雪花亮晶晶的
留下了我踩过的痕迹
回到温暖的宿舍
Long-term monitoring of evapotranspiration using the SEBAL algorithm and Google Earth Engine cloud computing
蒸散发计算是目前较为热门的方向，特别是大尺度的蒸散发计算与分析。遥感云计算在大面积估算中确实有其强大的功能，这篇文章发表于遥感领域顶刊ISPRS，可以作为一个重要参考，特别是该文章提供了详细的代码，包括JS版本和Python版本。
Long-term monitoring of evapotranspiration using the SEBAL algorithm and Google Earth Engine cloud computing
Abstract：Accurate estimation of evapotranspiration (ET) is essential for several applications in water resources management. ET models using remote sensing data have flourished in recent years allowing spatial and temporal assessments at unprecedented resolutions. This study presents geeSEBAL, a new tool for automated estimation of
ET, based on the Surface Energy Balance Algorithm for Land (SEBAL) and a simplified version of the CIMEC (Calibration using Inverse Modeling at Extreme Conditions) process for the endmembers selection, developed
within the Google Earth Engine (GEE) environment. The tool framework is introduced, and case studies across
multiple biomes in Brazil are presented by comparing daily ET estimates with eddy covariance (EC) data from 10
flux towers. Based on 224 Landsat images using ERA5 Land as meteorological inputs, daily ET estimates of geeSEBAL yielded an average root mean squared difference (RMSD) of 0.67 mm day− 1 when compared to EC
data corrected for the energy balance closure. Additional analyses indicate a low geeSEBAL sensitivity to
meteorological inputs, yielding an average RMSD of 0.71 mm day− 1 when driven by in situ meteorological
measurements. On the other hand, we found a higher sensitivity of the automated CIMEC algorithm to the selection of endmembers for internal calibration. For instance, by adjusting the endmembers percentiles to tropical
biomes we found an error that was 36% lower compared to the standard CIMEC percentiles. Finally, we assessed
the long-term effects (1984–2020) of land cover changes on surface energy fluxes and water use in agriculture for
key areas in Brazil, from deforested areas in the Amazon to irrigated crops in the Pampas and Cerrado biomes. A
comparison with a land surface temperature-based (SSEBop) and a vegetation-based (MOD16) model was also
performed to assess relative advantages and disadvantages. This analysis showed that geeSEBAL has a significant
potential for long-term assessment of ET in data-scarce areas, due to its lower sensitivity to meteorological inputs. geeSEBAL codes are written in Python and JavaScript and are freely available on GitHub (https://github.com/etbrasil/geesebal). geeSEBAL also includes a graphical user interface (https://etbrasil.org/geesebal),
allowing important advances in water resources management at regional scales.
Keywords:
Cloud computation;
ERA5 land;
geeSEBAL;
Google earth engine;
Landsat; Meteorological reanalysis 气象再分析 关键词3-5个，从题目摘要中来
摘要：【背景】准确估算蒸散发对于水资源管理中的一些应用至关重要 。使用遥感数据的 ET 模型近年来蓬勃发展，使空间和时间评估具有前所未有的分辨率。【方法】本研究介绍了 geeSEBAL，一种基于陆地表面能量平衡算法(SEBAL)的 ET 自动估计新工具，以及在 Google Earth Engine (GEE)环境中开发的用于端部选择的 CIMEC (极端条件下反向建模校准)简化版本。介绍了该工具的框架，并通过对10个通量塔的每日 ET 估计值与涡度相关(EC)数据的比较，介绍了巴西多个生物群落的案例研究。【结论】根据使用 ERA5大陆作为气象输入的224张陆地卫星图像，与经能量平衡闭合校正的欧共体数据相比，geeSEBAL 的每日 ET 估计值产生了0.67 mm 的平均均方根差(RMSD)。其他分析表明，geeSEBAL 对气象输入的敏感度较低，在现场气象测量的驱动下，平均 RMSD 为0.71毫米/天。另一方面，我们发现自动 CIMEC 算法对于内部校准端元的选择具有较高的灵敏度。例如，通过将端元百分位数调整到热带生物群落，我们发现与 CIMEC 标准百分位数相比，误差降低了36% 。最后，我们评估了土地覆盖变化对巴西关键地区地表能量通量和农业用水的长期影响(1984-2020年) ，这些地区包括亚马逊森林砍伐区和潘帕斯和塞拉多生物群落的灌溉作物。并与基于地表温度的(SSEBop)和基于植被的(MOD16)模型进行了比较，以评估相对优势和劣势。【展望】这一分析表明，geeSEBAL 对于数据匮乏地区的 ET 长期评估具有重大潜力，因为它对气象输入的敏感性较低。geeSEBAL 代码是用 Python 和 JavaScript 编写的，可以在 GitHub 上免费获得。geeSEBAL 还包括一个水资源管理图形用户界面，使水资源管理在区域规模上取得重要进展。
1.Introduction
Evapotranspiration (ET) is the process by which water vapor is
transferred from the surface to the atmosphere by plant transpiration
and evaporation from wet canopy, soil, rivers, lakes and wetlands. ET
has a major role in the terrestrial climate system, since it connects the
energy, water and carbon cycles (Anderson et al., 2012; Fisher et al.,
2017; Zhang et al., 2016). It also essential for water resources management, supporting decision-makers in efforts to improve water
security (Allen et al., 2011) and to achieve a better allocation of water
use in agricultural crops (Bastiaanssen et al., 1999; Biggs et al., 2015).
