Seasonal forecasts to support cropping decisions: To whom are these useful, and when?

Farmers in irrigated agriculture depend on the water allocated to them through the irrigation season. In drought years, when allocations may be curtailed, early information on water availability is of use to them in supporting their decisions on what to plant and when to plant. Seasonal forecasts provide such early information of water availability, and have been shown to be skilful, though effective lead times depend on the predictability of the local climate as well as the memory of the hydrological system. However, if the information provided is indeed useful, may depend on who uses it, the decisions they take, and the outcomes of those decisions.

Here, we explore seasonal forecasts in supporting decisions farmers take in an area with irrigated agriculture in the Ebro basin in Spain. We develop a simple decision model, considering the preferences farmers have and the crop choices they make depending on if water is expected to be abundant or if it is expected to be scarce. The model also considers the interconnected water allocation decision by operators of the reservoir feeding the irrigation area, and the expectation they have of the balance between supply and demand to the end of season. Demand is informed by the (expected) choices farmers make, while supply is predicted using bias corrected ECMWF System 5 seasonal forecasts and a simple hydrological model.

To understand to whom the forecast is useful, we consider farmers with differing levels of technical capability, which allows them to plant either one or two crops per season; as well as with differing levels of risk averseness. Decisions informed by the seasonal forecasts for each of these farmer types are then compared to those made using perfect information, and to those made using current allocation practice. We then evaluate (relative) benefit through simulating the outcome of the forecast decisions using the observed climate and a crop model to predict yield, and the market price versus investment costs of crops planted to predict net profits.

Results show that seasonal predictions of water availability in the area are skilful, attributed largely to catchment memory, though skill varies; with poorer skill early in the season and around the spring snowmelt. The decision timelines through the season vary per farmer type. Risk averse farmers with less technical capability take key decisions earlier in the season. While forecasts are potentially more useful to those early decisions, we find these are also more sensitive to uncertainty in the forecast. The more technical farmers take decisions late in the season, where skill is higher. They can then rely more on the information provided, though the added value of the forecast to them is lower. These results show that not all are served equally by seasonal forecast information. Some stand to benefit more than others, depending on the decisions they make, and when they take these.