A framework for estimating actual evapotranspiration at weather stations without flux observations by combining data from MODIS and flux towers through a machine learning approach

Zhang, Chen; Luo, Geping; Hellwich, Olaf; Chen, Chunbo; Zhang, Wenqiang; Xie, Mingjuan; He, Haijun; Shi, Haiyang; Wang, Yuangang

doi:10.1016/j.jhydrol.2021.127047

Cited by 20 publications

(11 citation statements)

References 83 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…6) appear to be insignificant. This seems to be related to the fact that in large-scale water flux simulations, the sites of similar PFTs are selected such as for modeling multiple forest sites across Europe (Van Wijk and Bouten, 1999) which focus on "forest" and multiple grassland sites across arid northern China (Xie et al, 2021;Zhang et al, 2021) which focus on "grassland", rather than mixing different PFT types to train models as is done in machine learning modeling of carbon fluxes (Zeng et al, 2020). In terms of the timescales of the models, the 4 d, 8 d, and monthly scales appear to correspond to higher accuracy compared to the halfhourly and daily scales.…”

Section: Other Model Featuresmentioning

confidence: 99%

“…Currently, there are three main approaches for simulation and spatial and temporal prediction of ET: (i) physical models based on remote sensing, such as surface energy balance models (Minacapilli et al, 2009;Wagle et al, 2017), the Penman-Monteith equation (Mu et al, 2011;Zhang et al, 2010), and the Priestley-Taylor equation (Miralles et al, 2011); (ii) process-based land surface models, biogeochemical models, and hydrological models (Barman et al, 2014;Pan et al, 2015;Sándor et al, 2016;Chen et al, 2019); and (iii) the observation-based machine learning modeling approach with in situ eddy-covariance (EC) observations of water flux (Jung et al, 2011;Li et al, 2018;Van Wijk and Bouten, 1999;Xie et al, 2021;Xu et al, 2018;Yang et al, 2006;Zhang et al, 2021). For remote-sensing-based physical models and process-based land surface models, some physical processes have not been well characterized due to the lack of understanding of the detailed mechanisms influencing ET under different environmental conditions.…”

mentioning

confidence: 99%

“…For example, the inaccurate representation and estimation of stomatal conductance (Li et al, 2019) and the linearization (McColl, 2020) of the Clausius-Clapeyron relation in the Penman-Monteith equation may introduce both empirical and conceptual errors into estimates of ET. Limited by complicated assumptions and model parametrizations, these process-based models face challenges in the accuracy of their ET estimations over heterogeneous landscapes (Pan et al, 2020;Zhang et al, 2021). Therefore, many researchers have used datadriven approaches for the simulation and prediction of ET with the accumulation of a large volume of measured observational data of water fluxes in the past decades.…”

mentioning

confidence: 99%

“…Currently, the most common ways of model validation include spatial, temporal, and random cross-validation. Spatial validation is mainly to evaluate the ability of the model to be applied in different regions or flux sites with different PFT types, and one of the common methods is "leave one site out" (Fang et al, 2020;Papale et al, 2015;Zhang et al, 2021). If the data of the site left out for validation differ significantly from the distribution of the training dataset, the expected accuracy of the model applied at that site may be low because the trained model may not capture the specific and local relationships between the water flux and the various predictor variables at that site.…”

mentioning

confidence: 99%

See 3 more Smart Citations

Evaluation of water flux predictive models developed using eddy-covariance observations and machine learning: a meta-analysis

Shi

Luo

Hellwich

et al. 2022

Hydrol. Earth Syst. Sci.

Self Cite

View full text Add to dashboard Cite

With the rapid accumulation of water flux observations from global eddy-covariance flux sites, many studies have used data-driven approaches to model water fluxes, with various predictors and machine learning algorithms used. However, it is unclear how various model features affect prediction accuracy. To fill this gap, we evaluated this issue based on records of 139 developed models collected from 32 such studies. Support vector machines (SVMs; average R-squared = 0.82) and RF (random forest; average R-squared = 0.81) outperformed other evaluated algorithms with sufficient sample size in both cross-study and intrastudy (with the same data) comparisons. The average accuracy of the model applied to arid regions is higher than in other climate types. The average accuracy of the model was slightly lower for forest sites (average R-squared = 0.76) than for croplands and grasslands (average R-squared = 0.8 and 0.79) but higher than for shrubland sites (average Rsquared = 0.67). Using R n /R s , precipitation, T a , and the fraction of absorbed photosynthetically active radiation (FA-PAR) improved the model accuracy. The combined use of T a and R n /R s is very effective, especially in forests, while in grasslands the combination of W s and R n /R s is also effective. Random cross-validation showed higher model accuracy than spatial cross-validation and temporal crossvalidation, but spatial cross-validation is more important in spatial extrapolation. The findings of this study are promising to guide future research on such machine-learning-based modeling. IntroductionEvapotranspiration (ET) is one of the most important components of the water cycle in terrestrial ecosystems. It also represents the key variable in linking ecosystem functioning, carbon and climate feedback, agricultural management, and water resources (Fisher et al., 2017). The quantification of ET for regions, continents, or the globe can improve our understanding of water, heat, and carbon interactions, which is important for global change research (Xu et al., 2018). Information on ET has been used in many fields, including, but not limited to, droughts and heat waves (Miralles et al., 2014), regional water balance closures (Chen et al., 2014;Sahoo et al., 2011), agricultural management (Allen et al., 2011, water resources management (Anderson et al., 2012),

show abstract

Section: Other Model Featuresmentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