蒸散发是植物蒸腾和蒸发从湿林冠、土壤、河流、湖泊和湿地将水蒸气从地表转移到大气中的过程。ET 在陆地气候系统中扮演着重要角色，因为它连接了能量、水和碳循环(Anderson 等人，2012; Fisher 等人，2017; Zhang 等人，2016)。它还对水资源管理至关重要，支持决策者努力改善水安全(Allen 等人，2011年) ，并实现农作物用水的更好分配(bastianssen 等人，1999年; Biggs 等人，2015年)。
For large areas or over heterogeneous surfaces, ET estimation poses a
great challenge since current ground measurements networks are not
able to provide spatial trends at large-scales (Biggs et al., 2015; McShane
et al., 2017). Over the last two decades, remote sensing-based ET models
have been an alternative to estimate ET for multiple temporal and spatial scales with consistent estimates at pixel and image scales
(Anderson et al., 2011; Ershadi et al., 2013; McCabe and Wood, 2006).
These models can be separated into two main groups: (1) vegetation
(VI)-based models, including the Moderate Resolution Imaging Spectroradiometer (MODIS) Land Surface Evapotranspiration (MOD16) (Mu
et al., 2011), the Global Land-Surface Evaporation Amsterdam Methodology (GLEAM) (Martens et al., 2017), the Priestley-Taylor Jet Propulsion Laboratory (PT-JPL) (Fisher et al., 2008), and the Satellite
Irrigation Management Support (SIMS) system (Melton et al., 2012;
Pereira et al., 2020), and (2) land surface temperature (Ts)-based
models, including the Surface Energy Balance Algorithm for Land model
(SEBAL) (Bastiaanssen et al., 1998a), Mapping Evapotranspiration at
High Resolution with Internalized Calibration (METRIC) (Allen et al.,
2007), Atmosphere-Land Exchange Inverse (ALEXI) and its associated
flux disaggregation technique (DisALEXI) (Anderson et al., 2011),
Simplified Surface Energy Balance (SSEBop) (Senay et al., 2013), and
Surface Energy Balance System (SEBS) (Su, 1988).
