Application of a remote sensing-based soil water balance for the accounting of groundwater abstractions in large irrigation areas

Abstract

The sustainability of groundwater abstractions for irrigation practices must be monitored to achieve a long-term equilibrium in aquifers. The accounting of irrigation water requirements in river basin management plans is commonly and mainly calculated by combining the average multiannual irrigated surface estimates and the unitary crop water requirements. However, remote sensing approaches allow water managers to incorporate more dynamic knowledge of a territory by monitoring irrigated crops. Hence, time series of biophysical products processed from Earth Observation data for 4 years (2010–2013) were incorporated into a remote sensing-based soil water balance to estimate spatially distributed irrigation water requirements on a monthly time scale over a semiarid environment, where agricultural practices greatly depend on groundwater resources. The simulated monthly water abstractions were then evaluated regarding monthly groundwater level changes recorded from a piezometric network. The results indicated that groundwater level changes on a monthly scale could be explained in more than 75% of the cases. Therefore, a simple remote sensing-based approach brings temporally and spatially distributed information of great practical value to river basin water managers according to their management necessities.

Annex: RS_SWB description

HidroMORE^® is a software that performs a distributed Remote Sensing-based Soil Water Balance (RS_SWB) over large areas. The not novelty, but proposed as an operational application advance for the accounting of groundwater abstractions, is the use of Earth Observation (EO) data to feed the soil water balance model. The goal is to monitor over time and space the vegetation development from irrigated crops, using biophysical parameters derived from satellite images. The RS_SWB approach has been validated and tested at different scales and crop types such us in rain-fed grapes plots (Campos et al. 2012), drip-irrigated table grape vineyards (Balbontín et al. 2017), drip-irrigated apple orchards (Odi-Lara et al. 2016), using in situ soil water content measurements from the REMEDHUS network of 23 stations located in the central semi-arid zone of the Spanish Duero river basin, in where rain-fed cereals, summer irrigated crops and vineyards were located (Sánchez et al. 2010), or at river basin scale by comparing, for those irrigated areas, the estimated IWR against irrigation water demand assigned by the river basin management plans over the principal Spanish mainland river basins (Garrido-Rubio et al. 2018). This annex briefly explains the methodology applied.

A soil water balance driven by remote sensing at root depth, based on the FAO56 methodology (Allen et al. 1998), has been implemented. The goal is to estimate from Eq. 2 (expressed in mm), at daily time scale and at pixel spatial scale, maps of irrigation water requirements (I). This thematic cartography is derived after the estimation of the other soil water balance components, such us crop evapotranspiration (ET_c), deep percolation (DP), precipitation (P), run-off (RO), and soil water depletion along the root zone (D_r). In operational terms, the soil explored by roots defines the depth limits in where RS_SWB is applied, and it is limited by its maximum soil depth taken from the soil geodatabases (see Application of RS_SWB over the Mancha Oriental System in “HidroMORE: a software that enables RS_SWB over wide areas”). Moreover, to start the balance and regarding the previous year 2009, as a humid year (Spanish Meteorological Agency, http://www.aemet.es), the soil profile is considered full of water, and hence, initial condition on soil water depletion (D_r,i−1) is 0 mm. Besides CR is considered negligible in study zone due to depth of vadose zone, so it is removed from equation, and in the same way RO is not considered, as the model does not take into account lateral fluxes. In parallel, DP estimations derived by HidroMORE^® as the water that exceeds the soil water content, which can be retained along the root layer, could be included as vertical inflows into the groundwater resources. However vertical fluxes in the aquifer are not well known yet. Hence, and considering the saturated zone normally deeper than 50 m (Fig. 2), such inflow would not reach the aquifer water storage in the incoming irrigation campaign and therefore, out of the previous presented evaluation. In the study, daily values of irrigation water requirements were temporal aggregated at pixel-based scale into a monthly time step, to be compared with monthly time series of groundwater level changes (λ_gli):

$$D_{{{\text{r}},i}} = D_{{{\text{r}},i - 1}} - (P - {\text{RO}})_{i} - I_{i} - {\text{CR}}_{i} + {\text{ET}}_{i} + {\text{DP}}_{i} .$$

(2)

