Remotely sensed estimates of evaporation for ... - Hydrologie.org

286

Remote Sensing and Hydrology' 2000 (Proceedings of a symposium held at Santa Fe, N e w Mexico, U S A , April 2000). IAHS Publ. no. 267, 2 0 0 1 .

Remotely sensed estimates of evaporation for irrigated crops in northern Mexico

JAIME GARATUZA-PAYAN Insliluto Mexico

Tecnologico

de Sonora,

5 de Febrero

SI8 Sur, Ciudad

Obregon,

Sonora

85000,

e-mail: [email protected]

W. J A M E S S H U T T L E W O R T H Department of Hydrology Arizona, Tucson, Arizona

and Water Resources, 85721, USA

Harshbarger

Building

11, The University

of

R A C H E L T. PINKER Department of Meteorology, Maryland 20742, USA

Space

Sciences

Building,

University

of Maryland,

College

Park,

C H R I S T O P H E R J. W A T T S Instituto Sonora,

del Medio Mexico

Ambienle

y Desarrollo

de Sonora,

Reyes y Aguascalienles,

Hermosillo,

Abstract Hourly estimates of solar radiation were derived from satellite data for the Yaqui Valley in Mexico, made on a 50 km grid using the GEWEX/SRB algorithm applied with GOES-East data and, on a 4 km grid, using a high-resolution development of the algorithm with GOES-West data. On average, values derived from GOES-East are 18% greater, while those from GOES-West are 9% lower than field measurements. After re-calibration, random differences between hourly satellite estimates and surface observa tions remained. These were markedly reduced when daily-average values were compared. Root mean square error (RMSE) between the satellite and the surface measurements is lower for the high-resolution satellite estimates than it is for the low-resolution estimates, and there is a noticeable increase in apparent structure with the high-resolution data. Finally, the application of the high-resolution estimates of solar radiation to calculate daily estimates of crop evaporation for wheat and cotton fields is demonstrated. K e y w o r d s cotton; évapotranspiration; G O E S ; Makkink equation; Mexico; remote sensing; solar radiation; Sonora; wheat

INTRODUCTION Knowledge of evaporation is essential to define irrigation water requirements when planning, designing, and scheduling irrigation schemes. The Yaqui Valley irrigation scheme in Mexico is the focus of attention for the research described in this paper, where timely crop evaporation estimates could provide important information for the more efficient use of water. Satellite observations should be of value for estimating evaporation. One success ful study, the TiSDat (Timely Satellite Data for Agricultural Management) project (Diak et al., 1998), used multiple GOES (Geostationary Operational Environmental Satellite) satellite images available each day to estimate regional evaporation for


287

irrigation scheduling. This encourages further investigation of this approach. Incoming short-wave solar radiation is the main factor controlling potential evaporation, and previous studies in the Yaqui Valley have shown that geostationary satellite data can be used to estimate solar radiation (Stewart et al, 1999). It also has been shown (Garatuza-Payan et al., 1998) that potential evaporation (XE ), in the Yaqui Valley, can be estimated from solar radiation and temperature using a locally-calibrated version of the Makkink equation. The present work explores the potential use of GOES data to provide daily estimates of crop evaporation in the Yaqui Valley for wheat and cotton fields through a growing season using the Makkink equation. P

EXPERIMENTAL AREA, MODELS, MEASUREMENTS AND METHODS The Yaqui Valley irrigation scheme is located on the coastal plain near Ciudad Obregon (27°28'N, 109°59'W) in northwest Mexico. The average annual rainfall is 270 mm. The irrigation scheme covers an area of 40 x 70 km and is made up of a patchwork of fields of between 25 and 400 ha for different irrigated crops. The main crops are wheat and cotton with some maize and vegetables, which are grown through-out the year. The Penman-Monteith equation (Monteith, 1965) is the most widely recognized, physically-based equation for describing évapotranspiration from uniform vegetation. However, in the case of irrigated agricultural crops, canopy cover is not always complete and net radiation is usually strongly determined by solar radiation. For these reasons, a simpler model has been proposed (the Makkink equation) which has proven reliable for estimating XE in the Yaqui Valley region (Garatuza-Payan et al, 1998), and has the form: P

