Irrigation management in lowland areas

As SPHY produces spatial outputs for the soil moisture content in the root zone and the potential and actual evapotranspiration (ET), it is a useful tool for application in agricultural water management decision support. By facilitating easy integration of remote sensing data, crop growth stages can be spatially assessed at different moments in time. The SPHY dynamic vegetation module ensures that all relevant soil water fluxes correspond to crop development stages throughout the growing season. Spatially distributed maps of root water content and ET deficit can be produced, enabling both the identification of locations where irrigation is required and a quantitative assessment of crop water stress.

SPHY has been applied with the purpose of providing field-specific irrigation advice for a large-scale farm in western Romania, comprising 380 individual fields and approximately ten different crops. Contrary to the other case studies highlighted in this paper, a high spatial resolution is very relevant for supporting decisions on variable-rate irrigation. The model has therefore been set up using a 30m resolution, covering the 2013 and 2014 cropping seasons on a daily time step. Optical satellite data from Landsat 8 (USGS 2013) were used as input to the dynamic vegetation module. Soil properties were derived from the Harmonized World Soil Database (Batjes et al. 2012), which for Romania contains data from the Soil Geographical Database for Europe (Lambert et al. 2003). Using the Van Genuchten equation (Van Genuchten 1980), soil saturated water content, field capacity, and wilting point were determined for the HWSD classes occurring at the study site. Elevation data was obtained from the EU-DEM data set (EEA 2014), and air temperature was measured by two on-farm weather stations.

In irrigation management applications like these, a model should be capable of simulating the moisture stress experienced by the crop due to insufficient soil moisture contents, which manifests itself by an evapotranspiration deficit (potential ETactual ET0). Figure 4 shows the spatial distribution of ET deficit, as simulated by the SPHY model for the entire farm on 03 April 2014. When SPHY is run in an operational setting, this spatial information can be included in a decision support system that aids the farmer in irrigation planning for the coming days.

Figure 6: Spatial distribution of evapotranspiration (ET) deficit, as simulated by the SPHY model for a Romanian farm on 03 April 2014. Transparency means no ET deficit.

For calibration purposes, field measurements of soil moisture and/or actual ET are desired. In this case study, one capacitance soil moisture sensor was installed in a soybean field to monitor root-zone water content shortly after 01 May 2014, which is the start of the soybean growing season. The sensor measures volumetric moisture content for every 10cm of the soil profile up to a depth of 60cm. It is also equipped with a rain gauge measuring the sum of rainfall and applied irrigation water, which was used as an input to SPHY. Soil moisture measured over the extent covered by the crop root depth was averaged and compared to simulated values (Figure 5).

Since this study was a demonstration project, only an initial model calibration was performed. The model was in this case most sensitive for the crop coefficient (Kc), affecting the evaporative demand for water. As can be seen in Figure 5, the temporal patterns as measured by the soil moisture sensor are well simulated by the SPHY model. Based on daily soil moisture values, a Nash–Sutcliffe (Nash and Sutcliffe 1970) model efficiency coefficient of 0.6 was found, indicating that the quality of prediction of the SPHY model is “good” (Foglia et al. 2009). Soil moisture simulations could be further improved by conducting a full model calibration, adjusting the soil physical parameters , , , and . Remotely sensed sensed evapotranspiration can be used in the calibration process (Immerzeel and Droogers 2008), although such data are often not available on these small scales as ET is a very complex variable to assess (Samain et al. 2012). It should also be noted that soil moisture content is typically highly variable in space; a very high correlation between point measurements and grid-cell simulations of soil moisture may therefore not always be feasible (Bramer et al., 2013).

Figure 7: Measured and simulated daily root-zone soil moisture content during the 2014 growing season. Rainfall irrigation had been measured by the rain fauge that was attached to the moisture sensor.