Get Complete Project Material File(s) Now! »

Remote sensing data relevant to crop water budget monitoring

Remote sensing offers the only possibility for monitoring land surface variables at different spatial resolutions and temporal frequencies, thus facilitating a systematic and comprehensive observation over extended areas. Furthermore, remote sensing observations are especially practical in areas where man-made measurements are difficult to perform or simply unavailable (Li et al., 2009; Rango, 1995). Remote sensing has played an important role in the development and application of several models over extended areas for monitoring water resources, being able to map ET and its associated variables, such as vegetation cover, land surface temperature and soil moisture (Fig. 1.2). Remote sensing has the particular interest of being cost effective and operational in its implementation over extended areas, allowing estimating energy-water balance components and its associated variables at multiple spatial and temporal scales. This advantage allows coupling remote sensing data, water and surface energy model in order to better understand the hydrological processes at different scales. The remote sensing data especially relevant for the monitoring of water resources are presented in Table 1.1 and detailed below.

Visible – Near Infrared data

Visible and near infrared (VNIR) reflectances have the advantages of monitoring vegetation/crops in terms of phenology, health and vigor among others. This is because green plant leaves show very low reflectance in visible regions (0.4 – 0.7 μm) due to a strong absorptance by photosynthetic and plant pigments and very high reflectance in the near infrared regions (0.7 – 1.3 μm) due to a low absorptance by subcellular particles or pigments and as well as a considerable scattering at mesophyll cell wall interfaces (Gausman, 1977). These characteristics have served as the basis for many applications of remote sensing to crop management by using mainly vegetation indices (VI) (i.e. differences, ratios, or linear combinations of reflectances in visible and near infrared wavebands). VI have shown good correlations with plant growth parameters such as green biomass (Pinter et al., 2003), leaf area index (Duchemin et al., 2006), and fraction of absorbed photosynthetically active radiation (Pinter et al., 1994), among others.
In crop water management, VI have been widely used to derive crop coefficients (e.g. defined as the ratio of ET and a reference ET value in optimal ET conditions) (Bausch and Neale, 1987; Choudhury et al., 1994; Singh and Irmak, 2009). This is because crop coefficients primarily depend on the dynamics of canopies (cover fraction, leaf area index, greenness and phenology). Hence, VI-based crop coefficients have been of great value in ET and irrigation scheduling algorithms in order to estimate the crop water requirements (Allen et al., 2011; Pereira et al., 2015; Singh and Irmak, 2009). Several studies have proven that local adjustment by phenology and crop coefficient are expected to be more suitable for estimating ET and crop water needs than the use of tabulated crop coefficient values (Allen et al., 2011; Bausch, 1995; Pereira et al., 2015). Such local adjustments usually rely on site-specific measurements or observations of crop growth and, consequently, VI based approaches are recommended for crop coefficients and irrigation management.
In addition, VNIR have received an especial interest for energy balance applications, providing robust estimates of the fraction of net radiation going into soil heat flux by means of VI (Daughtry et al., 1990) or for estimating surface albedo (Liang, 2001; Qu et al., 2015). VI are also essential auxiliary data in the estimation of surface emissivity to estimate the land surface temperature (LST) from thermal infrared data (Jiménez-Muñoz et al., 2006; José A Sobrino et al., 2008). Furthermore, VNIR are needed to detect the full range of surface conditions in vegetation cover needed in several methods based on the contextual information in remotely sensed LST and VI data (Merlin, 2013; Merlin et al., 2014; Moran et al., 1994).
One of the main advantages of VNIR sensors over other spectral sensors is the high spatial resolution suitable for crop monitoring. Resolution less than 100 m (e.g. Landsat, ASTER, Sentinel-2) allows only one to six observations per month in orbital cycle. However, SPOT series or other commercial satellites (e.g. QuickBird, Worldview, GeoEye) with very high (< 10 m) spatial resolution are generally cost prohibitive and hence they are not useful for operational implementations. The launch of Sentinel-2A/B represents a breakthrough for freely available VNIR missions, providing VNIR data at ~10 m resolution to systematically monitor crops at a weekly repeat cycle (from 5 to 12 days).
Despite plant water stress and senescence period can be detected by VI time series (Adams et al., 1999), water stress-induced impact in these wavelengths is not sufficiently large over biologically significant changes in plant water content for practical uses in the monitoring of water stress in the field (Bowman, 1989; Carter, 1991). Unlike VNIR, thermal infrared data have proven to be very useful in assessing the crop water stress (Jackson, 1982) as it is presented in the next section.

