WATER RESOURCES RESEARCH, VOL. 27, NO. 7, PAGES 1657-1663, JULY 1991
A Daily Hydroclimatological
Data Set for the ContinentalUnited States
JAMES R. WALLIS
IBM Thomas J. Watson Research Laboratory, Yorktown Heights, New York
DENNIS
P. LETTENMAIER
Departmentof Civil Engineering,Universityof Washington,Seattle
ERic
F. WOOD
Department
qf CivilEngineering
andOperations
Research,
Princeton
University,
Princeton,
NewJersey
Previousattemptsto validategeneralcirculation
modelsimulations
of landsurfacehydrologyhave
oftenbeenlimitedby theabsence
of systematic
historical
data,especially
forrunoff,precipitation,
and
temperature.
Because
hydrological
response
timesfor unregulated
watersheds
in theUnitedStates
varyfroma fewhourstoa fewdaysat most,climatological
studies
dealing
withlandsurface
hydrology
requiredataat relativelyshorttimeintervals.
We describe
a setof 1009U.S. Geological
Survey
streamflowstations,and 1036National Oceanicand AtmosphericAdministration
climatological
stations,for whichlong-term(1948-1988)
observations
havebeenassembled
intoa consistent
daily
databasewithmissing
observations
estimated
usinga simpleclosest-station
prorating
rule.Carewas
taken in selectionof the streamflowstationsto assurethat the recordswere essentiallyfree from
regulation.
Theclimatological
stations
area subset
of thehistorical
climatology
network
forwhich
monthlydataare described
by Quinlanet al. (1987).The dataformatis provided
to facilitate
development
of alternative
dataretrievalalgorithms.
Estimated
values
formissing
data,aswellas
suspicious
observations,
areflagged.
Thedataareretrievable
bystation
list,state,latitude-longitude
range,
andhydrologic
unitcodefromcompact
digital
read-only
memory
(CDROM).CD-ROMcopies
are available from the second author.
usingselectedhistoricalobservations
with corresponding
GCM seasonal
predictions
for two transects
(latitudebands)
INTRODUCTION
Predictions
of long-termclimatechange,especiallythose
thathaveresultedfrom numericalgeneralcirculationmodels
(GCMs),havebeenthe recentsubjectof considerablescientific
interestandcontroversy.GCMs are basedon descriptionsof massand energy balancesin the atmosphere,as
described
by differentialequations,which are solvedusing
gridmeshes
withspatialdimensions
of the orderof hundreds
of kilometers(see, for example, Mitchell [1989]). GCMs
characterize
land surfaceprocesses,includingevaporative
fluxes
andrunoff,usinghighly simplifiedland surfacerepresentations.
To date, most attemptsto validate GCMs have
focused
onlong-termsimulations
of averageclimaticconditions
(see,for example,Delworthand Manabe [1989]).The
fewattempts
to validateGCM simulationsof presentsurface
meteorological
conditionsusing historical surfacemeteorological
data[e.g., Rind, 1989;Grotch, 1988]have focused
primarily
on long-termclimaticaveragesover large areas.
Theyhavebeensomewhatlimitedby the absenceof systematichistoricaldata and methods for spatially averaging
across the continental United States. These comparisons
showedsomesystematicproblemswith the GCM simulationsof runoffandprecipitation.
However,the analysiswas
limitedto long-term
average
conditions
dueto theeffectsof
reservoirregulationon the streamflowrecordsused.Further, the comparisons
wereof necessityqualitative,since
the comparisons
werebetweenGCM grid cell simulations
and observedpoint(precipitation)
and watershed(runoff
historical data.
In additionto attemptsto validateGCMs, a numberof
studieshave analyzedhistoricalclimatic data (primarily
surface
temperature)
in attempts
to identifytheexistence
of
climatechangesignals[e.g.,Diaz and Quayle,1980;Karl,
1985;Wigleyet al., 1985;Joneset al., 1986].Because
of the
considerable
interestin long-termtemperaturetrends,and
becauseof a numberof problemswith the analysisof
observationalrecordsdue to station moves, changesin
instrumentation,
and otherproblems,Quinlanet al. [1987]
identified
approximately
1200long-term
monthlyprecipitation
and
temperature
records
for
the
continental
United
tions.
States,whichtheytermedthehistorical
climatology
network
Attemptsto validate GCM hydrologicpredictionshave (HCN). For eachof the HCN stations,
corrections
for
been
evenmorelimited,in partdueto problems
of identify- systematic
biases
due
to
changes
in
location,
instrumentaingsuitable
historicalrecordsof runoffthat are free of water
tion changes,
andchanges
in observation
timeswerepromanagement
effects. Neilson et al. [1989] compared the
vided.Ourdailyprecipitation/temperature
sitesarea subset
observationaldata that are consistent with GCM formula-
average
seasonaldistributionof precipitationand runoff of the HCN sites.
Identificationof streamflowrecordssuitablefor climate
Copyright
1991by theAmerican
Geophysical
Union.
Paper
number
91WR00977.
