Elsevier

Agricultural Water Management

Volume 271, 1 September 2022, 107789
Agricultural Water Management

Sensitivity of METRIC-based tree crop evapotranspiration estimation to meteorology, land surface parameters and domain size

https://doi.org/10.1016/j.agwat.2022.107789Get rights and content

Highlights

  • Daily evapotranspiration of tree crops was estimated by METRIC model.

  • The impacts of variation in meteorological variables on METRIC-based ET were evaluated throughout the entire growing seasons.

  • The impacts of variation in land surface parameters on METRIC-based ET were assessed throughout the entire growing seasons.

  • The impact of varying study sizes on METRIC-based ET estimates was tested.

Abstract

Evapotranspiration (ET), a key component of crop water balance, is essential for growers to manage agricultural irrigation. METRIC model is increasingly used to estimate and map ET with satellite observations, especially at daily and field scales. However, the analysis of its uncertainties on tree crop ET estimation is very limited. We used METRIC model to estimate daily ET over the almond and pistachio orchards throughout the entire growing season in the San Joaquin Valley, California, and evaluated the impacts of variations in meteorological variables, land surface parameters and different study sizes on ET estimation. The results showed that the METRIC-based daily ET estimates of the almond and pistachio trees were most sensitive to net radiation (Rn), followed by instantaneous reference ET (ETo), land surface temperature (LST), and near-surface temperature difference (dT). When a 5% error was added to Rn, the uncertainty in daily ET estimates was up to 8.44 ± 1.24%, corresponding to a mean increase of 0.45 ± 0.02 mm/day throughout the entire growing season. A 5% or 0.5 K error in instantaneous ETo, LST, and dT also caused changes in daily ET estimates by 4.15 ± 3.75%, 2.97 ± 0.14%, and 2.31 ± 1.20%, respectively. ET estimates with METRIC were relatively much less sensitive to other variables. The results from different study sizes indicated that daily ET estimates in Size-2 were 0.33 ± 0.08 mm/day overestimation compared with those from Size-1 in the almond site. Similarly, a mean overestimation of 0.30 ± 0.14 mm/day was found between Size-1 and Size-3 in the pistachio sites throughout the entire growing season. This study demonstrates the potential to further improve the accuracy of remote sensing based ET estimation for complex tree crops with METRIC model, by optimizing key parameters and choosing a proper domain size for study area.

Introduction

Evapotranspiration (ET) is an important component of the hydrological cycle and surface energy balance that links water, energy, and carbon cycles (Campbell and Norman, 1998, Jung et al., 2010, Zhang et al., 2019). It is closely related to crop water use in the context of agricultural irrigation management (Rana et al., 2005, Williams and Ayars, 2005, French et al., 2015, He et al., 2017, Xue et al., 2020). Accurate estimation of crop ET can improve crop water use efficiency and rationally allocate water resources (Qiu et al., 2019, Han et al., 2021, Wang et al., 2021), especially for water scarce regions. Compared with ground-based methods to measure ET, such as eddy covariance (Twine et al., 2000, Gu et al., 2012), lysimeters (Allen et al., 1991), and surface renewal techniques (McElrone et al., 2013, Shapland et al., 2014), remote sensing-based ET models have a great advantage of estimating and mapping ET in the temporal and spatial patterns. It would be providing a scientific guidance for improving water resource management.

