HomeEducationDoctorate (PhD & EngD)For current candidatesPhD infoUpcoming public defencesPhD Defence Lilin Zhang | Enhancing regional estimates of evapotranspiration with Earth observation data

PhD Defence Lilin Zhang | Enhancing regional estimates of evapotranspiration with Earth observation data

Enhancing regional estimates of evapotranspiration with Earth observation data

The PhD defence of Lilin zhang will take place in the Waaier building of the University of Twente and can be followed by a live stream.
Live Stream

Lilin Zhang is a PhD student in the department of Natural Resources. (Co)Promotors are prof. A.D. Nelson, dr. M.T. Marshall, and dr.ir. A. Vrieling from the faculty of Geo-Information Science and Earth Observation (ITC).

Food security and food sustainability are high on the global policy agenda. Reliable information on crop water use and terrestrial water stress are important to ensure an optimal use of available water resources and for enhancing crop production. Remote sensing provides a feasible avenue to estimate regional evapotranspiration (ET), which can be employed to assess terrestrial water stress. However, the heterogeneity of land surfaces and the accumulated errors from various inputs often result in substantial biases in most global or regional ET models across different landscapes. Reducing uncertainties in available ET products or remote sensing (RS)-based models and obtaining regional ET estimates with improved accuracy is important for effectively using ET to support agricultural monitoring and water resources managements.

This thesis first compared different Priestly-Taylor (PT)-based methods that use three Earth observation-based alternatives - apparent thermal inertia (ATI), microwave soil moisture (SM), and optical spectral indices based on shortwave infrared (SWIR) to assess soil evaporation over cropland and grassland regions. Using FLUXNET data as ET reference, the results illustrated that the incorporation of the SWIR-based soil moisture divergence index (SMDI) and microwave-based SM into monthly soil evaporation led to 6% and 5% increase in explained ET variances and reduced RMSE by 23.2% and 13.1% for cropland and grassland, respectively, as compared to PT-JPL using atmospheric reanalysis data only. The results suggested that a combination of optical SWIR and microwave SM has good potential to improve the PT-JPL model accuracy for agricultural landscapes.

Based on the performance of different PT-based methods, ET estimates derived from the revised PT method were used to assess water budgets across 53 catchments in central-western Europe with a humid climate and were compared with three additional ET data sources (MOD16, GLEAM, and PT-JPL). Surprisingly, all RS-based ET estimates significantly diverged from water balance-based ET (ETWB) in 45 humid catchments, whereas most previous studies that focussed on arid catchments or on the global scale found significantly less divergence. Using ET retrievals from the Budyko framework and upscaled ET from FLUXCOM as references, the closure errors of water budgets were sensitive to errors arising from precipitation data in humid regions and the water balance approach was found to overestimate ET during heavy rainfall events. Instead, the Budyko framework that describes the partitioning of precipitation to ET as a functional balance between atmospheric water supply (precipitation, P) and demand (potential evapotranspiration, PET) had good correlation with ET ensemble from multiple data sources and upscaled ET from FLUXCOM product. The results indicated that errors from precipitation and terrestrial water storage anomalies introduce large uncertainties in ETWB, thereby complicating water balance validation in humid regions across multiple timesteps. To improve the application of ETWB for benchmarking ETEB in humid regions, high-quality input data should be used or – like the Budyko framework – energy constraints should be considered.

The thesis then proceeds to explore causes for the notable deviations between observed and Budyko-predicted water balances in certain catchments. The results revealed that for humid catchments, topography and seasonal cumulative moisture surplus can explain the spatial distributions of Budyko scatter with r higher than 0.65, whereas soil properties and vegetation indices explained little of the variance (r≤0.30). Temporally, the interannual variability of Budyko scatter was negatively correlated with annual average vegetation indices, particularly for catchments with relatively low vegetation cover. This thesis provides valuable insights to the interpretation of the Budyko framework and offers possible solutions to improve its performance to predict the spatio-temporal variability of water balances.

Lastly, to address the deviations from the predictive Budyko curve, additional controls of hydrological partitioning were introduced to correct Budyko scatter between catchments and between years. The results illustrated that the use of catchment climatic seasonality properties and topography attributes is effective in reproducing the Budyko parameter (w) with an r of 0.76 and RMSE of 0.49 for all 45 catchments in central-western Europe. After the correction of temporal Budyko scatter using interannual variability of vegetation information and the fraction of precipitation falling as snow, the performance of the modified Budyko-type equation improves with respect to the original Budyko framework, in comparison to ETWB at catchment scale (∆r of 0.26 and ∆RMSE of 19.19 mm/yr). When compared with the gridded ET ensemble using energy balance, the enhanced Budyko framework is generally effective to reproduce the spatial distribution of ET with good similarity, even in ungauged regions. Overall, the revised Budyko framework shows improved performance in predicting water balances and can be applied to assess crop water use and terrestrial water stress at regional scale, particularly in ungauged areas.