0043-1397/91/91
WR-00977
$05.00
studies
presents
a somewhat
different
setof problems
than
existsfor theprecipitation
andtemperature
records.Almost
1657
WALLIS
ETAL.:DAILYHYDROCLIMATOLOGICAL
DATASET
1658
all gauged streams in the continentalUnited States are
affectedto someextentby man'sactivities,themostsignificantof whichare upstream
reservoirs,
upstream
diversions
50-
for beneficialuseswithin the basin, and, especiallyin the
West, interbasintransfersof water. The U.S. Geological
Survey(USGS)hasrecognized
the needto providebaseline
data on natural runoff for catchments that are free of water
40-
managementeffects. The USGS Benchmarknetwork was
establishedin part to provide the basisfor long-termassess-
mentsof hydrologicchange[Cobb and Biesecker,1971].
However, the number of stations(lessthan 50) is inadequate
to characterize runoff for the entire United States, especially
30-
sincethe drainageareasof the Benchmarkstationsarequite
small.
20-
Langbeinand Slack [1982] identifiedapproximately200
streamgaugingstationsin the continentalUnited Statesthat
130
120
1• 0
100
90
80
70
could be used to evaluatelong-termstreamflowvariations.
Fig. 1. Geographicdistributionof runoff stations.
They groupedthe stationsinto three classes:thosewith no
reportedregulationor diversions
(ClassI), diversions
and/or
storagecapacityless than 10% of the mean annualrunoff modeling.This dictatedthat the streamsbe unaffected
by
(ClassII), and others(ClassIII). Amongthe difficultieswith
Langbeinand Slack's stationlist is that many of the stations
are no longeractive, and somewere discontinuedas earlyas
1913. A further update of this concept is currently in
progress[Landwehrand Slack, 1990];at thistime the list of
sitesand periods of record is not available. It is likely that
these data, like the original Langbein-Slackrecords,will be
limited
to streamflow.
Unfortunately, stream gauging and climatological networks in the United Stateshave evolved independently;it is
only recently, with the advent of remote data transmission
capabilities,that meteorologicaland streamgaugingsystems
have begunto be coupled,albeit primarily for floodforecasting purposes. Further, although the U.S. Geological Survey's daily values for streamflow are computerized for
virtually all long-term stations for the entire period of
historic record, daily climatological records maintained by
the National Climatic Data Center (NCDC) are only available in electronic form for the period 1948 to present for
most stations. Therefore, to be useful for such purposesas
long-term characterization of land surface processes,attention-needsto be focused on stream gaugingstationsthat have
been active during the post-1948 period. With the increased
attention given to validation of GCMs, large-scalemodeling
efforts will need to be undertaken, for which consistent
records of surface meteorological variables, and runoff, will
diversionsand/or regulation at the storm responsescale;for
practical purposesthis implies daily data. To identify
streams meeting this criterion, we proceeded as follows.
First, we screenedthe USGS daily values file for all stations
that beganoperationprior to water year 1948and hadatleast
40 yearsof record(the screeningoperationswere performed
using commercialsot!ware for compact digital/readonly
memory(CD-ROM) retrieval of the USGS daily valuesfiles).
This resulted in a maximum
candidate
list of about $000
stations.We then examined the comments paragraphofthe
USGS station remarks file tbr each station and classifiedthe
stations as follows: no upstream diversions or regulation
(ClassI), minimal upstreamdiversionsand regulation(Class
II), upstreamdiversionsand regulations, extent/effectsunknown (Class III), substantialupstream diversionsand/or
regulation,stationprobablyunusable(Class IV), andmajor
upstreamdiversionsand/or regulation, almost certainlyunusable (Class V). In addition, stations with substantialnat-
ural upstreamstorage(e.g., lakes) were assignedto a Class
VI. Class IV, V, and VI stations were eliminatedfrom
furtherconsideration;the relatively small numberof Class
III stationswere also eliminated, but a list of thesestations
was retainedfor future reference. The remaining 1413Class
I and I! stations were further screened for stations that were
discontinued
priorto 1978.Thisresultedin a finallistof1009
stations,
the geographic
distributionof whichis shown
in
be essential.Some preliminary effortsalongthese lineshave Figure1. Althoughthereare somestationsin eachstate,
the
been initiated [e.g., Lettenmaier et al., 1990]. However, our
attention in this paper is limited to the description of the
development of a consistent set of daily streamflow and
climatic data for the continental United States during the
41-year period 1948-1988, in the belief that others with
similar data requirements will have an interest in the characteristics of the raw data, the criteria used for station
selection, and the algorithms used for estimatingmissing
values.
stationdensitytendsto be higher in the more humidareas
(e.g., East and Northwest) than in arid and semiaridareas
(e.g., Southwest)where the effects of water management
are
greater.
ESTIMATION OF MISSING DAILY STREAMFLOW VALUES
Each of the 1009 station records was screened, andany
missing
dailyvalueswereidentified
andsubsequently
estimatedasfollows.
Thelong-term
monthlyaverage
flows
for
STATION
SELECTION:
nearest
boththetargetsite(station
withmissing
data)and
theneighboring
sitewerecomputed
for thoseyears
with
STREAMFLOW
As noted above, our major concern was to identify stream
gauging locations as free as possible from the effects of
upstream diversions and storage. One of our concernswas
thensimply
estimated
asthecorresponding
dailyflowatthe
that the stations be suitable for rainfall/snowmelt
monthly
meanat thetargetsiteto themeanat theneighbor'
runoff
observeddata. Missingvaluesat the target stations
were
neighboring
site, multipliedby the ratio of the long-term
WALLIS ET AL.: DAILY HYDROCLIMATOLOGICAL DATA SET
1659
ing
site.
