Behavior of completely automated evapotranspiration estimation (EEFlux)
The Earth Engine Evapotranspiration Flux (EEFlux) application was designed and developed on the google earth Engine platform and utilizes Landsat imagery archive.
We can request ET maps on EEFlux in a matter of seconds.
icon search Searchable Transcript
Toggle between list and paragraph view.
[00:00:00.820]Welcome to the UNL students research
[00:00:04.199]My name is Atiqullah Atif and I am a
[00:00:06.799]graduate student doing my master's in
[00:00:08.940]water resources engineering
[00:00:10.431]Department of civil and environmental
[00:00:12.521]Engineering at UNL.
[00:00:13.699]In the next few minutes, I will walk
[00:00:15.800]through this poster presentation.
[00:00:17.760]The topic of this research poster is
[00:00:20.727]Behavior of completely automated
[00:00:23.173]evapotranspiration estimation or in
[00:00:27.033]short we say EEFlux.
[00:00:28.915]So, let's get started and I'll zoom in
[00:00:32.404]to the Introduction.
[00:00:35.556]Water is lost to the atmosphere from soil,
[00:00:37.916]ponds and vegetation in the form
of water vapor.
[00:00:41.000]This natural phenomenon is called
[00:00:43.151]Evapotranspiration or ET. The importance
[00:00:46.171]of ET is such that it is the second
[00:00:48.375]largest term in the water balance equation
[00:00:52.343]ET is responsible for 75-90 percent
[00:00:55.950]consumption of the annual precipitation
[00:00:58.623]in the U.S. It also dictates the
[00:01:01.475]recharge and discharge of aquifers.
[00:01:04.220]We have several techniques for estimating
[00:01:06.640]ET, which are time consuming and have
[00:01:09.168]their own limitations. The Earth Engine
[00:01:12.119]Evapotranspiration Flux or we call it
[00:01:16.119]EEFlux application was designed and
[00:01:19.495]developed on the Google Earth Engine
[00:01:22.001]platform and utilizes Landsat
[00:01:25.988]We can request ET maps on EEFlux in a
matter of seconds.
[00:01:31.020]Now, let's walk through the
[00:01:32.923]background work on ET estimation.
[00:01:36.185]There are accurate techniques
to estimate ET
[00:01:39.174]but they cannot be used on regional
scale due to the high number and
[00:01:43.724]operational cost requirements on
[00:01:46.888]To overcome this obstacle, surface
[00:01:50.416]technique is used in satellite-based image
[00:01:52.823]processing models like Mapping
[00:01:55.121]Evapotranspiration at high Resolution
[00:01:57.807]with Internalized Calibration or in
[00:02:00.338]short we call it METRIC to calculate ET
as a residual of
[00:02:04.746]surface energy balance.
[00:02:06.406]However, despite being accurate,
[00:02:09.634]METRIC model requires skilled personnel
[00:02:13.634]and run for each Landsat scene
and image date,
[00:02:16.891]and this proves it to be time consuming.
[00:02:19.587]To minimize human interference,
[00:02:22.222]automated calibration algorithms were
[00:02:26.425]METRIC model. The automated application
[00:02:30.709]Google Earth Engine as a platform and is
referred to as
[00:02:34.821]Earth Engine Evapotranspiration Flux or we
[00:02:38.150]can say EEFlux and eeMETRIC.
[00:02:40.935]Now, we talk about the objectives.
[00:02:43.043]The objectives of this research study
[00:02:47.043]Observe the behavior of eeMETRIC.
[00:02:50.612]Compare eeMETRIC results with the manual
[00:02:53.992]results to asses the utility and accuracy
[00:03:02.857]Since, eeMETRIC is based on manual METRIC,
[00:03:06.058]we will examine the difference of their
products to confirm
[00:03:09.804]its functioning on Earth Engine.
