Real-time optimization of irrigation of citrus fields near Valencia (Spain)
In the agriculture sector, farmers were already affected by water restrictions or will most probably be affected in a near future. Deficits in water supply result from the conjunction of a possible precipitation decrease, an increase in the climatic evaporative demand and an increasing water demand of other sectors (urban, industry). There is a need of improving irrigation scheduling tools to optimize the water use efficiency. Such tools can be disseminated directly to farmers or through farmer´s organizations.
In this project, we aim at introducing better knowledge on soil moisture conditions in existing irrigation decision support software. The approach will be tested and demonstrated for a test site and the Huragis software will host the improved determination of soil moisture. The work plan is as follows:
- Acquisition of soil moisture data within the test site, using cosmic ray probes and capacitance probes. The use of cosmic ray data for irrigation optimization is a novel aspect in this study.
- The use of data assimilation for combining optimally land surface model predictions and measurement data. This is the main novel aspect of the study as currently in irrigation optimization studies no real-time soil moisture information is combined with model predictions using a formal methodology like sequential data assimilation.
- Use of weather forecasts including their uncertainty for the prediction of the future evolution of soil moisture contents.
- Calculation of the required irrigation amount with help of the predicted evolution of soil moisture contents and target soil moisture values.
- The required irrigation amount is used as input to a hydraulic model that takes into account logistic restrictions as well.
- Evaluation of the optimized irrigation strategy at the test site.
The study area is an agricultural region with citrus trees near Picassent (Valencia, Spain). For this area irrigation data, meteorological data as well as soil moisture measurements are available and additional equipment is currently installed. One cosmic ray probe will be installed. Cosmic ray probes have the advantage that they can measure soil moisture content for a larger area (with a diameter of around 600 meters). Therefore, one cosmic ray probe is able to capture the average soil moisture conditions over quite a large part of the study area. However, for field specific irrigation more information is needed and therefore 12 additional capacitance probes (with measurements at four different depths in the root zone and slightly below the root zone of the citrus trees) will be installed, which deliver local information on soil moisture conditions.
The land surface model CLM 4 is used to model the exchange of water, energy and carbon dioxide between the land surface and the atmosphere, as well as water and energy flow in the soil and vegetation processes. The CLM model has been set up for the Picassent area. The soil properties implemented in the model still have to be improved on the basis of soil samples.
In 2011 also a data assimilation framework has been developed and implemented in combination with CLM. The local ensemble transform Kalman filter (LETKF), has been implemented and is a new variant of the ensemble Kalman filter (EnKF). The LETKF is a very efficient algorithm for data assimilation, because it treats the different grid cells one by one. The grid cell by grid cell analysis method can easily be parallelized and is useful to decrease the computational burden for the large scale data assimilation. The LETKF uses a local analysis scheme and only considers the observations located in a local region surrounding the analysis grid cell. This makes the LETKF also more robust against filter inbreeding and filter divergence. The observations are selected for each grid cell.
CLM was modified so that multiple stochastic realizations can be processed with LETKF. Uncertainty in forcing terms (precipitation and weather) as well as soil and vegetation parameters can be taken into account. This extension of CLM has been parallelized and efforts have been dedicated to increasing the efficiency of the program by improving the handling of input/output operations in the parallel environment. Finally, LETKF was modified so that parameters can be jointly updated together with the states. This allows for example for our project area in Valencia updating soil and vegetation parameters. The data assimilation framework in combination with CLM is now ready for assimilating measurement data in our model. Test runs were already made with the model and it was possible to generate an ensemble of soil moisture predictions with help of CLM.
CLM 4 also allows for the optimization of irrigation requirements. Calculations have been made with an ensemble of soil moisture predictions, to optimize the irrigation for the next 1, 2, 3 or 4 days. Preliminary results indicate that the amount of irrigation could be lower than the amounts currently applied at the site. However, further calculations are needed to corroborate this. It is still important that the parameterization of CLM improves, that we work with realistic ensembles of model forcings and a better model for plant growth and drip irrigation. These last two aspects are novel and originally not foreseen in the project.
Weblink to project (start: May 2011):
Dr. Xujun Han
Forschungszentrum Jülich GmbH
Tel. +49 02461 8668
Alternative contact person:
Prof. Dr. Harrie-Jan Hendricks Franssen
Forschungszentrum Jülich GmbH
Tel. +49 02461 4462