Comparing different FAO approaches for assessing irrigation needs and scheduling: application on a maize field in Mediterranean area

Selecting the best irrigation management is required for improved use of water resources and for achieving sustainable crop productions. That selection implies accurate predictions of crop water requirement in response to meteorological variables and phenological stages. A plethora of irrigation models are reported to date in literature, many of which are based on three different approaches proposed by the FAO organization, the single and double crop coefficient methods and, the canopy-cover curve determination included in the AquaCrop model.

The objective of this study is to compare irrigation needs and scheduling obtained by the three aforementioned approaches in the agricultural context of the Po River Plain (northern Italy). The first and the second approaches were simulated respectively by Sim1Kc and IdrAgra models, which implemented the algorithms and crop parameterizations reported in FAO paper 56 for a crop water requirement estimation. While the third approach was simulated by the open source version of AquaCrop software.

Models were tested on a maize plot located in the lower-east part of Lombardy Po River Plain characterized by a humid sub-tropical climate, according Köppen classification. A single sandy-loam layer profile of medium-textured soil 1 m deep was considered for the simulation. Crop parameters values in Sim1Kc and AcquaCrop models were mutuated from the IdrAgra model, which is routinely applied in the region as the reference model for the assessment of crop water requirements.

Actual evapotranspiration and irrigation needs were evaluated respectively in rainfed and irrigated simulations. These latter were performed replacing soil moisture at the field capacity when 70% of TAW was reached. Results achieved in three agrarian seasons characterized by low, medium and high rainfall volumes (from June to September) were compared (respectively the years 2009, 2002 and 2014 with about 41 mm, 116 mm and 152 mm of rainfall).

The results show that in rainfed conditions, for each year, actual evapotranspiration simulated by the models were consistent with each other, with an average RMSE, calculated comparing the models in pairs, of about 1 mm over the season. Differences among the models were mainly observed in the first part of the season (respectively before the thirtieth day after the sowing) and for each year, probably caused by a still limited crop and root development, which highlights the differences in simulating water fluxes exchanges in soil-vegetation domain proposed by three modeling approaches.

Concerning irrigations, IdrAgra and AquaCrop appear very consistent with each other in volumes and frequency, especially during mid-crop stages and in all years with a total irrigation volume of about 400, 300 and 180 mm and with 10, 12 and 8 irrigation interventions respectively in the years 2009, 2002 and 2014. Results of Sim1Kc were consistent with those obtained by the other models only in mid and end crop season, whereas no irrigations were suggested in the first part of the season.