Crop Water Use Estimation Using Remote Sensing and Google Earth Engine
The ability to reliably estimate crop water use i.e., actual evapotranspiration (ETa) is imperative for developing long-term water management strategies in water-stressed regions. Remote sensing (RS) techniques have proven to have valuable contributions for retrieving distributed ETa information at...
Saved in:
| Main Author: | |
|---|---|
| Contributors: | |
| Format: | Doctoral Thesis |
| Language: | English |
| Published: |
Philipps-Universität Marburg
2023
|
| Subjects: | |
| Online Access: | PDF Full Text |
| Tags: |
No Tags, Be the first to tag this record!
|
| Summary: | The ability to reliably estimate crop water use i.e., actual evapotranspiration (ETa) is imperative for developing long-term water management strategies in water-stressed regions. Remote sensing (RS) techniques have proven to have valuable contributions for retrieving distributed ETa information at diverse spatial and temporal scales, particularly, in regions where in situ ETa data is scarce. There are two main RS-based approaches of ETa estimation: (a) surface energy balance methods (SEBs) and (b) vegetation-index-based methods (ET-VIs). SEBs require large computational effort while VI-based approaches are considered to be fast to apply. Although many VI-based algorithms have been developed for specific irrigation districts, no single ET-VI method has been extensively tested and evaluated across various croplands and crop types. The main goal of this Ph.D. thesis was to develop simple and transferable ET-VI-based models for estimating ETa at the field scale across croplands in different regions, utilizing cloud computing and Landsat imagery. Specifically, this thesis focused on the application of distinct VIs, such as the Normalized Difference Vegetation Index (NDVI), the Enhanced Vegetation Index (EVI), and the two-band Enhanced Vegetation Index (EVI2) in the Google Earth Engine (GEE) cloud computing platform for estimating ETa across croplands. The main objectives of this research were: (1) to modify ET-VI methods developed originally for riparian vegetation to croplands through cross-sensor translation; (2) to select the best ET-VI to fit the purpose by comparing different ET-VI-based methods in terms of transferability; (3) to compare the selected ET-VI-based method with the well-established RS-based products (OpenET) and in situ data at the regional scale; (4) to validate the selected ET-VI-based method against flux towers measurements and compare its performance against other RS-based products at various (experimental) sites worldwide. The first two objectives of this thesis focused on developing different ET-VIs over croplands in the Zayandehrud River Basin (ZRB), Iran. The first objective aimed to sensor-transform, model, and map ETa using EVI and EVI2, and evaluate the impacts of dynamic harvested area changes on ETa. Since EVI and EVI2 were developed for the MODIS sensor, using these VIs with Landsat sensors required a cross-sensor transformation to allow for their use in the ET-VI algorithm. Compared to crop ET values, ETa predictions from continuity-corrected Landsat-EVI2 followed by Landsat-EVI performed slightly better across croplands than those of Landsat-EVI and -EVI2 without transformation, showing the necessity of cross-sensor transformation before calculating ETa. After learning about the necessity of cross-sensor translation when using different sensors and the importance of the dynamic nature of plantation area and its significant impact on ETa estimation, three crop-independent NDVI-based approaches (ET-NDVI*, ET-NDVI*scaled, ET-NDVIKc) were employed to calculate ETa and compared to that of selected ET-VI (ET-EVI2) while considering ease of transferability of these methods. NDVI* was applied over croplands in a large irrigation district for the first time. NDVI-based approaches were assessed in terms of their applicability and translation to other regions. NDVI* was scaled (NDVI*scaled) to match the Kc range. Comparing three ET-NDVIs and ET-EVI2 showed that annual ET-EVI2 and ET-NDVI*scaled estimations were close. ET-NDVI* and ET-NDVI*scaled are scene-dependent and require more parameterization and localization; this makes them less user-friendly for croplands’ ETa estimation. The finding that ET-EVI2 can be used in many regions raised the question of how this model performs compared to other well-established RS-based ETa products. In the first attempt, ET-EVI2 was modified (METEVI2) and tested for mapping croplands’ water use dynamics in the Lower Colorado River Basin and compared with seven well-established RS-based products of OpenET and in situ data. Results confirmed that the METEVI2 rates were comparable to ETa estimated by OpenET methods and observed ETa values and showed almost a similar monthly ETa pattern with varying values. Besides model strengths, the calculation of the soil-evaporation component was determined to be the weakness of the VI-based model. The promising results encouraged testing the METEVI2 in various regions across the world. To reach the last objective, METEVI2 was compared with monthly ETa data measured at 15 flux towers located in different regions and two well-established SEB models (WaPOR and SSEBOp) over two different study regions (Bekaa and Gezira). Overall ETa estimates showed moderate to high accuracy compared to ETa data from flux towers located in croplands, with RMSE ranging between 12.4 and 40.2 mm/month. The METEVI2 could capture the vegetation dynamics and estimated ETa followed a similar seasonal pattern to the observations. With a field spatial resolution, METEVI2 offers a potentially useful tool for agricultural water management applications and therefore, can be utilized interchangeably due to its acceptable performance in comparison to ground measurements and RS-based products. The primary obstacle to utilizing ET-VIs lies in their inability to account for the impacts of stress or soil evaporation. Hence, in future investigations, it is imperative to incorporate water stress and evaporation factors. Because it is implemented in GEE, METEVI2 can be used to estimate the ETa at the global scale and improve the understanding of crop water use as well as improve the ETa estimates as a major component of the hydrologic cycle. Altogether, this dissertation provides methods for a comprehensive ETa estimation of agricultural water use at 30 m spatial resolution and offers a monitoring tool for cropping areas and their water consumption, especially in the absence of systematic in situ data. Vegetation-index-based ETa estimation provides robust results for crops grown in arid regions and may therefore play a key role in informing land and water management in regions lacking in situ data. |
|---|---|
| DOI: | 10.17192/z2023.0504 |