Mapping evapotranspiration at high resolution with internalized calibration (METRIC) was proposed by Allen et al. (2007a) based on the theoretical and computational basis of surface energy balance algorithms for land (SEBAL, Bastiaanssen et al., 1998) for estimating ET by incorporating satellite observations and a few of high quality weather data. METRIC model has its significant advantage over SEBAL in that using the weather station-based reference ET (ETr) to serve as a real ground reference on the satellite image-based actual ET (Allen et al., 2007a). The ETr was used to build surface energy balance conditions at the cold and hot pixels for a linear relationship between land surface temperature (LST) and near-surface temperature difference (dT) in sensible heat flux (H) estimation. Also, METRIC model applies fraction of reference ET (ETrF) as a real-time crop coefficient for extrapolating from instantaneous ET (satellite overpassing time) to daily ET (satellite overpassing date) (Allen et al., 2007a, Allen et al., 2011). Contrasting with other remote sensing ET models, such as the LST-based Atmosphere Land Exchange Inverse (ALEXI) model at continental scales (Anderson et al., 2007) and Simplified Surface Energy Balance (SSEB) model with MODIS images (Senay et al., 2007), METRIC model was initially designed for regions smaller than a few hundred kilometers in scale and produced high accurate ET maps at high spatial resolution (Allen et al., 2007a). For the relatively complex terrain and weather conditions, it was suggested to divide the satellite image into several parts and to calculate ET separately since the weather of study area need to be relatively uniform for cold and hot pixels selection in the METRIC model assumes (Allen et al., 2014).

The METRIC model is a feasible tool for estimating and mapping ET in temporal-spatial variability. It has been widely used for water consumption from natural vegetation to agricultural crops, determination of crop coefficients, irrigation schedules, and water rights accounting in Idaho, California, and New Mexico (Allen et al., 2007b). Case studies concluded that errors for seasonal ET estimates were less than 4% in the growing seasons and errors for daily ET estimates were about 13%−20% for irrigated meadow and sugar beet crop (Allen et al., 2007b). French et al. (2015) found METRIC model was used to provide good estimates of daily ET over a cotton study site, and they suggested METRIC model was preferred when sufficient ancillary data were not available. He et al. (2017) showed that monthly ET estimates had a 9.68% relative mean difference between METRIC-based versus observed ET over an almond orchard. Ortega-Salazar et al. (2021) found that the estimated errors within the METRIC model for daily ET estimates were only between 4% and 6% of the measurements and the root mean square error (RMSE) was 0.42 mm/day for a drip-irrigated olive orchard.

Sensitivity analysis is very useful to identify the sensitive variables in a model and the suitable study size for improving model accuracy (Zhang et al., 2017, Long et al., 2011). Valipour et al. (2020) analyzed the effect of meteorological variables on the reference ET (ETo). They found that wind speed was the most important variable driving the trends of ETo in most regions. Anwar et al. (2021) examined the sensitivity of the simulated summer potential ET (PET) of tropical Africa to LAI and vegetation-runoff systems using the RegCM4 model. Other studies focused on the sensitivity of actual ET estimates to bias in calibration and LST of METRIC model (Morton et al., 2013, Kilic et al., 2016), selection of cold and hot pixels of the triangle, SEBAL and METRIC models (Long and Singh, 2013), and meteorological factors of Penman method at a region scale (Yang et al., 2019; Biazar et al., 2019). Allen et al. (2013) performed a sensitivity analysis on METRIC model in mountainous areas. They showed that increasing aerodynamic roughness increased convective heat transfer and decreased ET estimation. French et al. (2015) revealed that the sensitivity of METRIC model was related to the season. The relatively great sensitivity showed in the early growth of cotton at a fine spatial resolution.

However, tree crops have complex canopy structure and physiology which cause the energy exchange processes more complicated (Jin et al., 2018). The sensitivity analysis of tree crop ET has received less attention compared to agricultural row crops. We are aware of no studies focusing on the quantifying errors of METRIC-based daily ET estimates over orchards during the entire growth period. Also, it is not clear how different study sizes impact on daily ET estimates in the METRIC model. In this study, we performed a comprehensive sensitivity analysis of tree crop daily ET estimates by using METRIC model on two types of commercial orchards throughout the growth period. The objective was to quantify the sensitivity of daily ET estimates to errors in various meteorological variables and land surface parameters. In addition, appropriate study size could improve the estimated accuracy and it is crucial for correctly selecting cold and hot pixels in the METRIC model. Therefore, we also investigated the impact of different study sizes on daily ET estimates of an almond and three pistachio orchards in the San Joaquin Valley, California, USA.

