Soil electrical resistivity monitoring as a practical tool for evaluating irrigation systems efficiency at the orchard scale: a case study in a vineyard in Central Chile

Abstract

In many orchards, irrigation scheduling is designed based on data from meteorological networks and considering homogeneous soil properties. Such assumptions may result in inefficient irrigation, which is difficult to constrain without expensive or invasive techniques. Here we have evaluated the ability of the electrical resistivity tomography (ERT) for detecting meter-scale irrigation uniformity and deep percolation during irrigation. The spatiotemporal variability of soil volumetric water content (VWC) in a vineyard located near Santiago (Chile) was inferred using ERT monitoring of two irrigation cycles. The electrical resistivity structure up to 4 m depth was estimated using two-dimensional inversion of ERT data. ERT results were verified by comparing resistivity models with VWC measured with soil moisture sensors, soil properties mapped in a 2 m-depth soil pit, and the spatiotemporal evolution of VWC obtained by solving numerically Richards equation. Largest temporal variations of resistivity were observed within the root depth (1 m) and are consistent with expected relative changes in VWC during irrigation. ERT images exhibit lateral changes in resistivity at these depths, likely indicating non-uniform infiltration of water controlled by observed soil texture variations. Resistivity changes were also observed below the root zone, suggesting that a fraction of the irrigation water percolates downward. These findings can be explained by an excess of irrigation water applied during the monitoring, which was planned considering regional evapotranspiration (ET) data that overestimated the actual ET measured at the vineyard. Altogether, our results suggest that ERT monitoring during irrigation is a cost-effective tool to constrain the performance of irrigation systems.

Appendix: ERT data fit

Tables 4 and 5 show the misfit between observations and preferred inversion models’ responses for all ERT surveys made with N = 4 and N = 1, respectively. Initial RMS error is the average misfit between observations and the response of the starting model (a 70 Ωm homogenous half space). The data misfit range indicates the differences between measured data and model responses observed overall quadripoles, expressed as percentage of the measured data value (Figs. 10, 11, 12).

Table 4 Misfit for preferred models of ERT surveys made with N = 4

Table 5 Misfit for preferred models of ERT surveys made with N = 1

Fig. 10

Comparison between observations and preferred model responses for ERT surveys made with N = 4. For each day, upper panel shows the pseudosection calculated from observed data, middle panel shows the pseudosection calculated from the inversion resistivity model, and lower panel shows the inversion resistivity model. Black dots in pseudosections indicate the estimated location of apparent resistivity points measured with each quadripole

Fig. 11

Comparison between observations and preferred model responses for ERT surveys made with N = 1. For each day, upper panel shows the pseudosection calculated from observed data, middle panel shows the pseudosection calculated from the inversion resistivity model, and lower panel shows the inversion resistivity model. Black dots in pseudosections indicate the estimated location of apparent resistivity points measured with each quadripole

Fig. 12

Sensitivity test of resistivity feature C2 for ERT surveys made with N = 4. Comparison between observations and responses of a modified version of the preferred model, where a 100 Ωm rectangle replaces the area roughly occupied by C2 in the preferred model. For each day, upper panel shows the pseudosection calculated from observed data, middle panel shows the pseudosection calculated from the inversion resistivity model, and lower panel shows the inversion resistivity model. Dashed-line polygons in pseudsections panels outline the areas where the datafit is worse than in the preferred model responses. Dotted-line polygons in model panels outline the 100 Ωm rectangle that was added to the preferred model. Global RMS error for each model is indicated