由于目前的地面测量网络不能提供大尺度的空间趋势，对于大面积或不均匀地表，ET 估算是一个巨大的挑战。在过去的二十年中，基于遥感的 ET 模型已经成为一种在像素和图像尺度上一致估计多种时间和空间尺度的 ET 的替代方法(Anderson 等人，2011; Ershadi 等人，2013; McCabe 和 Wood，2006)。这些模型可以分为两大类: (1)基于植被(VI)的模型，包括中分辨率成像光谱仪(MODIS)陆地表面蒸发散(MOD16)(Mu 等人，2011) ,全球陆地表面蒸发阿姆斯特丹方法(GLEAM)(Martens et al. ，2017) ，Priestley-Taylor 喷气推进实验室(pt-taylor 喷气推进实验室)(Fisher et al. ，2008) ，卫星灌溉管理支持系统(SIMS)(Melton et al. ，2012; Pereira et al. ，2020) ，和(2)基于陆地表面温度(Ts)的模型,包括陆地模式的表面能量平衡算法(SEBAL)(bastianssen 等人，1998年 a) ，内部校准的高分辨率蒸发散映射(度量)(Allen 等人，2007年) ,大气-陆地交换反演(ALEXI)及其相关通量解聚技术(DisALEXI)(Anderson 等人，2011) ，简化表面能量平衡(SSEBop)(Senay 等人，2013) ，以及表面能量平衡系统(SEBS)(Su，1988)。
The choice of an ET model depends on several factors, including type
of application, spatial and temporal resolutions, remote sensing and
meteorological inputs, and the expected model advantages and limitations (Zhang et al., 2016). VI models rely on vegetation greenness
indices (such as the Normalized Difference Vegetation Index (NDVI) or
leaf area index (LAI)) and meteorological inputs (mainly net radiation
(Rn), vapor pressure deficit (VPD) and air temperature (Tair)) to estimate
ET. For instance, MOD16 is a widely used VI − based model due to its
global coverage, which allows several large-scale applications with
moderate accuracy (Biggs et al., 2016; Kiptala et al., 2013; Yang et al.,
2013). However, several studies demonstrated that it has limitations to
estimate ET in areas with high soil moisture availability (Bhattarai et al.,
2017; Biggs et al., 2016; Souza et al., 2019), since the main proxy for
water availability is VPD and this inaccuracy can be related to uncertainties in global meteorological reanalysis (Li et al., 2019; Yang
et al., 2015) and land cover parameterization (Ruhoff et al., 2012). On
the other hand, Ts models are often used to support water resources
applications at multiple scales, since Ts serves as an effective proxy for
soil moisture (Anderson et al., 2011). The different Ts models have
divergent sensitivity to meteorological inputs. For instance, METRIC and
SSEBop need high quality hourly and daily meteorological data (Allen
et al., 2011; Bhattarai et al., 2016; Tasumi et al., 2005), which strongly
influences the estimation of reference ET (ETr), usually computed with
the Penman-Monteith equation (Allen et al., 1998), for upscaling
instantaneous fluxes to daily ET. SEBAL, on the other hand, has a lower
sensitivity to meteorological inputs and a higher sensitivity to Ts and the
near-surface temperature gradient (dT) (Laipelt et al., 2020; Long et al.,
2011). While some geographic regions operate extensive monitoring
networks that provide high-quality meteorological data (for instance,
North America and Europe), several areas worldwide are data-scarce,
with low density of ground measurements (Fick and Hijmans, 2017;
Menne et al., 2012) to enable accurate ET estimates.
ET 模型的选择取决于几个因素，包括应用类型、时空分辨率、遥感和气象投入，以及预期模型的优点和局限性(Zhang ET al. ，2016)。VI 模型依赖于植被绿度指数(如常态化差值植生指标指数(NDVI)或叶面积指数(LAI))和气象输入(主要是净辐射(Rn)、蒸汽压亏缺(VPD)和气温(Tair))来估算 ET。例如，MOD16是一个广泛使用的基于虚拟仪器的模型，由于它的全球覆盖范围，允许一些大规模的应用程序具有中等的准确性(Biggs 等人，2016; Kiptala 等人，2013; Yang 等人，2013)。然而，一些研究表明，它在估算高土壤湿度地区的 ET 方面有局限性(Bhattarai ET al. ，2017; Biggs ET al. ，2016; Souza ET al. ，2019) ，因为水可利用性的主要代理是 VPD，这种不准确性可能与全球气象再分析的不确定性有关(Li ET al. ，2019; Yang ET al. ，2015)和土地覆盖参量化(Ruhoff ET al. ，2012)。另一方面，Ts 模型经常用于支持水资源应用在多种尺度，因为 Ts 作为一个有效的代理土壤湿度(Anderson 等人，2011年)。不同的 Ts 模式对气象输入具有不同的敏感性。例如，METRIC 和 SSEBop 需要高质量的每小时和每日气象数据(Allen 等人，2011; Bhattarai 等人，2016; Tasumi 等人，2005) ，这强烈影响参考 ET (ETr)的估算，通常用 Penman-Monteith 方程(Allen 等人，1998)计算每日 ET 的瞬时通量。另一方面，SEBAL 对气象输入的敏感性较低，对 Ts 和近地表温度梯度的敏感性较高(Laipelt et al. ，2020; Long et al. ，2011)。虽然一些地理区域运行着广泛的监测网络，提供高质量的气象数据(例如，北美和欧洲) ，但全世界的一些地区数据匮乏，地面测量密度低(Fick 和 Hijmans，2017; Menne 等人，2012) ，以便能够准确估计ET。