XE =C -^-R P

M

S

(1)

A+y 2

where R is the short-wave radiation (W ra ), À is the slope of the saturation vapour pressure curve (Pa K" ), y is the psychometric constant (Pa K" ), X is the latent heat of vaporization (J kg" ), and CM is a locally relevant empirical calibration constant. The value of CM has been determined to be approximately 0.65 on a yearly basis in the Yaqui Valley (Garatuza-Payan et al, 1992). The term À/(À + y), taking account of the temperature dependency of the saturation vapour pressure, changes monotonically by 1-2% for a 1°C change in temperature. This implies that the Makkink equation can be used to derive a daily estimate of XE using only climatological air temperature data and measurements of the daily total energy incident as solar radiation. The GEWEX/SRB (Global Energy and Water Cycle Experiment/Surface Radiation Budget) algorithm (Whitlock et al, 1995), based on the model developed by Pinker & Lazslo (1992), was used to infer the downward solar radiation at the ground from satellite observations. In context of the G E W E X Continental Scale International Project (GCIP), the G E W E X / S R B algorithm is used with data from the GOES-East satellite to provide satellite estimates of surface and top-of-the-atmosphere radiative fluxes within one to two days of image capture at one-hourly intervals on an equal area 0.5° grid (Pinker et al, 2001). hi the present study, these data were used to provide the 50 km grid GOES-East satellite-based estimates of solar radiation for the period November 1998-March 1999. S

1

1

1

P

288

Jaime Garatuza-Payan et al.

A modified version of the GEWEX/SRJ3 algorithm was implemented and used to provide estimates on a 4 km grid using data from the GOES-West satellite that were received at the Instituto Tecnologico de Sonora (ITSON), for the same period. Halfhourly visible images for the area 22.5-36°N and 102-117.5°W were collected, preprocessed, and stored. Image pre-processing involves applying a cloud detection algorithm, which assumes that any partial cloud cover in a selected target area increases the spatial variance in the visible radiance over the area. All pixels in a 4 x 4 target area were allocated between clear-sky and totally cloud-covered categories as explained in Garatuza-Payan et al. (2001). The daily estimates of potential evaporation were combined with locally-calibrated crop factors to provide estimates of the actual daily evaporation for individual areas of crop in the irrigation scheme, following: F

= K

^ crop, total

where 7\

J

C v

c ^ M

A

(R , A s lot

(2)

A +y

1

] is the daily total solar radiation estimated from satellite data in J day" , Ticrop.totai is the estimated crop evaporation in m m day" , and K is the appropriate crop factor for the specific crop and day in the crop growth cycle as provided by GaratuzaPayan er: al. (1998). To provide validation, the satellite-based estimates of solar radiation were compared with surface observations in the irrigation region. Solar radiation was measured through out the study period with Eppley pyranometers at two sites, Site 910 (27.37°N, 109.92°W) and Site 1517 (27.20°N; 110.18°W), both sites characterized by irrigated crops. S]tota

1

c

R E S U L T S A N D DISCUSSION Table 1 gives the monthly average solar radiation as estimated by the two satellites using the current satellite calibrations. In both cases, there is an obvious, systematic, and persistent discrepancy as compared to ground measurements. When the values are averaged over the whole period for which data are available, the estimates derived from GOES-East are 18% greater than field data, while those from GOES-West are 9% lower. For the present study, recognizing the currently poor calibration of GOES-East and GOES-West, the two satellite estimates were re-calibrated (by - 1 8 % and + 9 % , respectively) to give time-average agreement with ground observations. Figure 1 Table 1 Average solar radiation (W m" ) over each month from November 1998 to March 1999 and the whole study period observed at two sites, Sites 910 and 1517, in the Yaqui Valley irrigation scheme compared with equivalent estimates made from data from the GOES-East and GOES-West satellites. Site 910: Ground

GOES-W

GOES-E

Site 1517: Ground

November December February March

429 375 467 596

397 312 427 564

524 498 557 668

417 366 483 607

Total

466

426

562

468

GOES-W

GOES-E

405 312 414 552 421

530 495 545 664 559

289


Hourly; Site 9 1 0

(a)