If missing
values
wereencountered
forthesamedays) provided, in
some cases, a greater total than the
unadjusted
monthly
HCN totals. The approach used to
date
atboth
sites,
thenthesecond
orthird(ifneeded)
nearest
identify
and
flag
these
problemsis discussedin more detailin
neighbors
wereused
instead.
Onrareoccasions
whenall
four
sites
hadconcurrent
missing
values,
thefill-invalue
was the next section.Whenever possible, the missingdata were
thelong-term
dailyflowforthatmonth
andsite,whichwas estimatedand adjustedin a manner to yield a monthly total
that is consistent with the unadjusted HCN data set. In
flagged
sothatit could
bereadily
recognized.
addition, we calculatedthe adjustmentfactor to go from the
unadjustedto the adjustedHCN data for each month of
PRECIPITATIONAND TEMPERATURE
every year. This allows for the construction of daily precipTheinitialsetofprecipitation
andtemperature
siteswere itation and maximum and minimum temperature data that
those
identified
byQuinlanet al. [1987],whichconsisted
of are consistentwith the adjustedmonthly HCN values.
1228
temperature
stationsand 1220precipitation
stations. We encounteredadditionalproblemswith the daily climaOne
precipitation
station
forwhichonly4 years
ofdatawere tological data. In some cases the daily observationswere
present
ontheQuinlan
etal.station
listwaseliminated
atthe missingduring monthsthat the HCN data set reported as
outset.
We alsoeliminated
at the outsetthe seventempera- complete.In suchcasesthe missingdayswere estimatedso
turestations
for whichmonthlyprecipitation
wasnotavail- as to be consistent with the HCN monthly totals. In the
able.Wethenretrievedthe dailyprecipitation
andtemper- processof verifyingthe computerizeddailyrecordsfor such
ature
datafor eachof theremaining
stations.We foundthat cases,spot checkingof the originalpaper recordsshowed
daily
datain electronic
formwerenot available
for both that someof the daysshownas missingin the computerized
precipitation
andtemperature
fora number
of stations;
this records were in fact observed. These data were nonetheless
further
reducedthe numberof candidatestations,as did the
treatedasmissing;noattemptwasmadeto editthe raw daily
requirement
thatnomorethan20%ofthedayshavemissing data. The missingvalueswere estimatedusingthe following
values.The final list had 1036 precipitation/temperature three-step procedure.
sites.
1.
The closest three stations were identified, and the
Thehandling
of missingdatafor precipitation
andtemper- missingdayswere estimatedfrom the closeststationwith
aturewas more difficult than for streamflow becauseof two
factors:
first, unlike streamflowrecordsfor which the num-
data. The observationsfrom this station was adjusted by the
ratio of the long-termmeansfor that month(for precipita-
berof missingperiodstend to be 'few, but for which the tion) or the differencein the long-termmeansfor that month
temperature).If noneof the three
lengths
of the gapsmay be long,theretendto be more, (for maximum/minimum
shorter
gapsin precipitation
andtemperature
observations, closest stationshad data, the long-term mean from that
andthesegapstendto be scatteredthroughout
the record. month was used to estimate the missing day.
2. Takingthe resultsfrom step!, the estimatedmonthly
Therefore
whilegapsin the streamflowrecordstend to last
or monthlymean(for temperature)
formonthsto years, gapsin precipitationand temperature total (for precipitation)
to theunadjusted
HCN monthlyvalues,and
recordstend to last for only few days, and in many cases wascompared
onlyoneday.The secondcomplicating
factorlbr precipita- the differencebetweenthesetwo monthlyvaluesis provided
tionandtemperature
was the desirabilityto have the daily as a correction factor. Becausethere are different methods
thesedifferences
(inconsistencies)
withinthe
data,afterestimationof missingvalues, be consistentwith for distributing
differentmethodscanhavea largeeffect
themonthlydata setsof Quinlan et al. [1987]. As noted month,andbecause
circumstances,
we decidednotto adjustthe
above,Quinlanet al. provided two sets of monthly data for particular
records:
unadjusted
or "raw" data, and data adjustedfor data but instead only to indicate the magnitudeof the
systematic
biasesdue to station moves, changesin instrumentation,
andtime of observation.We refer to the two data
setsasthe unadjustedand adjustedHCN data sets, respec-
inconsistencies.
3. Informationrequiredto further modifythe daily data
setfrom step2 to reflectdifferences
betweenthe adjusted
tively.We adjustedthe daily data, throughthe processof HCN monthlyvaluesand the unadjustedHCN monthly
If implemented,
this stepwill resultin
estimating
missingobservations
and otherchecksdescribed valuesis provided.
for the systematic
biasesdue to
below,to be consistent with the unadjusted HCN data. daily valuescorrected
in instrumentation,
asidentified
However,we also provide monthly adjustmentfactors to stationmovesandchanges
produce
dailydataconsistentwith the HCN adjusteddata. It by Quinlaneta!. [1987].In thecaseof bothsteps2 and3,
are not calculated
in caseswherethe HCN
should
be notedthat an updateto the Quinlanet al. monthly adjustments
datahasrecentlybeenreleasedwhichcontainsadjustments monthly data are missing.
forurbantemperature
stations
thoughtto be affectedby the
so-called
"heat island" effect, as describedby Karl et al.