[00:03:12.643]Talking about materials, we will need
[00:03:15.883]materials for this study
[00:03:17.358]number one, daily, monthly and seasonal ET
[00:03:20.883]maps will be needed form both models.
[00:03:25.563]Number two ArcGIS application for
[00:03:30.091]eeMETRIC and EEFLux
[00:03:32.258]a web-based interface for requesting
[00:03:35.531]within seconds. And Microsoft EXCEL to do
[00:03:41.212]Here is figure 1 showing the material
we will need for this study.
[00:03:46.954]Now what is the study area?
[00:03:48.876]The study is performed on the
[00:03:51.961]highlighted in Figure 2 because we
[00:03:55.505]METRIC data already available for
[00:03:59.289]Here we can see the locations
[00:04:02.685]highlighted with these red rectangles.
[00:04:06.279]Here are the location names and
Landsat path/row specifications.
[00:04:11.446]Here are the remaining two locations
of this study area
[00:04:19.525]Now we come to the Methodology and
[00:04:22.943]the following steps:
First is obtaining ET maps
[00:04:26.379]from both models.
Second we add the maps to
[00:04:29.785]ArcMap file as raster layers.
[00:04:33.053]We might need reprojections and or
[00:04:37.577]Then we choose sampling fields with the
help of NLCD Land Use/Land Cover layer.
[00:04:44.276]Then we extract values for each
[00:04:48.276]and summarize them into a table.
[00:04:50.621]Then export the table to an
[00:04:54.675]And finally we create scatterplot to see
[00:04:58.455]correlation, slope of regression line,
[00:05:01.554]and calculate index agreement.
[00:05:03.960]Talking about results, so far,
the study is only
[00:05:07.060]performed on Snake River Plain data.
[00:05:09.435]And it will be performed for the
[00:05:13.435]mentioned in study area. Here are
some examples products.
[00:05:17.200]The left one is the daily ET map from
[00:05:19.869]manual MATRIC application for snake river
[00:05:22.733]plain and the right one is the daily
ET map from eeMETRIC.
[00:05:28.796]Here is the scatterplot you can see.
[00:05:31.836]it's showing strong correlation,
[00:05:33.867]the R square value is close to one.
[00:05:36.487]And the Index of agreement is
also close to one.
[00:05:40.765]Now here we discuss some the primary
results from this study.
[00:05:46.353]the average values for R square
and Index of
[00:05:49.997]Agreement for Snake River Rlain data
for 12 overpass
[00:05:54.494]dates are respectively 0.949 and 0.941.
[00:06:01.574]These results can further improve after
cloud masking is applied.
[00:06:06.817]Also, the study of the time-based
[00:06:08.744]interpolation methods will provide further
[00:06:11.397]insight about the behavior of eeMETRIC.
[00:06:14.774]Now we come to the conclusion.
[00:06:17.156]The so far results show that eeMETRIC gives ET
[00:06:20.958]estimations comparable with manual METRIC.
[00:06:23.523]We expect similar results from the
reaming 28 Landsat
[00:06:28.105]image sites, which will be studied later.
[00:06:30.723]And larger differences are noted for
non-agriculture land use.
[00:06:35.497]And the reason could be that eeMATRIC
[00:06:38.315]struggles to account for background
[00:06:40.452]evaporation at the hot pixel calibration
[00:06:44.543]or it could be due to a bias introduced
[00:06:47.708]application of the Evaporative
Fraction to extrapolate
[00:06:51.278]instantaneous ETrf to daily ETrf.
[00:06:55.717]At the end we acknowledge and appreciate
the support and
[00:06:59.965]funding by google earth engine to
initiate EEFlux and eeMETRIC.
[00:07:04.372]Also, thanks to the USGS, NASA
and Windward Fund to
[00:07:08.687]support Landsat team. And last but
not the least, a special
[00:07:13.514]thanks to IIE and Fulbright for
funding this study.
[00:07:19.825]And with this we come to the end of this
[00:07:25.422]Thank you all for your attention.
Log in to post comments