Hourly; S i t e 1 5 1 7

£ g

J

800 700

«600 O «500 c

o '« 4 0 0 TD CO 300 •o 0 1

I

200

LU

100

200

400

600

200

Daily-average: Site 9 1 0

100

400

600

800

Dairy-average: Site 1 5 1 7

200 300 400 500 600 O b s e r v e d Radiation (W m - ) 1

700

100

200 300 400 500 600 O b s e r v e d Radiation (W m )

700

!

Fig. 1 Comparison between ground measurements of solar radiation at two sites (910 and 1517) in the Yaqui Valley and the equivalent solar radiation estimated from satellite: (a) derived at a 4 km grid resolution from GOES-West, and (b) derived at a 50 km grid resolution from GOES-East.

shows a comparison between the resulting hourly average and daily average satellite estimates vs observed solar radiation at two surface sites. Table 2 shows the root mean square error (RMSE) of the mean value for hourly and daily averages of solar radia tion. The R M S E for daily estimates from GOES-West are in the range 7 - 1 4 % (30-55 W m" ) and are consistent with results given by other studies (Stewart et al., 1999). Few authors have reported hourly estimates; however Dedieu et al. (1987) reported a R M S E of 2 0 % for a study in France, while Stewart et al. (1999) obtained a R M S E of 20.2% in the Yaqui Valley using a different radiation model. These compare with the equivalent values found during the present study that lie in the range 9 - 1 4 % (47-96 W m" ). The daily R S M E values are significantly less than the hourly R M S E because the satellite provides an instantaneous area-average estimate, while the field observation is a time-average single-point measurement. Results given in Table 2 and Fig. 1 show that the high-resolution solar radiation estimates based on GOES-West data show somewhat better agreement with the surface measurements than do the lower-resolution estimates based on GOES-East data. The greater resolution in the spatial distribution given by the 4 km grid should provide more accurate estimates of incoming solar radiation since more accurate estimates may be obtained for clear and cloudy areas. This should lead to more reliable estimates of evaporation using field scale information on crop cover. 2

2

Jaime Garatuza-Payan et al.

290

2

Table 2 Root mean squared errors in W m' , and as a percentage of the average flux (in brackets) between hourly average and daily average estimates based on GOES-East and GOES-West data, relative to ground-based observations at sites 910 and 1517.

November December February March Total

Site 1517: GOES-West

Site 910: GOES-West

GOES-East

Hourly

Daily

Hourly

Daily

Hourly

Daily

Hourly

Daily

53.73 (10.35) 66.67 (13.39) 79.69 (13.95) 95.97 (14.23)

42.97 (9.91) 55.36 (14.75) 62.15 (13.48) 53.00 (8.92)

62.24 (12.71) 75.92 (15.79) 91.68 (18.17) 89.57 (14.78)

34.96 (7.78) 41.68 (11.15) 45.95 (9.94) 53.72 (9.11)

46.93 (9.32) 49.03 (10.84) 79.42 (13.4) 77.10 (11.43)

29.93 (7.07) 50.28 (13.75) 63.75 (13.48) 78.08 (12.90)

63.51 (13.17) 71.45 (15.34) 98.48 (18.66) 98.24 (18.54)

46.67 (11.28) 48.96 (13.35) 73.55 (16.17) 54.80 (9.12)

72.83 (13.22)

50.91 (11.17)

80.52 (15.52)

39.21 (9.61)

65.80 (12.06)

53.59 (11.78)

83.65 (16.27)

55.61 (12.35)

GOES-East

Figure 2 demonstrates the feasibility of using satellite data to estimate the evapora tion from the component crops in the Yaqui Valley irrigation scheme. This shows the evaporation estimates during a growth season for two fields, one centred on 27.37°N, 109.92°W which was planted with wheat on 16 November 1998, and one centred on 27.20°N, 110.18°W which was planted with cotton on 1 January 1999. The evapora-

Fig. 2 Evaporation estimates calculated for two fields (a) planted with wheat on 16 November 1998, and (b) planted with cotton on 1 January 1999. In each case, the daily average net radiation is also shown.


