=1

τE0c(k l)

τE0v(k l)

ϕvegE0v =1

nsumk=1

2suml=1

τE0v(k l) (10)

)middotϕatmE0v

3 Results

4 Discussion

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

τE0(k) (5)

FPAR=(Sminus Smin)

rs(NDVI)=(

)minus1

ϕatmE0v =1

nsumk=1

2suml=1

τE0c(k l)

τE0v(k l)

ϕvegE0v =1

nsumk=1

2suml=1

τE0v(k l) (10)

)middotϕatmE0v

3 Results

4 Discussion

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

)middotϕatmE0v

3 Results

4 Discussion

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

4 Discussion

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

4 Discussion

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

4 Discussion

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

4 Discussion

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

Abstract

Introduction

Data and methods










Results








Discussion




Data availability


Competing interests

Acknowledgements

References

Documents

Why has catchment evaporation increased in the past 40