Evapotranspiration Estimates from Remote Sensing for Irrigation Water Management - Satellite-based Applications on Climate Change - page 198

Table 13.1 Weather stations providing meteorological data for the three water districts

(ORMVAs)
ORMVA Weather station Elevation (m) Latitude (N) Longitude (W) Data quality

Tadla Oulad Gnaou 450 32.290 6.522 Good

Doukkala Khemis Zemamra 165 32.622 8.699 Good

Haouz Ounasda 430 32.117 7.350 Good

using the traditional SEBAL approach (Bastiaanssen et al. 1998 ; Hong 2008 )
because METRIC required calibration coefficients that are not available for
Morocco. These and other details of the study are fully described in the project
final report (Riverside Technology, Inc. 2010 ).
Morocco has a network of weather stations that provided meteorological data for
agriculture, water management, and other applications. Bi-hourly data from three of
these stations (Table 13.1 ) were acquired for this study and included the following
parameters: actual vapor pressure, solar radiation, wind speed and direction, rela-
tive humidity, air temperature and dew point, and computed reference ET (alfalfa).
The meteorological data required quality control, including some filling and error
correction, but overall were of good quality and adequate for this study.
A complete inventory of available Landsat 5 images and key meteorological
inputs was conducted for each of the three Morocco water districts. A total of 18
Landsat images were acquired and processed for the 2006-2007 study period; an
example Landsat image of the Doukkala district is shown in Fig. 13.3 . Landsat 5
images are acquired every 16 days, whereas the lower resolution MODIS images
are acquired much more frequently. For most places on the earth, two MODIS
measurements per day are collected, one in the morning (Terra) and another in the
afternoon (Aqua). However, the view angle on some days can be as much as 57 off
nadir, reducing the clarity of image detail and mixing information among adjacent
pixels. Typically an image is collected every 4 days with each MODIS satellite
when the view angle is within 20 of nadir, yielding high detail (up to 250 m in

the red and near infrared bands). For parts of the study region, the time periods
between cloud-free Landsat images were 32 or more days, requiring interpolation
of ET r F . Because of these excessive time gaps between useful Landsat images,
Allen and Trezza ( 2009 ) developed a data fusion method for extending the
METRIC-based ET record using MODIS as a transfer mechanism between Landsat
images. For this technique, ratios of ET r F were developed from the cloud-free
Landsat images (30-m ground resolution) to the underlying NDVI from MODIS
images (250-m resolution) acquired on, or about, the same day. The ratios were then
interpolated between Landsat images, and for every available MODIS date, ET r F
was estimated by multiplying MODIS NDVI by the interpolated ratios. The result
was a time-integrated ET r F that followed the evolution of NDVI as monitored by
MODIS, but retained much of the 30-m resolution of Landsat. This technique does
not account directly for evaporation from precipitation or irrigation occurring
between Landsat image dates; rather, it estimates evaporation for the image gap
period based on the wetting frequencies prior to and beyond the endpoint Landsat
image dates.