To monitor vegetation development and its daily crop evapotranspiration, ET_c is calculated following the “dual crop coefficient” methodology (Wright 1982), which has been suggested as the most suitable approach when crops have partial ground cover or are under frequent irrigation (Allen et al. 1998, 2005). This approach multiplies ET_o by a crop coefficient (K_c) that has two contributions: the soil evaporation coefficient (K_e in Eqs. 3 and 8), which describes soil water evaporation, and the basal crop coefficient (K_cb in Eqs. 3 and 4), which describes potential crop transpiration. Additionally, the crop water stress coefficient is considered (K_s in Eqs. 3 and 6), which limits the transpiration when water is not readily available for roots. Therefore, the evapotranspiration adjusted for soil water conditions (ET_cadj) is calculated as shown in Eq. 3:

$${\text{ET}}_{\text{cadj}} = {\text{ET}}{}_{\text{o}} \cdot (K_{\text{cb}} \cdot K_{\text{s}} + K_{\text{e}} ).$$

(3)

The basal crop coefficient (K_cb) and the green cover fraction (f_c) link the FAO56 model and EO data as they are derived from optical reflectance provided by the remote sensing images. Both parameters were mapped daily applying the linear relationships NDVI-K_cb and NDVI-f_cv (obtained by means of interpolating the daily synthetic images). For the first relationship, applications among several crops have been proposed, such as maize (Bausch 1993; Choudhury et al. 1994), alfalfa (Bausch and Neale 1987), and grapes (Campos et al. 2010). The NDVI-K_cb relation presented in Eq. 4 was chosen in this study because it was developed in the study area (Campos et al. 2010). Regarding the second relationship (NDVI-f_cv), many authors have proposed different examples (Johnson and Trout 2012; López-Urrea et al. 2009; Xiao and Moody 2005) and Eq. 5 was selected because it was developed for the same crops in the study area (González-Piqueras 2006). Consequently, the soil water balance as explained above is considered as a RS_SWB:

$$K_{\text{cb}} = 1.44 \cdot {\text{NDVI}} - 0.1$$

(4)

$$f_{\text{cv}} = 1.19 \cdot {\text{NDVI}} - 0.16.$$

(5)

Finally, the way on how HidroMORE^® determines the irrigation water requirements is a consequence of the actual soil water content and the crop water demand. Consequently, HidroMORE^® daily monitories the soil water depletion in order to check if soil water content is good enough for crop water demands. In that sense, the FAO56 model calculates K_s distinguishing between two soil water depletion stages. Both phases follow a linear relationship between Total Available Water (TAW) and Readily Available Water (RAW) at the soil root depth (Eq. 6). TAW is obtained by Eq. 7. θ_FC and θ_WP are the volumetric soil water content (cm³/cm³) at field capacity and the wilting point, respectively. Z_r refers to the actual root depth (mm). RAW is calculated from an average fraction of TAW (p) that can be depleted from root depth without inferring crop water stress. The soil evaporation reduction coefficient (K_r) is calculated from a soil evaporation modification (Torres and Calera 2010) in parallel with the K_e calculation (Eq. 8). It has been demonstrated that K_r overestimates soil evaporation under high atmospheric demands because it occurs in the study zone. Therefore, a correction coefficient (m) equal to 0.3 was used in Eq. 9, selecting the minimum value between two relationships: (1) a coefficient between Readily Evaporable Water (REW) and ET_o and (2) a linear relationship between Total Evaporable Water (TEW), REW, and m, considering depletion from the soil surface layer (D_e). Regarding the study objectives, the RS_SWB model was computed to estimate irrigation amounts necessary to maintain crop transpiration at potential rates during the growing season. Therefore, D_r values were kept over RAW, whereas K_s was maintained equal to 1 during the growing period:

$$\begin{aligned} K_{\text{s}} = 1 \leftrightarrow D_{{{\text{r}},i}} \le {\text{RAW}}\quad ({\text{Stage I}}) \hfill \\ K_{\text{s}} = \frac{{{\text{TAW}} - D_{{{\text{r}},i}} }}{{{\text{TAW}} - {\text{RAW}}}} = \frac{{{\text{TAW}} - D_{{{\text{r}},i}} }}{{(1 - p) \cdot {\text{RAW}}}} \leftrightarrow D_{{{\text{r}},i}} > {\text{RAW}}\quad ({\text{Stage II}}) \hfill \\ \end{aligned}$$

(6)

$${\text{TAW}} = 1000 \cdot Z_{\text{r}} \cdot (\theta_{\text{FC}} - \theta_{\text{WP}} )$$

(7)

$$K_{\text{e}} = K_{\text{r}} \cdot (K_{\text{cmax}} - K_{\text{cb}} ) \le (1 - f_{\text{c}} ) \cdot K_{\text{cmax}}$$

(8)

$$K_{\text{r}} = \hbox{min} \left\{ {\frac{\text{REW}}{{{\text{ET}}_{\text{o}} }}} \right.,\left. {m \cdot \frac{{{\text{TEW}} - D_{{{\text{e}},i}} }}{{{\text{TEW}} - {\text{REW}}}}} \right\}.$$

(9)