DevelopingaLong-termGlobalReservoirProductSeriesbyFusingMulti-SatelliteObservations

October15-19,2018MODIS/VIIRSScienceTeamMeeting

1

PI:HuilinGaoDepartmentofCivilEngineering

TexasA&MUniversity

GlobalReservoirandDam(GRanD)Database

Lehner etal.,2011

Globalcapacity>6200km3

Chaoetal.,2008

GlobalReservoirs

1

ReservoirStoragevs.ReservoirEvaporation

Hydropower

Irrigation/watersupply

Floodcontrol

Recreation

• Evaporation>Industrial+Domestic

• ~265km3 in2010

https://www.flickr.com/photos/89241789@N00/172762894

• LakeMeadannualevaporation=791,000acre-feet

2

• We hardly know anything about the reservoir storage in some regions (e.g. developing regions with conflicts of interest)

• Different reservoirs are operated using different rules, but information about operating rules are usually not shared

• In situ reservoir evaporation data is extremely limited

Motivation

• Drought and flood monitoring• Water resources management• Hydrological modeling • Coupled atmospheric-hydrologic

modeling• Hydrodynamic modeling• Ecosystem service

Limited data availability Targeted Science Applications

3

WaterSurfaceElevationfromRadarAltimetry

https://ipad.fas.usda.gov/cropexplorer/global_reservoir/ 4

Objectives

Togenerateacomprehensive,coherent,andlongtermglobalreservoirproduct seriesatimprovedspatialcoverage(andatimprovedtemporalresolution)bycombiningMODIS/VIIRS observationswithdatafromothersatellitesensors.

• Developalongtermreservoirstoragevariationdatasetunderall-weatherconditions.

• Generateafirst longtermrecordofthereservoirevaporationrateandtheevaporationloss.

• Validatethereservoirproductseriesandquantifytheuncertaintiesassociatedwiththedatasets.

5

9

Storage variations

Elevation Altimetry (Radar/Lidar)

Area

Optical(MODIS/VIIRS)

Microwave(AMSR-E, AMSR2)

Area-elevationcurve

Reservoir Storage Variations

Dataset Number Sensors Period

Cretaux etal2011(Adv.SpaceRes) 15

Variousimagers+radaraltimeters

1992-2011(present)

Gaoetal.2012(WRR) 34

MODIS+radaraltimeters

1992-2010

Buskeretal.2018,(HESS)

135(includinglakes)

Landsat+radaraltimeters 1984-2011

Availableglobalreservoirstoragevariationdatasets

6

Vo = Vc – (Ac+Ao)(hc-ho)/2

SouthAsia

(b)

Improving the Spatial Coverage of the Product

60km

Radaraltimetertrack

7

SouthAsia

4km

ICESat altimetertrack

MODIS+Radaraltimeter=34reservoirs(1164km3,19%ofglobalcapacity)

MODIS+ICESat =21reservoirs(83.9km3,28%ofregionalcapacity)

Gaoetal.,2012

Zhangetal.,2014

Improving the Spatial Coverage of the Product

8

MODIS+Radaraltimeter&ICESat =180reservoirs(2880km3,46.5%ofglobalcapacity)

339.78km2 590.50km2(74%morewater)

k meansclassification Enhancedclassification

9

Reservoir Area from MODIS

k meansclassification Enhancedclassification

350.56km2 360.38km2(3%morewater)

MODISNDVIonday097in2005

MODISNDVIonday273in2005

Zhang et al., 2014

Towards a Continuous Storage Data Record

StorageEstimation

Area-elevationrelationshiph = f(A)

10

V = g(A)

ICESat/radarelevatio

n(m

)

Gao,2015

Lietal,2017

MODIS+radaraltimeter

MODIS+ICESat

LakeMead

PongReservoir(India)

Monitoring Reservoir Storage under All-Weather ConditionsMODISandAMSR-Eat16-day

11Gaoetal.,2006,JHM

ZhangandGao,2016Rengali Reservoir(India)

1

1

( )

( )

T

S

N

i iiN

jjj

W TbH TWHR

W TbH S=

=

´=

´

åå

Validation Results

12

observation altimetry estimated MODIS smoothed

Enhanced algorithm after Zhang et al., 2014Algorithm after Gao et al., 2012

(Gao et al., 2012)

(Zhang et al., 2014)

MODIS+radar

ICESat+radar

Uncertainty Analysis

13

UncertaintiesduetoMODISarea+radarelevation

UncertaintiesduetoMODISarea

UncertaintiesduetoICESat elevation

Gaoetal.,2012 Zhangetal.,2014)

Evaporation loss

Evaporation rate

MODIS LST, area+ meteorology data

Area

Optical(MODIS/NPP)

Microwave(AMSR-E, AMSR2)

Reservoir Evaporation Estimation

FUTURENEEDS.Uniform,coherent,andlong-termmeasurements.

Friedrichetal.,2017BAMS

14

AreaX evaporationrate

𝑬 =𝒔(𝑹𝒏−∆𝑼) + 𝜸𝒇(𝒖𝟐)𝜹𝒆

𝝀𝒗(𝒔 + 𝜸)

Heatstorageeffect:SpringandSummer– absorbheatFallandWinter– releaseheat

AirTem

perature(ºF)

Evaporation(in

ch)

Jul

SepLakeTahoe2014

PenmanEquation

𝒔: slopeofthesaturationvaporpressurecurve(kPa·ºC-1)𝑹𝒏:netradiation(MJ·m-2·d-1)𝜸:psychrometricconstant(kPa·ºC-1)𝒇(𝒖𝟐): windfunction(s·m-1)𝜹𝒆:vaporpressuredeficit(kPa)𝝀𝒗:latentheatofvaporization(MJ·kg-1)

Heat Storage Effect on the Evaporation Rate

15

𝐸 =𝑠(𝑅7−∆𝑈) + 𝛾𝑓(𝑢<) 𝑒> − 𝑒?

𝜆A(𝑠 + 𝛾)

∆𝑈 = ℎ𝜌𝑐[𝑇G 𝑡 − 𝑇G 𝑡 − ∆𝑡 ]

Improving the Estimation of Reservoir Evaporation Rate

TwfromMODISLSTLake Tahoe

(Friedrich et al., 2017)16

(L)

DynamicMODISArea

𝑓 𝑢< = 𝜆A(2.33 + 1.65𝑢<)𝐿QR.S

(McJannet etal.2012)

LakeMead(AZ) WhiteBearLake(MN)

RossBarnett(MS) LakeCalm(FL) LakeFive-O(FL)

Validation Results & Uncertainties

17

Shadingrepresentsuncertaintiesfromforcings:

a-c:Eddycovariance;d-e:Bowenration

Validation Results & Uncertainties

18

Trends of Reservoir Evaporation

ZhaoandGao,tobesubmitted. 19

1984-2015

Planed Schedule of Activities

SWOTICESat-2

Sentinel-3

20

Summary

21

ü Long term, consistent, remotely sensed reservoir storage and evaporation productscan be used to support many applications.

ü By leveraging MODIS/VIIRS observations along with satellite altimetry data, thespatial and temporal coverage of remotely sensed global reservoirs can besignificantly improved.

ü The estimation of reservoir evaporation rate can be improved using the PenmanEquation, with the heat storage and fetch considered. The lake heat storage can beestimated using MODIS/VIIRS LST, while the fetch is a function of the dynamic lakearea.

ü Both the storage and evaporation algorithms have been validated using in situ data.

ü More comprehensive uncertainty analyses are needed for the final product.