[1988].
The numberof stationsso affectedis relatively small,
INCONSISTENTAND SUSPICIOUSPRECIPITATION
TEMPERATURE VALUES
sowehavenot adjustedthe daily data to be consistentwith
thelatest
HCN heatislandadjustments.
However,sufficient As notedabove, we had difficultymakingthe sum (or
ofthevalues
forthedaysina monthequaltheHCN
information
is providedso that suchadjustments
could average)
monthly
values
inmanycases.
Weexpected
suchdifferences
easilybe made.
Thislatter requirementof makingour daily data set to occur for the adjustedHCN values but did not expect
consistent
with the HCN data sets was more complicated
suchproblems
for the unadjusted
values,particularly
in
dayswerereported.Someof the
thanoriginally
envisioned
becauseof inconsistencies
be- monthswhereno missing
maybe theresultof differences
in timeof day
tween
themonthlyHCN valuesandthe archivaldailyvalues problem
for reporting
monthlyvaluesin theHCN source
forthesame
stations.
Forexample,
themonthly
totalsbased conventions
ontheincomplete
dailyprecipitation
data(thenonmissingdata and the NCDC records.Initially, we intendedto pro-
!660
WALLISETAL.:DAILYHYDROCLIMATOLOGICAL
DATASET
statistics
are provided
of the fractionof missing
datafor
runoffstationsandfor climatological
stationsby variable
50-
(precipitation,
dailyminimum
andmaximum
temperature)
and,inthecaseofprecipitation,
forthepercentage
ofdays
for whichonly accumulated
amountswere reported,
as
describedin the previoussection.All missingdataand
precipitation
accumulation
statisticsare givenfor boththe
entireperiodof record(1948-1988)and the first31 years
(1948-1978).The reasonfor inclusionof the 1948-!978
40-
30-
20....
130
Fig. 2.
I
I
I
I
I
I
120
110
1 O0
90
80
70
Geographicdistributionof climatologicalstations.
vide adjustmentsthat would make the sum of the days and
the HCN monthly values consistent.Becauseof problems
with adjustment of missing values we finally decided to
provide only the monthly differences.
In addition to the problem with sum of the day and
monthly value inconsistencies, we encountered a number of
suspiciousvalues. Thesevalueswere flaggedbut not altered.
Suspiciousvaluesfell into the followingcategories:(!) daily
statistics
for missingdataand precipitationaccumulations
is
that, in the caseof runoffstations,there were a number
of
stations
for whicha disproportionate
amountof missing
data
occurredin the last 10 years (1979-!988) becauseof station
discontinuation.
Generally,Table 1 showsthat missing
data
were more uniformly distributedthroughoutthe periodof
record for the climatological records.
The statisticalanalysis shows that the median catchment
areafor all runoffstationsin our data set is 294 squaremiles
and that when segregatedby region the median area varies
from !03 to 575 squaremiles. The median distancebetween
a streamflow and climatological station for the entire United
Statesis about20 miles, and 99% of the streamgauges
are
withinabout60 milesof a climatologicalstation.Exceptfor
hydrologicunits 17 and 18, the climatological stationstendto
beat a higherelevationthanthe streamflowstations.Finally,
for the entire United States, over 50% of the runoff stations
hadno missingobservations,while 50% of the climatological
(rangprecipitation< 0 or > 20 inches,(2) dailyTruax
< Train,(3) stationshad lessthan about2% missingobservations
rmax - rmin > 100øF,and (4) Truax< 20øFduringthe period ing from 1.62% for precipitation to 2.07% for maximum
May-September.
temperature).
Thirty-eightpercentof the precipitation/temperature
sites
hadoneor morevaluesin the abovecategories,
althoughthe
DATA AVAILABILITY
total numberof such occurrencesin the 41 years of record
The data are available 'from the second author on CDfor most of these sites was one or two. In a few cases,
ROM.
The data are retrievableby stationlist, state,hydrohowever,therewere as manyas 100suchflags.Perhapsof
range. In addition,
greater concern is the number of errors that must exist in the logicunit code, and latitude-longitude
archival data but which are not obviousenoughto fail the soi•wareis providedto identify and retrieve all climatologiscreeningcriteria.At the very least,vigilanceon the partof cal and/orstreamflowstationswithin a given distanceofa
datausersis suggested;
we encouragethoseresponsible
for specificstation.The data format is providedalongwiththe
collectingand archivingof climatological
data to develop retrievalprogramsto facilitate developmentof alternative
and employimprovedquality assuranceproceduressothat retrieval algorithms. For the streamflow data there is a
summaryfile listingthe periodsof missingdata (replaced
by
the number of data errors can be reduced.
