|M.Sc Student||Alaa Jamal|
|Subject||Utilizing probabilistic weather forecasting for optimal|
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Linker Raphael|
|Dr. Housh Mashor|
|Full Thesis text|
Optimal use of water resources is one of the most important aims to face the future climate change and food security concerns. In this study we developed a Decision Support System (DSS) to optimize the water use in the agriculture sector by deriving optimal irrigation schedule on the field scale. The purpose of irrigation scheduling is to coordinate the timing and amount of irrigation in a way that deals with two conflicting objectives of minimizing irrigation water use while maximizing the yield. While many irrigation DSSs exist, the majority of these tools are based on deterministic models in which all the parameters of the optimization problem are assumed to be known with certainty. In the deterministic irrigation scheduling problem, parameters like rainfall, radiation, wind speed and other meteorological variables are assumed to be perfectly known along the optimization horizon. The scheduling problem is then defined as finding the timing and the amount of water while knowing what is going to happen in the future with full certainty. However, in reality the actual weather will differ from the assumed weather forecast and the performance of the irrigation schedule will be lower than expected. In order to reduce the weather uncertainty, optimization based on a number of weather scenarios should be used.
In this research we developed a stochastic optimization approach using probabilistic seasonal weather forecasts, based on ten weather scenarios with uniform probability of occurrence. Three approaches were investigated: implicit approach, single-stage approach and two-stage approach. In addition to the stochastic approaches, a deterministic approach using either perfect seasonal forecasts or perfect weekly forecast is presented for comparison purposes.
The solution search in the optimization sub-model was conducted by Genetic Algorithm. The decision variables are the daily irrigation amounts. The solutions were used as input for the simulation model SWAP to evaluate the profit resulting from each irrigation schedule. In our case there were 47 potential irrigation days, meaning 47 decision variables. To reduce this number of decision variables, weekly aggregation was performed after the first next week, so that every seven days have the same daily irrigation amount.
To evaluate the benefit of the developed approaches, benchmark solutions are presented. These approaches are: Real-Field schedule, Agriculture Extension Service of Israel schedule, No-Stress approach and No-Irrigation approach. All these approaches were tested for a chickpea field of the HaZorea kibbutz. The results show that the most beneficial approach was the perfect seasonal forecast, which is not surprising as it considered the longest forecast horizon and the most certain one. Using the stochastic optimization approach resuld in the highest profit between all the presented approaches after the perfect seasonal approach. The perfect weekly approach yielded high profits compared to the irrigation schedule implemented in the actual field and the two approaches of No-Stress and No-Irrigation. The suggested stochastic approaches increased the profit significantly compared to the profit expected from the Real-field schedule and AESI schedule.