Reservoirs allow water to be stored, so that it can be used whenever needed. They are very important for the irrigation of cultivated lands. However, it is important to control the rate of water release from these reservoirs, so that the water usage can be optimized according to current requirements. These requirements are dependent on current moisture level in the soil, the rainfall forecast and so on.
This issue has been investigated by Sangeeta Kumari, a graduate student of Civil Engineering in Indian Institute of Science (IISc) along with her advisor, Prof Pradeep Mujumdar. Prof. Mujumdar points out, “In India, about 75% of water is used in agriculture. But water-use efficiencies in the country are very low in the agriculture sector – just 35-40%. Improvement in scheduling of water applications is therefore necessary.”
Reservoir operations are scheduled for a crop season, by dividing the entire season into short periods of 7 or 10 days. Also, the field area under consideration needs to be divided into grids. Such localization with respect to both space and time are important to ensure that water release is optimum. The water release is planned for each grid and period adaptively, by considering various factors, such as how much of the released water is utilized by the crops. The researchers in IISc have taken this line of research further by including factors such as rainfall forecasts and soil moisture level into the scheduling policy.
Recently, reliable medium-range rainfall forecasts have become available at the grid level from National Centre for Medium Range Weather Forecasting (under Government of India), and soil moisture level can also be estimated from satellite images. These have been used by the researchers for improving the water release schedules.
The researchers have developed a mathematical model for the volume of water to be released for every grid and every period. The model proposed by the researchers is based on known mathematical models of reservoir storage, soil moisture and also ponding depth -- the minimum amount of water that must be present above the soil for growth of crops like paddy.
These models, in turn, take into account factors like rainfall, water intake by the crops, percolation of water through the soil, irrigation from the reservoir and evaporation of water from the reservoir. The researchers have also taken into account the uncertainty in reservoir storage and the soil moisture, by using a mathematical technique called 'fuzzy logic'. Another technique called Dynamic Programming has been used to compute the optimal water release policy. At every period, this dynamic programming is run to compute the optimal release policy from that period till the end of the season, using reservoir storage and soil moisture at the beginning of that period as well as grid-level rainfall forecast. The policy is then implemented for that particular period, taking into account the actual rainfall for each grid. Water is released from the reservoir according to the policy, and divided among the grids as par crop requirements. The soil moisture and reservoir storage at the end of that period will be used to initialize the dynamic programming for the next period.
“The crops compete for a limited amount of water, and such competition is addressed through the crop yield functions that account for sensitivity of a crop to deficit moisture in a given time period”, says Prof Mujumdar.
The model was tested by doing a simulation study for Lakkavalli dam on river Bhadra, in Chikmagalur district of Karnataka. The reservoir inflow and outflow rate, the soil moisture levels and rainfall forecast data for 10-day periods is available from various sources. Four major crops are cultivated in this area -- paddy, maize, citrus and sugarcane. The performance of the proposed policy was compared against a standard operating policy of the reservoir, where the release rate for each period is decided simply based on the requirements of the crop, -- without taking into account ponding depths and rainfall forecasts. The importance of including the ponding depth and the uncertainty in the reservoir storage and soil moisture level were studied separately.
According to Prof Mujumdar, “The experiments demonstrate that use of satellite estimated soil moisture along with medium range weather forecasts in scheduling irrigation could result in significant savings in water by increasing the efficiency of water use.”
About the Authors:
Sangeeta Kumari is a PhD student in Civil Engineering department of IISc, and Dr. Pradeep Mujumdar is a professor in the same department.
About the Paper: “Fuzzy State Real-Time Reservoir Operation Model for Irrigation with Gridded Rainfall Forecasts” appeared in Journal of Irrigation and Drainage Engineering, September 2015