291

tion estimates are derived by multiplying the daily estimates of potential evaporation for each field by the relevant crop factor as specified by Garatuza-Payan et al. (1998). The estimate of potential evaporation used in these calculations is that given by the Makkink equation from the high-resolution (4 km) satellite estimates of solar radiation for each field.

Acknowledgements Primary support for this analysis at The University of Arizona was provided under N A S A Grant NAG8-1531. In addition, Jaime Garatuza-Payan received support under a C O N A C Y T Fellowship and project 29340T. The satellite data from GOES-West were gathered as part of research supported by the European Union (CI1-CT94-0059).

REFERENCES Dedieu, G., D e s c h a m p s , Y. & Kerr, Y. H. ( 1 9 8 7 ) Satellite estimation of solar irradiance at the surface of the Earth and surface a l b e d o using a physical model applied to Meteosat data. J. Appl. Met. 26, 7 9 - 8 7 . Diak, G. R., A n d e r s o n , M . C , Bland, W . L., N o r m a n , J. M . , Mecikalski, J. M . & A u n e , R. M . ( 1 9 9 8 ) Agricultural m a n a g e m e n t decision aids driven b y real-time satellite data. Bull. Am. Met. Soc. 7 9 ( 7 ) , 1 3 4 5 - 1 3 5 5 . G a r a t u z a - P a y a n , J., Pinker, R. T. & Shuttleworth, W. J. ( 2 0 0 1 ) High resolution cloud observations for N o r t h w e s t M e x i c o from G O E S - 7 satellite data. J. Atmos.

Ocean

Technol.\0S(\),

39-56.

G a r a t u z a - P a y a n , J., Shuttleworth, W . J., M c N e i l , D . D., Stewart, J. B . , D e Bruin, H. & W a t t s , C. ( 1 9 9 8 ) M e a s u r e m e n t and m o d e l l i n g evaporation for irrigated c r o p s in N o r t h w e s t M e x i c o . Hydrol.

Processes

12, 1 3 9 7 - 1 4 1 8 .

G a r a t u z a - P a y a n , J., W a t t s , C. J. & Cervera, L. E. ( 1 9 9 2 ) Estimacion d e evapotranspiracion p o t e n t i a l en el Valle del Yaqui (Potential évapotranspiration estimates in the Yaqui Valley). In: XII Congreso de Hidraulica (Puerto Vallarta, Mexico). Monteifh, J. L. ( 1 9 6 5 ) E v a p o r a t i o n and environment. In: Proc. Symp. of the Society

of Experimental

Biology

19, 2 0 5 - 2 3 4 .

Pinker, R. T., Tarpley, J. D . , Laszlo, I. & Mitchell, K. ( 2 0 0 1 ) Surface radiation b u d g e t s in s u p p o r t of the G E W E X Continental Scale International Project ( G C I P ) . J. Geophys.

Res. (submitted).

Pinker, R. T. & L a z l o , I. ( 1 9 9 2 ) M o d e l i n g surface solar irradiance for satellite applications o n a global scale. J. Appl. Met. 31(2), 194-211. Stewart, J. B . , W a t t s , C. J., Rodriguez, J. C , D e Bruin, H. A. R., V a n den B e r g , A. R., G a r a t u z a - P a y a n , J. ( 1 9 9 9 ) U s e of satellite data to estimate radiation and evaporation for northwest M e x i c o . Agric. Wat. Manage. 3 8 , 1 8 1 - 1 9 3 . W h i t l o c k , C. H., Charlock, T. P., Staylor, W . F., Pinker, R. T., Laszlo, L, O h m u r a , A., Gilgen, H., K o n z e l m a n , T., D i P a s q u a l e , R. C , M o a t s , C. D., LeCroy, S. R. & Ritchey, N . A. ( 1 9 9 5 ) First global W C R P s h o r t w a v e surface radiation b u d g e t data set. Bull. Am. Met. Soc. 7 6 ( 6 ) , 1-18.