Among several Ts-based models, SEBAL has been successfully
applied worldwide across different climates and land cover conditions
for water resources management, yielding consistent and accurate results (Bastiaanssen et al., 2005; Bhattarai et al., 2012; Ruhoff et al.,
2012; Tang et al., 2013; Wagle et al., 2017; Yang et al., 2012). SEBAL
estimates ET as the residual of the surface energy balance. The main
component of the algorithm is the internal process to estimate dT, which
is based on the selection of endmembers that represent the extremes of
the wet (cold) and dry (hot) ET spectrum (Allen et al., 2007; Bastiaanssen et al., 1998a). While this was manually performed in early
applications, an algorithm based on Ts and NDVI percentiles using the
Calibration using Inverse Modeling at Extreme Conditions (CIMEC) has
allowed the automatic selection of endmember candidates (Allen et al.,
2013). This algorithm enabled the use of multiple images within a fully
automated ET framework, improving and facilitating the assessment of
time series based on remote sensing images (Bhattarai et al., 2017;
Dhungel and Barber, 2018; Jaafar and Ahmad, 2019).
在几个基于 ts 的模型中，SEBAL 已经成功地在全球范围内应用于不同气候和土地覆盖条件下的水资源管理，产生了一致和准确的结果(bastianssen et al. ，2005; Bhattarai et al. ，2012; Ruhoff et al. ，2012; Tang et al. ，2013; Wagle et al. ，2017; Yang et al. ，2012)。SEBAL 估计 ET 是表面能量平衡的残余。该算法的主要组成部分是估计 dT 的内部过程，这是基于选择代表湿(冷)和干(热) ET 光谱极端值的端元(Allen 等人，2007; bastianssen 等人，1998a)。虽然这是在早期应用中人工执行的，但是一种基于 t 和 NDVI 百分比的算法，使用了在极端条件下使用反向建模的校准(CIMEC) ，允许自动选择端元候选人(Allen 等人，2013年)。该算法使得在一个完全自动化的 ET 框架内使用多幅图像成为可能，从而改进和便利了基于遥感图像的时间序列评估(Bhattarai 等人，2017; Dhungel and Barber，2018; Jaafar 和 Ahmad，2019)。
The global availability of state-of-the-art global reanalysis dataset for
land applications (Munoz-Sabater ˜ et al., 2021) further allows the use of
surface energy balance algorithms like SEBAL for long-term applications
without significantly decreasing its accuracy (Laipelt et al., 2020).
Considering the importance of accurate ET estimates worldwide for
water resources management and the critical need for large-scale
availability of remote sensing and reanalysis data, the main goals of
this study are to (1) describe an open-source SEBAL framework implemented within the Google Earth Engine (GEE) application Programming
Interface (API), hereafter named geeSEBAL, to estimate long-term ET;
(2) validate geeSEBAL estimates using eddy covariance (EC) data from
multiple biomes and land cover conditions in Brazil; (3) compare
geeSEBAL estimates against other well validated Ts (SSEBop) and
VI-based (MOD16) ET models; and (4) investigate long-term ET changes
in key areas in Brazil, which have undergone major land cover changes
over the last decades, from deforested areas in the Amazon to irrigated
crops in the Pampas and Cerrado biomes. We also assessed the sensitivity of Ts-based models to meteorological inputs, when driven by
global reanalysis and ground measurements from the flux towers, to
evaluate residual errors attributed to meteorological inputs and model
structure.