Evaluation of water flux predictive models developed using eddy-covariance observations and machine learning: a meta-analysis

Shi

Luo

Hellwich

et al. 2022

Hydrol. Earth Syst. Sci.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Low air and soil temperatures, high relative humidity, and moderate wind speed are known to be a favorable condition for dew condensation (Monteith, 1957;Zangvil, 1996). On the other hand, random forest (RF) model has proven to be an effective machine learning model in eco-hydrological variable reconstruction (Xu et al, 2018;Fu et al, 2021;Zhang et al, 2021). However, there are few studies using meteorological variables to reconstruct the dew condensation process over a long period with an RF model.…”

Section: Introductionmentioning

confidence: 99%

Dew amount and its long-term variation in the Kunes River Valley, Northwest China

et al. 2022

View full text Add to dashboard Cite

Dew is an essential water resource for the survival and reproduction of organisms in arid and semi-arid regions. Yet estimating the dew amount and quantifying its long-term variation are challenging. In this study, we elucidate the dew amount and its long-term variation in the Kunes River Valley, Northwest China, based on the measured daily dew amount and reconstructed values (using meteorological data from 1980 to 2021), respectively. Four key results were found: (1) the daily mean dew amount was 0.05 mm during the observation period (4 July-12 August and 13 September-7 October of 2021). In 35 d of the observation period (i.e., 73% of the observation period), the daily dew amount exceeded the threshold (>0.03 mm/d) for microorganisms; (2) air temperature, relative humidity, and wind speed had significant impacts on the daily dew amount based on the relationships between the measured dew amount and meteorological variables; (3) for estimating the daily dew amount, random forest (RF) model outperformed multiple linear regression (MLR) model given its larger R 2 and lower MAE and RMSE; and (4) the dew amount during June-October and in each month did not vary significantly from 1980 to the beginning of the 21 st century. It then significantly decreased for about a decade, after it increased slightly from 2013 to 2021. For the whole meteorological period of 1980-2021, the dew amount decreased significantly during June-October and in July and September, and there was no significant variation in June, August, and October. Variation in the dew amount in the Kunes River Valley was mainly driven by relative humidity. This study illustrates that RF model can be used to reconstruct long-term variation in the dew amount, which provides valuable information for us to better understand the dew amount and its relationship with climate change.

show abstract

Moisture conditions are limiting evapotranspiration changes of Alpine mountains of Qilian Mountains

Wang,

Li,

Xue

et al. 2024

Hydrological Processes

View full text Add to dashboard Cite

Variations in evapotranspiration and their sensitivity to controlling variables are pivotal for comprehending water balance dynamics and climate change, particularly in high‐altitude regions such as the Qilian mountains. Environmental shifts are bound to disrupt local water cycles and balance, with significant implications for these alpine areas. To enhance our understanding of evapotranspiration variability across different altitudes within the Qilian Mountains' high‐elevation region and to assess the model's adaptability and responsiveness to environmental factors, our study involved measuring actual evapotranspiration at three distinct elevations. This was achieved using meteorological stations and continuous data from a weighing‐type microlysimeter at the Shaliu River basin's gradients of 3797, 4250 and 4303 m, spanning the growing seasons from June 2020 to October 2022. We utilized 10 models to calculate the value of reference evapotranspiration, which were then matched against actual evapotranspiration data to identify the most appropriate model. Our research found that across the three elevation gradients, the daily average evapotranspiration were 3.663, 3.845 and 4.317 mm day−1, respectively. Across the three elevations, with consistent intra‐annual fluctuations. Notably, August experienced the highest monthly evapotranspiration at 4.750 mm day−1, and reach peak at 10:00 and 15:00 on the three elevation gradients. The results from the simulation of the 10 models indicate that the Dalton model is more suitable for our study area compared with the other models, showing the best R2, root mean square error and percentage error values. Partial least squares regression analysis, coupled with an enhanced regression tree model, identified precipitation as the most critical factor, with a variable importance in projection score of 2.079, contributing 52.6% to evapotranspiration. Collectively, precipitation were identified as key factors influencing evapotranspiration variability within our research area. Our study's insights are valuable for anticipating the impacts of future climate change. This conclusion is instrumental for refining water budget projections in Alpine regions under climate change scenarios.

show abstract

A framework for estimating actual evapotranspiration at weather stations without flux observations by combining data from MODIS and flux towers through a machine learning approach

Cited by 20 publications

References 83 publications

Evaluation of water flux predictive models developed using eddy-covariance observations and machine learning: a meta-analysis

Evaluation of water flux predictive models developed using eddy-covariance observations and machine learning: a meta-analysis

Dew amount and its long-term variation in the Kunes River Valley, Northwest China

Moisture conditions are limiting evapotranspiration changes of Alpine mountains of Qilian Mountains

Contact Info

Product

Resources

About