Thermal infrared data

Land surface temperature (LST) is an essential variable that modulates radiative, latent and sensible heat fluxes at the soil-plant-atmosphere interface. LST can be obtained globally and operationally from thermal infrared remote sensing observations. Hence, LST is a useful variable for monitoring the carbon, water and energy fluxes from field to regional scales (Anderson et al., 2008).
LST data have been a key land surface variable as input for many environmental and hydro-meteorological applications, including climatological studies (Anderson et al., 2007; Hansen et al., 2010), extreme weather monitoring such drought monitoring (Anderson et al., 2011; Jiménez-Muñoz et al., 2016; McVicar and Jupp, 1998), soil moisture estimates (Amazirh et al., 2018; Merlin et al., 2012b) and irrigation and water resource management (Anderson et al., 2012b; Bastiaanssen et al., 2007; Droogers et al., 2010). LST is particularly useful for the monitoring of crop water management since it is very sensitive to plant water stress and a strong indicator of changes in root-zone soil moisture (Anderson et al., 2012a, 1997; Moran et al., 2009). Thus, LST can be related to the root-zone soil moisture (RZSM) by means of the canopy temperature and its associated plant transpiration (Boulet et al., 2007; Hain et al., 2009; Moran et al., 1994) given the coupling between the surface energy and water balance (e.g. Wetzel et al., 1984).
LST can be derived from satellite thermal sensors at different spatial and temporal scales. However, the main limitation in the existing thermal missions is the unavailability of high spatial and temporal resolutions at the same time. For instance, missions offering high revisit time (e.g. MODIS, AVHRR, MSG/SEVIRI, VIIRS and Sentinel-3) usually provide a low spatial resolution, and conversely, those offering high spatial resolution (e.g, Landsat and ASTER) provide a low temporal resolution (Fig. 1.3). Therefore, the ability for monitoring water resource at crop field scale (~100 m) is limited by the low revisit time and even hampered by cloudy conditions, hence preventing the monitoring of rapid changes of the vegetation water status.
Recent studies have highlighted the importance of thermal observations at high resolution with a near daily revisit for vegetation water status monitoring (Cao et al., 2019; Guzinski and Nieto, 2019; Sobrino et al., 2016). Thus, ideally a constellation of polar orbiting satellites (e.g. Landsat, ASTER) would appear to be the best solution to meet these requirements, which is potentially achieved by the ECOSTRESS mission (Hulley et al., 2017), recently launched in June 2018, or the foreseen TRISHNA mission (Lagouarde and Bhattacharya, 2018). ECOSTRESS, onboard of International Space Station, will address critical questions on plant-water dynamics and future ecosystem changes with climate by means of LST, ET, Water Use Efficiency, and Evaporative Stress Index data products at ~60 m spatial resolution every few days (<5) at varying times of day. Consequently, the detection is further enhanced in heterogeneous environments (such as agricultural areas) by the high spatiotemporal resolution (Hulley et al., 2017). However the ECOSTRESS overpass time changes and does not offer global coverage, therefore it is not optimal for monitoring crop management under operational implementations. The TRISHNA mission will combine a high spatial resolution (50 m) and high revisit time (about 3 days) in the thermal domain with a global coverage. The two main scientific objectives driving the mission are the monitoring of energy and water budgets of the continental biosphere and the monitoring of coastal and continental waters (Lagouarde and Bhattacharya, 2018).
Fig. 1.3. Different spatial and temporal resolution of current and near future thermal satellite observations related to different target observation scales.