最先进的全球土地应用再分析数据集的全球可用性(Munoz-Sabater 等人，2021年)进一步允许在长期应用中使用 SEBAL 等地表能量平衡算法，而不会显著降低其准确性(Laipelt 等人，2020年)。考虑到全世界对水资源管理进行准确的 ET 估计的重要性以及对大规模提供遥感和重新分析数据的迫切需要，本研究的主要目标是(1)描述在 Google Earth Engine (GEE)应用程序编程接口(下称 geeSEBAL)内实现的开放源码 SEBAL 框架，以估计长期 ET;(2)利用来自巴西多个生物群落和土地覆盖条件的涡度相关(EC)数据验证 geeSEBAL 估计值; (3)将 geeSEBAL 估计值与其他经过充分验证的 Ts (SSEBop)和基于 MODIS16的 ET 模型进行比较;(4)调查过去几十年来巴西主要地区土地覆盖发生重大变化的地区，从亚马逊森林砍伐地区到潘帕斯和塞拉多生物群落的灌溉农作物的长期 ET 变化。我们还评估了在全球重新分析和来自通量塔的地面测量驱动下，基于 Ts 的模式对气象输入的敏感性，以评估气象输入和模式结构造成的残余误差。
Following national and international initiatives to use large-scale ET
estimates for water resources management, such as the OpenET
(https://openetdata.org) and FAO Wapor (https://wapor.apps.fao.org/)
efforts, we used current (Landsat 8) and historical (Landsat 5 and 7)
images at 30 × 30 m and the state-of-the-art meteorological reanalysis
from ERA5 to estimate ET time series since 1984. We expect that opensource and operational initiatives will improve water resources management in sustainable ways, enabling decision makers and water
managers to benefit from accurate information that stems from the access to ET data.
根据国家和国际上对水资源管理使用大规模 ET 估算的倡议，例如 OpenET ( https://openetdata.org )和 FAO Wapor ( https://Wapor.apps.FAO.org/)的努力，我们使用当前的(Landsat 8)和历史的(Landsat 5和7)图像在30 × 30米处，并使用 ERA5最先进的气象重新分析来估算1984年以来 ET 时间序列。我们期望开放源代码和业务举措将以可持续的方式改善水资源管理，使决策者和水资源管理者能够从获取环境技术数据所产生的准确信息中受益。
2. Methods
3. Results
Fig. 7. Spatial patterns of four-month average evapotranspiration estimates derived from geeSEBAL, SSEBop and MOD16 models for four locations in Brazil:
(a)
Cerrado-Bananal, a floodplain in a transitional area between the Amazon and Cerrado biomes (estimates from May to September of 2018);
(b) Amazon-Rondonia, a
tropical forest subject to deforestation (estimates from May to September of 2018);
(c) Cerrado-Minas, an area with intense irrigated (center pivots) cropland
expansion (estimates from June to October of 2019);
(d) Pampa-Santa Maria, a natural grassland area with historical irrigated (flooded) croplands (estimates from
January to April of 2009).
5. Conclusions
In this study, the SEBAL algorithm for ET estimation was implemented within the GEE environment (named as geeSEBAL), which enables a fast and accurate way to estimate ET at regional scales using
Landsat images. geeSEBAL is available as open-source software implemented in Python and JavaScript with a user-friendly GUI, facilitating
worldwide applications. The comparison between EC data and geeSEBAL estimates in Brazil demonstrated its accuracy over different biomes
and land cover conditions, yielding consistent estimates with an average
RMSD of 0.67 mm day− 1 and R2 values higher than 0.4 for most of the
assessed flux tower sites. Furthermore, the comparison with other Ts and
VI-based models (SSEBop and MOD16, respectively) showed that while
geeSEBAL uses primarily reanalysis data as meteorological input data,
ET estimates are similar to estimates when the algorithm is driven by
ground measurements of meteorological conditions. This indicates that
geeSEBAL can provide data with reasonably accuracy for worldwide
applications over data scarce areas. SSEBop also demonstrated its potential to estimate ET at regional scales, however with slightly lower
accuracy than geeSEBAL and a higher sensitivity to meteorological
forcing data and requiring accurate meteorological input data to improve its accuracy. On the other hand, the use of large-scale ET
models (such as MOD16) may introduce some systematic biases for
different vegetation types and heterogeneous landscapes. Finally, the
long-term ET assessment in different Brazilian biomes showed the potential for geeSEBAL to support improved understanding of the impacts
of land cover changes on ET over recent decades. The understanding of
ET dynamics is imperative to mitigate impacts of freshwater depletion in
ecosystems due to increasing water demand for food production and
water supply. Given that geeSEBAL does not require any ground measurement as input, we anticipate that this tool may be useful for regional
and large-scale studies on water and energy balances, as well as for
water resources management in data scarce areas worldwide.