estimatedvaluesin the datafiles)for each station.Generally,
thenumberof periodsof missingdatais relativelysmallfor
SUMMARY CHARACTERISTICSAND STATISTICS
streamflow,but the periodstend to be lengthy,typically
a
The characteristicsof the data can be representedin a year or more. For the climatologicaldata there tendto be
variety of manners.For the purposeof validatingclimate more periodswith missingdata, of much shorter duration
modelsit is importantto establishthe representativeness
of (oftenonlya few days).Thereforethe data structureforthe
the observationlocations.Figure 1 givesthe locationof the climatological data is somewhat different than for the
streamflowstationswhile Figure 2 givesthe locationof the
meteorological stations. The station characteristicsare sum-
marizedin Table 1 as empiricalprobabilitydistributions
for
selectedsummary statisticsfor the entire continentalUnited
streamflow
data. Threebasicfiles are providedfor each
station(oneeachfor dailyprecipitation,
maximumtemperature,andminimumtemperature).Thesefiles containcodes
indicatingwhich days were estimatedand how theywere
Statesandby USGShydrologic
units,of whichthereare18, estimated,
aswellasflagsfor suspicious
values,asdescribed
as shown in Figure 3. Because of the small number of above.Additional
filesare providedwith monthlyadjuststationswithin particularhydrologicunit areas,someareas mentfactorsto makethe basicdatafully consistent
with
were combined in our summary statisticstables.
The summary statisticsreported in Table 1 are the drain-
age area in squaremilesfor runoff stations(Area USGS),
station elevation in feet (Elevation USGS and Elevation
eithertheunadjusted
or adjusted
HCN monthlydata.
SUMMARY
NCDC for runoffandclimatological
stations,
respectively), A dailyhydrometeorological
datasetfor the continental
•nd minimumdistancein miles betweena streamflowand
United Stateswas constructedfor use in climatestudies.
meteorological
station(Distance).In addition,summary Thedataweretaken
fromUSGSdailystreamflow
data
and
WALLIS ET AL.' DAILY HYDROCLIMATOLOGICALDATA SET
1661
TABLE 1. StatisticalSummaryof Streamflow(USGS) and Climatological(NCDC) StationRecords
Percentile
1
5
10
50
90
95
99
All Stations (1009 USGS, 1036 NCDC)
Area
USGS
(square
miles)
6.
Elevation
USGS
(feetabove
msl)
4.
Elevation
NCDC
(feet
above
msl)
10.
Elevation
difference
(feet)
-2802.
Distance
(miles)
0.65
Runoff
percent
missing,
!948-1988
0.00
Runoff
percent
missing,
1948-1978
0.00
Precipitation
percent
missing,
1948-1988
0.01
Precipitation
percent
accumulated,
1948-1988 0.00
Train
percent
missing,
1948-1988
Traax
percent
missing,
1948-1988
Precipitation
percent
missing,
1948-1978
Precipitation
percent
accumulated,
1948-1978
Train
percent
missing,
1948--1978
Truax
percent
missing,
1948-1978
0.01
0.02
23.
24.
40.
-743.
2.98
0.00
0.00
0.03
0.00
0.!1
0.15
0.00
0.00
0.02
0.00
0.01
0.01
0.07
0.08
HUC 8+10+11+12+13,
Area
USGS
(square
miles)
Elevation
USGS(feetabovemsl)
Elevation
NCDC(feetabove
msl)
Elevation
difference
(feet)
19.
5.
35.
-3282.
Distance
(miles)
Runoff
percent
missing,
1948-1988
Runoff
percent
missing,
1948-!
978
Precipitation
percent
mis
sing,
1948-1988
Precipitation
percent
accumulated,
1948-1988
0.54
0.00
0.00
0.01
0.00
Precipitation
percent
missing,
1948-1978
Precipitation
percent
accumulated
1948-!978
0.00
0.00
Tmi
npercent
missing,
1948-1988
Trax
percent
missing,
1948-1988
Train
percent
missing,
1948-1978
Tmax
percent
missing,
1948-1978
AreaUSGS (squaremiles)
ElevationUSGS (feet above msl)
ElevationNCDC (feet above msl)
Elevationdifference (feet)
Distance(miles)
Runoff
percentmissing,1948-1988
Runoff
percentmissing,1948-1978
Precipitation
percentmissing,1948-1988
Precipitation
percentaccumulated,1948-1988
rmin percentmissing,1948-1988
rraaxpercentmissing,1948-1988
Precipitation
percentmissing,1948-1978
Precipitation
percentaccumulated,1948-1978
Train
percentmissing,1948-1978
Traax
percentmissing,1948-1978
0.01
0.02
0.00
0.02
41.
85.
140.
-338.
6.49
0.00
0.00
0.13
0.00
0.30
0.32
0.05
0.00
0.!7
0.!9
-1133.
3.55
0.00
0.00
0.03
0.00
0.09
0.18
0.03
0.00
0.04
0.06
79.
100.
180.
-416.
9.93
0.00
0.00
0.19
0.00
0.31
0.31
0.05
0.00
0.12
0.18
0.05
0.00
0.08
0.06
0.02
0.00
0.04
0.06
0.29
0.00
0.32
0.35
0.08
0.00
0.13
0.19
HUC 14+ 15+ 16, (39 USGS, 99 NCDC)
9.
! 1.
56.
2029.
2177.
2523.
2050.
2205.
2650.
-3004.
-2824.
- 1920.