Remote sensing data relevant to crop water budget monitoring

Before the launch of TRISHNA mission, the disaggregation of existing low resolution LST data to high spatial resolution with a relatively satisfying accuracy can be performed. Disaggregation methods focus on decomposing pixel-based temperatures providing a better dataset of LST with finer temporal and spatial resolutions based on the information obtained from different sensors. Therefore, disaggregation methods aim to achieve appropriate LST data for monitoring crop water budget at crop field scale (illustrated in Fig. 1.3). The basic idea behind these methods is to establish either a statistical relationship or a physically based model between coarse scale LST and fine scale auxiliary variables. In these methods, satellite data in the VNIR wavelengths available at a resolution finer than that of most thermal sensors have been essential to bridge the gap between the low spatial resolution and the high temporal resolution of available LST observations (Zhan et al., 2013). Consequently, most common disaggregation LST methods have been based on a scale invariant relationship between LST and VI, largely related to the fractional vegetation cover. The VI-based methods are still the most used operational approaches due to the availability of data at high spatial and temporal resolutions, such as DisTrad, TsHarp, among other algorithms (Agam et al., 2007a; Bindhu et al., 2013; Kustas et al., 2003; Mukherjee et al., 2014; Zhan et al., 2013).
In addition to the use of VNIR data, other more complex disaggregation methods have proposed the use of the LST-VI feature space to derive soil water status indices that could better represent the variability in LST and hence improving the disaggregation accuracy over agricultural areas with high moisture content (Chen et al., 2010; Sandholt et al., 2002; Yang et al., 2010). This procedure has been further extended by using additional factors that modulate the LST, reflecting the soil moisture content and vegetation type (Amazirh et al., 2019; Merlin et al., 2012a, 2010; Yang et al., 2011). For instance, Merlin et al. (2010) distinguished between photosynthetically and non-photosynthetically active vegetation from time series of optical shortwave data to be included in the disaggregation procedure. Then soil moisture proxies derived from microwave data can take into account the soil moisture effects on the disaggregation of LST (Merlin et al., 2012a; Amazirh et al., 2019). Although these latter methods can provide better accuracies than using only LST-VI relationships, they require additional parameters, which make them difficult to be implemented operationally. Therefore, implementing disaggregation methods on an operational basis with reasonable accuracies implies new challenges in the methods.

READ  Elastic interior transmission eigenvalue problem 

Microwave data

Microwave wavelengths are one of the most sensitive to the variations in soil moisture given the large contrast of the emission from the earth’s surface between the water and land. Thus, surface soil moisture (SSM) can be estimated from remote sensing (Entekhabi et al., 1994; Kerr et al., 2010). However, remote sensing instruments are only able to collect soil moisture information to an estimated depth of approximately the first 5-10 cm of the surface. Indeed, the microwave emission in this frequency is severely attenuated in the soil porous medium (Entekhabi et al., 1994; Kerr et al., 2010).
According to Schmugge et al. (2002), microwave data are characterized by four unique advantages over other spectral regions: i) the atmosphere is effectively transparent providing all weather coverage; ii) vegetation is semi-transparent allowing the observation of underlying surfaces; iii) the microwave measurement is strongly dependent on the dielectric properties of the target, which for soil is a function of the amount of water present; and iv) the microwave measurement is independent of solar illumination, which allows day or night observation.
There are two microwave remote sensing techniques: the passive and active microwave sensors. The passive microwave sensors (radiometers) detect the naturally emitted microwave energy within its field of view using very sensitive detectors. However the amounts of energy are generally very small due to the wavelengths, which are much longer compared to optical wavelengths. Thus, the fields of view must be large to detect enough energy to record a signal. Most passive microwave sensors are therefore characterized by a low (~30 – 60 km) spatial resolution. Among satellite passive missions, the SMOS satellite, launched in 2009, has been widely used for SSM retrieval, with an accuracy requirement of 4%. It is based on an L-band (1.4 GHz) antenna and is the first space mission dedicated to observe SSM globally (Kerr et al., 2010). The AMSR-E mission, launched in 2002, provides brightness temperature measurements at six frequencies from 6.9 to 89 GHz in horizontal and vertical polarizations, of which C-band (6.9 GHz) and X-band (10.7GHz) channels are suitable for retrieving SSM (Njoku et al., 2003) at spatial resolutions ranging between 25 and 50 km. The SMAP mission, launched in 2015, combines a radiometer (passive) and a Synthetic Aperture Radar (SAR, active) instrument within the L-band range (1.20–1.41 GHz) to provide measurements of SSM moisture with a global coverage in 2–3 days. The ASCAT sensor is a C-band scatterometer (5.255 GHz, VV polarization) at a spatial resolution of about 50 km, operating on-board the Meteorological Operational (MetOp) satellite since 2006.
Regarding the active sensors, the most popular is the Sentinel-1 mission, launched in 2014, providing C-band SAR data at 20 m spatial resolution with an unprecedented repeat cycle of 6 days by combining both ascending and descending overpasses (3 days by combining the two satellites available since 2015). Although backscatter signals data have potential to monitor SSM (e.g. Amazirh et al., 2018; Gao et al., 2017; Zribi et al., 2011), there is currently no global operational SSM product at such fine resolution. This is notably due to the difficulty to model in time and over extended areas the impact of vegetation cover/structure and surface roughness on the backscatter signal (Zribi et al., 2011, 2008).

