TL;DR
This review paper provides a comprehensive overview of current methodologies for estimating regional evapotranspiration (ET) using remotely sensed data. It highlights the diverse approaches, from simplified empirical models to complex physically-based energy balance models, and discusses their inputs, assumptions, advantages, and drawbacks, offering a valuable resource for researchers and practitioners in hydrology, meteorology, and agriculture.
What Problem Does This Paper Address?
Accurate estimation of evapotranspiration (ET) is crucial for understanding water cycles, managing water resources, and modeling climate change. However, directly measuring ET over large, heterogeneous land areas is challenging and often impractical. While conventional ground-based methods provide accurate site-specific data, they are difficult to scale up. This paper addresses the need for reliable regional ET estimation by reviewing and comparing various remote sensing-based methodologies that overcome the limitations of point measurements.
Methods
The paper categorizes remote sensing-based ET models into two main types:
- (Semi-)Empirical Methods: These methods rely on empirical relationships and primarily use data derived from remote sensing observations with minimal ground-based measurements. A key example discussed is the Simplified Empirical Regression Method, which directly relates daily ET to the difference between instantaneous surface temperature (Ts) and air temperature (Ta), often using a regression equation.
- Analytical Methods: These methods involve establishing the physical processes at the scale of interest with varying complexity. They require a variety of direct and indirect measurements from remote sensing and ground-based instruments. This category includes:
- Residual Method of Surface Energy Balance: This approach uses the surface energy balance equation (LE = Rn - G - H) to estimate ET. It involves estimating net radiation (Rn), soil heat flux (G), and sensible heat flux (H) using a combination of remote sensing data (e.g., surface temperature, albedo, NDVI) and ground-based meteorological variables.
- Single-Source Models: These models treat the surface as a single layer. Examples include SEBI, SEBAL, SEBS, and METRIC.
- Dual-Source Models: These models separately model the heat and water exchange between soil and atmosphere and between vegetation and atmosphere, offering a more detailed approach for partially vegetated areas.
- VI-Ts Triangle/Trapezoidal Feature Space: This method utilizes the relationship between vegetation index (VI) and land surface temperature (Ts) to infer ET. It establishes relationships based on the "dry edge" (limited water availability) and "wet edge" (potential ET) within the feature space.
- Data Assimilation: This technique combines remote sensing data with numerical models to provide temporally continuous ET estimates, overcoming the instantaneous nature of satellite observations.
The paper also discusses methods for temporal scaling, converting instantaneous ET estimates to daily values, primarily through sine function approximations or constant evaporative fraction assumptions.
Datasets
Not explicitly detailed as training or evaluation datasets in the context of a specific model development. The paper discusses the use of various remotely sensed data sources, including:
- Multi-spectral data (visible to thermal infrared bands)
- Surface temperature (Ts)
- Vegetation indices (e.g., NDVI, SAVI)
- Surface albedo
- Emissivity
- Leaf Area Index (LAI)
- Fractional vegetation cover
Ground-based meteorological data such as air temperature, wind speed, humidity, and net radiation are also highlighted as crucial inputs for many models.
Key Insights and Ideas
- Remote Sensing as a Necessity: For regional ET estimation, remote sensing is recognized as the "only viable means" due to its ability to provide large, continuous spatial coverage economically and in areas where ground-based measurements are difficult or impossible.
- Trade-offs in Model Complexity: There's a clear trade-off between model simplicity (e.g., empirical methods) and accuracy/physical basis (e.g., physically-based energy balance models). Simpler models are easier to apply but may be less accurate or site-specific, while complex models require more data and computational resources.
- The Importance of Surface Temperature: Land surface temperature (Ts) is a direct indicator of energy and water availability and is a critical input for most remote sensing ET models. However, its accurate retrieval and interpretation are subject to atmospheric effects and viewing angle variations.
- The Challenge of Scaling: A fundamental challenge is bridging the gap between point-scale ground measurements and regional/global scale ET estimates. Temporal scaling (instantaneous to daily) and spatial scaling (local to regional) are critical but remain areas of active research.
- Data Assimilation as a Promising Avenue: Data assimilation techniques offer a powerful way to integrate diverse data sources (remote sensing, ground-based, model outputs) to overcome limitations like temporal discontinuity and improve ET estimates.
Key Results
The paper does not present new experimental results but rather synthesizes the findings of numerous studies. Key qualitative findings include:
- Simplified empirical methods are convenient but often require site-specific parameterization, limiting their broad applicability.
- Single-source energy balance models can be accurate over homogeneous areas but struggle with partially vegetated surfaces, where dual-source models are preferred.
- The VI-Ts triangle/trapezoid methods offer a way to estimate ET without extensive ground data, but their accuracy depends on the presence of clear dry and wet edges in the imagery.
- Dual-source models are more robust for diverse landscapes and require fewer ground-based measurements but can be computationally intensive and may require geo-stationary satellite data.
- Data assimilation can produce continuous ET estimates and merge information from various sources, but it is computationally demanding.
- The accuracy of ET estimates is significantly influenced by the accuracy of retrieved land surface variables (e.g., Ts, albedo, NDVI) and the handling of atmospheric effects.
Limitations and Future Work
Limitations:
- Site-Specificity: Many models, especially empirical ones, require site-specific parameterization, limiting their generalizability.
- Atmospheric Effects: Retrieving accurate surface variables from satellite data is challenging due to atmospheric interference, requiring corrections that can introduce errors.
- Spatial and Temporal Scaling: Bridging the gap between point measurements and regional/global scales, and converting instantaneous ET to daily/longer-term values, remains a significant challenge.
- Data Availability and Quality: The accuracy of ET estimates is limited by the quality and availability of remotely sensed data (e.g., cloud cover, spatial resolution) and ground-based meteorological data.
- Heterogeneous Surfaces: Accurately modeling ET over complex, heterogeneous landscapes is difficult for many models.
- Validation: Validating ET estimates at regional scales is problematic due to scaling effects and the difficulty of obtaining representative ground truth.
Future Work:
- New Parameterizations: Developing new parameterizations of land surface ET using only land surface variables and parameters directly or indirectly derived from satellite data.
- Improved Data Assimilation: Further development of data assimilation techniques to integrate satellite data with land surface models for continuous ET estimation.
- Modeling Land Surface Processes: Formulating models that describe soil-biosphere-atmosphere exchanges in terms of macroscopic parameters with significant physical meaning at regional scales and are measurable from satellites.
- Dialectical Modeling Approaches: Developing integrating and "autonomous" methods to model spatial-temporal variations of land surface processes at various scales.
- Phenomenological Analysis: Studying phenomenological relations between surface variables and spatial indicators to reveal new parameters and establish laws governing land surface processes.
- Reformulation of Energy Balance: Reformulating the energy balance at large scales by introducing "integrating parameters" and parameterizing diurnal variations of surface temperature.
- In-depth Research on Advection: Investigating the impact of advection on regional ET estimates, particularly over urban areas and deserts.
- Improved Accuracy of Retrieved Variables: Enhancing the accuracy of retrieved land surface variables (e.g., Ts, LAI, albedo) and their physical interpretation.
- Fusion of Multi-scale Data: Developing methodologies to calibrate ET simulated by land surface models with remote sensing ET values, utilizing multi-spectral, multi-temporal, and multi-spatial satellite technology.
- Validation at Pixel Scale: Developing robust validation techniques for ET and land surface variables at the satellite pixel scale.
Summary updated