4.65
8.57
9.40
0.00
0.00
0.00
Precipitation
percentmissing,1948-1988
0.00
0.00
0.00
Runoff
percentmissing,1948-1978
0.01
0.05
0.29
Precipitation
percentmissing,1948-1988
0.00
0.00
0.00
Precipitation
percentaccumulated,
1948-!988
0.00
0.07
0.29
Train
percent
missing,
1948-1988
0.00
0.05
0.32
Truax
percent
missing,1948-1988
0.00
0.02
0.06
Precipitation
percentmissing,
1948-!978
0.00
0.00
0.00
Precipitation
percentaccumulated,
!948-1978
0.00
0.04
0.12
Train
percent
missing,
1948-1978
0.00
0.05
0.12
rrnax
percent
missing,
1948-1978
AreaUSGS(squaremiles)
Elevation
USGS (feet above msl)
Elevation
NCDC (feet above msl)
Elevation
difference(feet)
Distance
(miles)
1.86
1.94
1.38
0.02
1.83
1.85
1542.
3201.
3300.
486.
37.06
11.12
0.10
5.74
0.12
6.63
6.58
5.00
0.10
6.07
6.07
2644.
1986.
6813.
6262.
449.
45.06
9.75
0.09
6.70
0.15
7.02
7.13
5.40
0.14
6.92
7.09
297!.
5957.
5510.
829.
44.50
17.05
0.15
9.O2
0.19
9.65
9.58
7.24
0.18
8.53
8.72
6564.
8160
7691
1742
60.63
24.37
0.22
17.37
0.59
18.21
18.35
17.46
0.44
18.39
!8.67
(!63 USGS, 284 NCD½)
39.
28.
38.
HUC 7+9, (136 USGS, I04 NCDC)
23.
59.
129.
357.
429.
466.
438.
512.
579.
-63.
- 15.
- 1.
0.22
6.12
8.04
0.00
0.00
0.00
0.00
0.00
0.00
0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.00
294.
700.
800.
62.
20.45
0.00
0.00
1.46
0.02
538.
933.
1052.
70.
25.83
0.00
0.00
!.66
0.03
1.94
2.03
1.44
0.02
1.92
0.93
575.
669.
810.
100.
21.98
0.00
0.00
1.85
0.03
2.24
2.26
1.80
0.03
2.08
2.16
268.
6400.
5740.
-211.
23.86
0.00
0.00
!.78
0.03
2.13
2.21
1.78
0.03
2.08
2.07
2492.
1083.
1236.
289.
42.24
2.44
0.02
8.85
0.16
10.11
10.26
7.27
0.16
8.25
8.15
4010.
7941.
7850.
1402.
40.42
14.62
0.13
8.95
0.16
10.68
10.72
7.23
0.15
7.99
8.04
8!53.
7687.
693.
59.83
14.62
0.13
9.65
0.25
10.71
10.62
7.42
0.25
8.70
8.32
4421.
! 177.
1437.
420.
51.21
11.01
0.10
12.27
0.29
13.55
13.50
11.36
0.29
11.36
11.37
5872.
8284.
9000.
1793.
43.66
19.50
0.17
12.69
0.31
14.39
14.24
!1.37
0.29
11.37
11.37
6901.
9235.
9065.
1649.
85.55
22.08
0.20
16.70
1.05
18.00
!8.33
16.05
0.78
19.13
18.94
7313.
!473.
1565.
642.
59.80
16.76
0.15
18.51
! .32
18.21
18.36
23.35
1.72
23.53
23.74
7896.
9617.
9065.
1925.
54.53
19.50
0.17
18.54
1.34
18.22
18.36
23.61
1.77
23.75
23.96
1662
WALLISETAL.: DAILYHYDROCLIMATOLOGICAL
DATASET
TABLE 1. (continued)
Percentile
1
Area USGS (squaremiles)
Elevation USGS (feet above msl)
Elevation NCDC (feet abovemsl)
Elevation difference(feet)
Distance (miles)
Runoff percent missing, 1948-1988
Runoff percentmissing,1948-1978
Precipitationpercentmissing,1948-1988
Precipitationpercent accumulated,1948-1988
Tminpercentmissing, 1948-1988
Tmax percent missing, 1948-1988
Precipitationpercentmissing,1948-1978
Precipitation percent accumulated, 1948-1978
50
90
95
131.
1320.
1120.
- 135.
18.99
0.00
0.00
1.85
0.03
2.13
2.17
1.65
888.
4741.
3966.
568.
36.56
14.62
0.13
8.67
0.19
8.84
8.58
7.01
1578.
6587.
5270.
1181.
43.78
21.94
0.19
11.51
0.40
11.32
11.24
10.97
9552.
7331.
6798.
3060.
69.95
24.37
0.22
18.09
1.33
18.83
18.87
20.23
0.33
11.27
10.83
1.12
20.80
21.11
0.13
0.16
0.03
2.04
2.11
HUC 3+6, (149 USGS, 120 NCDC)
22.
54.
76.
0.
8.
16.
5.
18.
40.
-500.
-156.
-72.
0.52
2.24
6.85
0.00
0.00
0.00
0.00
0.00
0.00
0.01
0.05
0.30
0.00
0.00
0.00
0.00
0.08
0.35
0.00
0.09
0.39
0.00
0.03
0.07
0.00
0.00
0.00
0.00
0.04
0.13
0.00
0.06
0.14
395.
311.
475.
103.
23.2 !