Table of contents :

1.1. General context
1.2. Remote sensing data relevant to crop water budget monitoring
1.2.1. Visible – Near Infrared data
1.2.2. Thermal infrared data
1.2.3. Microwave data
1.3. Modelling the crop water budget components from remote sensing data
1.3.1. Evapotranspiration modelling
1.3.2. Root-Zone Soil Moisture modelling
1.3.3. Irrigation modelling
1.4. Objectives
1.1. Contexte général
1.2. Données de télédétection pertinentes pour le suivi du bilan hydrique des cultures
1.2.1. Données Visible – Proche infrarouge
1.2.2. Données thermiques infrarouges
1.2.3. Données micro-ondes
1.3. Modélisation des composantes du bilan hydrique des cultures à l’aide de la télédétection
1.3.1. Modélisation de l’évapotranspiration
1.3.2. Modélisation de l’humidité en zone racinaire
1.3.3. Modélisation de l’irrigation
1.4. Objectifs
2.1. Introduction
2.2. Morocco: Haouz Plain
2.2.1. Meteorological data
2.2.2. Flux data (Eddy-covariance system)
2.2.3. Soil Moisture
2.2.4. Irrigation
2.2.5. Fractional green vegetation cover
2.2.6. Temperature data
2.3. Chile: Copiapó Valley
2.3.1. Meteorological data
2.3.2. Ground-based land surface temperature
2.4. Remote sensing data
2.4.1. Landsat data
2.4.2. ASTER Global Emissivity Datasets (ASTER GED)
2.4.3. MODIS data
2.5. Conclusion
3.1. Introduction
3.2. FAO-56 dual crop coefficient method
3.2.1. Basal crop coefficient (Kcb)
3.2.2. Evaporation reduction coefficient (Ke)
3.2.3. Water stress coefficient (Ks)
3.3. Remote sensing data integrated into FAO-2Kc
3.4. Estimating water budget components from ground-based optical/thermal data
3.4.1. Implementation of a contextual method at in situ level
3.4.2. Root-zone and soil surface water status from optical/thermal data: Ks and Kr estimation
3.4.3. First-guess water budget components
3.4.4. Re-analysis of water budget components
3.5. Summary and conclusions
3.6. ARTICLE: Estimating the water budget components of irrigated crops: Combining the FAO-56 dual crop coefficient with surface temperature and vegetation index data
4.1. Introduction
4.2. Issues for implementing the crop water balance modelling over large areas 91
4.3. Contextual methods for detecting soil and crop water status
4.4. Landsat-derived estimates integrated into a crop water balance model for irrigation retrieval
4.5. From pixel-scale to field-scale irrigation
4.6. Crop coefficients Kcb and Ke derived from contextual methods
4.7. Main results of the spatial application to Haouz Plain
4.8. Summary and conclusions
4.9. ARTICLE: Irrigation retrieval from Landsat optical/thermal data integrated into a crop water balance model: A case study over winter wheat fields in a semi-arid region
5.1. Introduction
5.2. Disaggregation of LST data
5.2.1. Operational method for disaggregating LST data
5.3. Application in Copiapo River Basin – Chile: main results
5.3.1. Disaggregated LST product
5.3.2. Operational estimation of ET every 8 days
5.4. Application over a winter-wheat field (R3) in Haouz Plain – Morocco
5.4.1. Disaggregated LST
5.4.2. Irrigation retrieval by using disaggregated LST
5.4.3. Daily RZSM and ET
5.5. Summary and conclusions
5.6. ARTICLE: An operational method for the disaggregation of land surface temperature to estimate actual evapotranspiration in the arid region of Chile
6.1. Summary of results
6.2. Identifying the main limitations of the methods
6.2.1. Irrigation retrieval approach
6.2.2. LST disaggregation method
6.3. Perspectives
6.3.1. Towards the improvement in spatial and temporal resolution
6.3.2. Towards the use of radar data for a better representation of hydrological processes
6.3.3. Partitioning soil/vegetation components
6.1. Résumé des résultats
6.2. Principales limites des méthodes
6.2.1. Approche d’estimation d’irrigation
6.2.2. Méthode de désagrégation LST
6.3. Perspectives
6.3.1. Vers l’amélioration de la résolution spatiale et temporelle
6.3.2. Vers l’utilisation des données radar pour une meilleure représentation des processus hydrologiques
6.3.3. Partition entre les composants de sol et de végétation


Related Posts