Section snippets

Study area

This study focused on two types of commercial mature orchards located in south San Joaquin Valley, California, US (Fig. 1), including one almond orchard (He et al., 2017) and three pistachio orchards (Jin et al., 2018). The study site has a typical Meditterranean climate of a hot and dry summer and a mild and wet winter. The almond trees were planted in 1999 with a well-watered management of drip and fanjet during the period 2009–2012. The canopy structure was homogeneous with the summer NDVI

Sensitivity analysis to meteorological variables

The METRIC-based daily ET estimates of both almond and pistachio trees were more sensitive to instantaneous ETo compared to Ta, VP and WS throughout the entire growing season (Fig. 2). Results indicated that an error of 5% in instantaneous ETo can result in a mean increase of 6.80 ± 4.41% (~0.31 ± 0.21 mm/day) in the daily ET of almond trees, corresponding to the changes of 36.39 ± 10.84 W/m2 in H estimates and 0.05 ± 0.03 in EToF. The change of daily ET in the pistachio sites was smaller than

Meteorological variables

It is worth noting that instantaneous ETo was a key variable in the METRIC model since it was used to internally calibrate to compute H and reference ET fraction (EToF) which is similar to the crop coefficient. When errors varying from − 50 to 50% were added to instantaneous ETo, H estimates decreased but EToF increased, thereby resulting in an increase in daily ET of both almond and pistachio trees. In the Fig. 3, the shape of the curves for almond trees was different from that for pistachio

Conclusions

In this study, we quantified the sensitivity of METRIC-based daily ET estimates of almond and pistachio trees to weather variables and land surface parameters, with the introduced errors up to ± 50% (∆=5%) or ± 5 K (∆=0.5 K). We also analyzed the influence of different study sizes on daily ET estimates throughout the entire growing season.

For the sensitivity to meteorological variables, our results showed that instantaneous ETo played an important role in determining METRIC-based daily ET

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This work is supported in part by the National Natural Science Foundation of China under Grant 41971300 and Grant 61901278; in part by the Key Project of Department of Education of Guangdong Province under Grant 2020ZDZX3045; in part by the Guangdong Basic and Applied Basic Research Foundation under Grant 2022A1515011290; in part by the Natural Science Foundation of Guangdong Province under Grant 2021A1515011413; in part by Shenzhen Scientific Research and Development Funding Program under

References (43)

  • R. Qiu et al.

    Evapotranspiration estimation using a modified Priestley-Taylor model in a rice-wheat rotation system

    Agr. Water Manag.

    (2019)
  • G. Rana et al.

    Measurement and modelling of evapotranspiration of irrigated citrus orchard under Mediterranean conditions

    Agr. For. Meteorol.

    (2005)
  • J.M. Sánchez et al.

    Modelling surface energy fluxes over maize using a two-source patch model and radiometric soil and canopy temperature observations

    Remote Sens. Environ.

    (2008)
  • T.M. Shapland et al.

    Thermocouple frequency response compensation leads to convergence of the surface renewal alpha calibration

    Agr. For. Meteorol.

    (2014)
  • T.E. Twine et al.

    Correcting eddy-covariance flux underestimates over a grassland

    Agr. For. Meteorol.

    (2000)
  • T. Wang et al.

    Evaluation of crop coefficient and evapotranspiration data for sugar beets from landsat surface reflectances using micrometeorological measurements and weighing lysimetry

    Agr. Water Manag.

    (2021)
  • L.E. Williams et al.

    Grapevine water use and the crop coefficient are linear functions of the shaded area measured beneath the canopy

    Agr. For. Meteorol.

    (2005)
  • J. Xue et al.

    Evaluation of remote sensing-based evapotranspiration models against surface renewal in almonds, tomatoes and maize

    Agr. Water Manag.

    (2020)
  • Y. Yang et al.

    Sensitivity of potential evapotranspiration to meteorological factors and their elevational gradients in the Qilian Mountains, northwestern China

    J. Hydrol.

    (2019)
  • R.G. Allen et al.

    Lysimeters for Evapotranspiration and Environmental Measurements

    (1991)
  • R.G. Allen et al.

    Satellite-based energy balance for mapping evapotranspiration with internalized calibration (METRIC)—Model

    J. Irrig. Drain. Eng.

    (2007)
  • View full text