0.00
0.00
1.80
0.03
2.12
2.12
1.65
0.03
1.93
2.06
0.00
0.01
Area U SGS (square miles)
Elevation USGS (feet above msl)
Elevation NCDC (feet above msl)
Elevation difference (feet)
Distance (miles)
Runoff percent missing, 1948-1988
Runoff percent missing, 1948-1978
Precipitationpercent missing,1948-1988
Precipitation percent accumulated, 1948-1988
Tminpercent missing, 1948-1988
Tmaxpercent missing, 1948-1988
Precipitationpercent missing,1948-1978
Precipitation percent accumulated, 1948-1978
Tminpercent missing, 1948-1978
Tmaxpercent missing, 1948--1978
0.04
0.06
HUC 1+2+0413-0419, (208 USGS, 138 NCDC)
2.
10.
2!.
103.
3.
13.
28.
414.
15.
44.
88.
490.
-661.
-317.
-240.
64.
0.98
2.72
5.61
16.17
0.00
0.00
0.00
0.00
Runoffpercentmissing,1948-1978
Precipitationpercentmissing,I948-1988
Precipitationpercentaccumulated,1948-1988
Tminpercentmissing,1948-1988
Tmaxpercentmissing,1948-1988
Precipitationpercentmissing,1948-1978
Precipitation
percentaccumulated,
1948-1978
Tmin percentmissing,1948-1988
Tmaxpercentmissing,1948-1978
Area USGS (squaremiles)
ElevationUSGS(feetabovemsl)
ElevationNCDC (feetabovemsl)
Elevationdifference
(feet)
10
HUC 17+18, (140 USGS, 164 NCDC)
5.
13.
22.
0.
90.
191.
5.
10.
41.
-6076.
-2734.
-1286.
0.50
2.14
4.10
0.00
0.00
0.00
0.00
0.00
0.00
0.01
0.05
0.24
0.00
0.00
0.00
0.00
0.07
0.33
0.01
0.10
0.35
0.01
0.03
0.06
0.00
0.00
0.00
Tminpercent missing, 1948-1978
Tmaxpercent missing, 1948-1978
Area USGS (square miles)
Elevation USGS (feet above msl)
Elevation NCDC (feet above msl)
Elevation difference(feet)
Distance(miles)
Runoff percentmissing,1948-1978
5
0.00
0.01
0.00
0.01
0.01
0.00
0.00
0.00
0.01
0.00
0.05
0.00
0.07
0.09
0.02
0.00
0.03
0.05
0.00
0.30
0.00
0.34
0.32
0.06
0.00
0.12
0.13
HUC 0400-0412+05,
(179USGS,126NCDC)
4.
55.
86.
290.
424.
535.
462.
550.
580.
-1327.
-298.
-90.
Distance
(miles)
Runoffpercentmissing,1948-1988
Runoffpercent
missing,
1948-1978
Precipitation
percent
missing,
1948-1988
Precipitation
percent
accumulated,
1948-1988
Tmi
n percent
missing,
1948-1988
Truax
percent
missing,
1948-1988
Precipitation
percent
missing,
1948-1978
Precipitation
percent
accumulated,
1948-1978
Trnin
percent
missing,
1948-1978
Trnax
percent
missing,
1948-1978
0.30
0.00
0.00
0.01
0.00
0.00
0.01
0.00
0.00
0.00
0.00
1.92
0.00
0.00
0.05
0.00
0.09
0.10
0.03
0.00
0.05
0.06
5.31
0.00
0.00
0.31
0.00
0.37
0.41
0.09
0.00
0.14
0.22
0.00
1.87
0.03
2.14
2.17
1.66
0.03
2.00
2.11
300.
765.
845.
65.
18.38
0.00
0.00
1.84
0.03
2.18
2.17
1.67
0.03
2.00
2.08
0.19
7.85
7.83
2650.
1792.
2160.
767.
38.76
7.31
0.06
8.93
0.16
9.50
9.73
7.23
0.15
7.90
8.00
4246.
645.
1097.
1380.
575.
32.72
15.16
0.13
8.77
0.!7
9.66
9.28
7.05
0.19
7.91
7.89
908.
! 281.
1735.
826.
1131.
1679.
1810.
442.
!516.
2086.
2238.
7!8.
99
11,215.
2034.
3300.
1267.
43.52
16.05
o. 14
12.60
2840.
3840.
2091.
47.48
25.23
0.22
18.40
0.29
14.22
14.09
11.27
0.29
11.30
1.27
18.16
18.35
22.52
1.56
22.80
11.35
23.04
2640.
1660.
2233.
1237.
35.81
23.15
0.20
12.73
0.31
49.69
28.60
0.25
18.27
1.33
14.40
14.26
10.75
0.29
11.31
19.28
19.24
21.58
1.38
21.98
22.25
11.32
2853.
2772.
3390.
1386.
35.54
7.30
0.06
8.81
0.17
40.88
14.62
0.13
12.11
0.39
75.90
22.04
0.19
18.36
1.33
9.88
9.39
7.22
0.18
7.93
7.92
13.35
12.21
11.09
0.33
11.35
11.34
18.15
18.35
22.21
1.50
22.53
22.77
1square
mileequals
259ha;1footequals
30.48
cm'1mileequals
1.609
km.NCDC,
National
Climatic
DataCenter;
HUC,hydrologic
unit code; msl, mean sea level.
WALLIS ET AL.: DAILY HYDROCLIMATOLOGICALDATA SET
1663
near-surfaceatmosphericvariability, J. Clim., 2(12), 1447-1462,
1989.
Diaz, H. F., and R. G. Quayle,The climate of the United States
since 1895;Spatialand temporalchanges,Mon. WeatherRev.,
108, 249-266, 1980.
Grotch, S. L., Regionalintercomparisons
of general circulation
model predictions and historical climate data, DOE/NBB-0084,
291 pp., Departmentof Energy, Washington,D.C., 1988.
Jones,P. D., S.C. B. Raper,R. S. Bradley,H. F. Diaz, P.M. Kelly,
and T. M. L. Wigley, Northern hemispheresurfaceair temperature variations 1851-1984,J. Clim. App. Meteorol., 25, 161-179,
1986.
Karl, T. R., Perspectiveon climate changesin North America
during the twentieth century, Phys. Geogr., 6, 207-229, 1985.
Karl, T. R., H. F. Diaz, and George Kukla, Urbanization: Its
Fig.3. Hydrologic
regions
ofthecontinental
UnitedStates.
NOAAdailyclimatological
datafor precipitationand maximumandminimumtemperature. Care was taken to assure
thatthestreamflowrecordswere essentiallyfree from regulation.
The climatologicalstationswere a subsetof the HCN
stations
for which monthly data are describedby Quinlan et
al. [1987].
Missingdata were estimatedusing observationsfrom
nearbygaugesand long-term monthly means. Approximately
5%of thedayswere missingin the recordsandhadto
beestimated.The daily meteorological data were further
detection and effect in the United States climate record, J. Clim.,
1, 1099-1123, 1988.
Landwehr, J. R., and J. R. Slack, Evidence of climate change or
climate variation in dischargerecordsfor the United States, paper
presented at Chapman Conference on Hydrologic Aspects of
Global Climate Change, AGU, Lake Chelan, Wash., June, 1990.
Langbein, W. B., and J. R. Slack, Yeafly variations in runoff and
frequency of dry years for the conterminous United States,
1911.-79, Geol. Surv. Open File Rep. 82-75I, 1982.
Lettenmaier, D. P., E. F. Wood, and J. R. Wallis, Evaluation of
general circulation model hydrologic representations using a
hydro-meteorologicaldata set for the continental United States,
paper presentedat Western Pacific GeophysicsMeeting, AGU,
Kanazawa, Japan, Aug., 1990.
Mitchell, J. F. B., The greenhouseeffect and climate change, Rev.
Geophys., 27(1), 115-139, 1989.
Nellson, R. P., G. A. King, R. L. DeVelice, J. Lenihan, D. Marks,
J. Dolph, W. Campbell, and G. Glick, Sensitivity of ecological
landscapesand regionsto globalclimatic change,190 pp., Environ. Res. Lab., U.S. Environ. Prot. Agency, Corvallis, Oregon,
adjusted
to be consistent
at the monthlylevel with the HCN
1989.
dataof Quinlanet al. [1987].We believethat thesedata are Quinlan, F. T., T. R. Karl, and C. N. Williams, United States
relativelyfree from anthropogeniceffects and should be
historical climatology network (HCN) serial temperature and
precipitationdata, Tech. Rep. NDP-019, Environ. Sci. Div.,
suitable
for studyinga wide range of hydroclimatological
Carbon Dioxide Information Analysis Cent., Oak Ridge Natl.
problems
where daily data are necessary.However, users
shouldbe aware that errors do exist, especially in the
climatological
data, and shouldexercise due caution.
Acknowledgments.The support of Pacific Northwest Laboratoryundercontract DE-AC06-76RLO
1830 with the U.S. Depart-
Lab., Oak Ridge, Tenn., June, 1987.
Rind, D., R. R. Goldberg, and R. Ruedy, Change in climate
variabilityin the 21stcentury,Clim. Change,14(1),5-37, 1989.
Wigley,T. M. L., J. K. Angell,and P. D. Jones,Analysisof the
temperaturerecord,in Detectingthe ClimaticEffectsof Increasing CarbonDioxide,editedby M. C. MacCrackenand F. M.
Luther, Rep. DOE/ER-0235,U.S. Dep. of Energy, Washington,
mentofEnergy,the U.S. GeologicalSurvey undergrant 14-08-0001D.C., 1985.
G1753,and the National Aeronautics and Space Administration
undergrant nagw-1392 to the third author, and the Andrew W.
D. P. Lettenmaier, Department of Civil Engineering,FX-10,
MellonFoundationthroughsupportto the Water ResourcesPro- Universityof Washington,Seattle,WA 98195.
gramat PrincetonUniversity for hydroclimatologystudies,is grateJ. R. Wallis, IBM ThomasJ. WatsonResearchLaboratory,P.O.
fullyacknowledged.
Box 218, Yorktown Heights, NY 10598.
E. F. Wood, Departmentof Civil Engineeringand Operations
REFERENCES
Cobb,
R. F., andJ. R. Biesecker,The nationalhydrologicbenchmarkprogram,U.S. Geol. Surv. Circ. 460-D, 1971.
Delworth,
T., and S. Manabe,The influenceof soil wetnesson
Research,PrincetonUniversity, Princeton, NJ 08544.
(Received June 1, 1990;
revised March 28, 1991;
acceptedApril 2, 1991.)