Previous Page Table of Contents Next Page

Remote sensing and simulation modelling for on-demand irrigation systems management

G. D'Urso, University of Molise, Dept. SAVA, Campobasso, Italy, M. Menenti, DLO-Staring Centrum, Wageningen, The Netherlands, and A. Santini, Univ. of Naples 'Federico II', Inst. Agric. Hydraulics, Portici, Naples, Italy


The objective of this paper is to describe a procedure for monitoring and improving the performance of on-demand irrigation networks, based on the integration of remote sensing techniques and simulation modelling of water flow in the soil and in the conveyance system. Multispectral satellite images are used to infer crop potential evapotranspiration, which is the main input for water balance simulations. Irrigation flow rate is thus determined from soil water deficit to accord with the hydraulic capabilities of the distribution network as well as farmers' scheduling criteria and preferences. The development of this procedure, which is currently being implemented in a pressurized on-demand irrigation system in southern Italy, is finalized to the realization of a tool for supporting the decision-making process in the management of water resources.


In the past decades, a consistent effort has been made in the field of agriculture research to improve the understanding of physical processes involved in an irrigation system. Water transport phenomena can be accurately predicted by means of numerical simulation algorithms (Feddes et al., 1988; Santini, 1992). More recently, the spread of modelling techniques using distributed parameters has largely encouraged the use of input data from remote sensing (Azzali et al., 1991), with the support of Geographical Information Systems (GIS) for manipulating large data sets. In principle, these tools could be used for supporting the decision-making process in the irrigation water management of large districts. This involves the correct schematization of the interested areas and of the water transport processes in each part of the system. Menenti et al. (1989) and D'Urso et al. (1992) have shown the potentiality of multispectral satellite images for the appraisal of irrigation management. Information such as land surface patterns, crop mapping, identification of irrigated areas and other crop related parameters may be surveyed and monitored extensively in space and time by means of satellites. By means of the appropriate interpretation methodologies, digital images may be used to produce multitemporal maps of crop water requirements over large areas.

This information, together with other ancillary data, constitutes the input set for the simulation of soil water flow, from which actual crop water requirements can be determined. The accurate knowledge of water demand in space and in time enables irrigation engineers and managers to issue the criteria for ameliorating water distribution, by matching: (i) resources availability; (ii) structural restrictions; and (iii) farmer needs. These three management levels can be linked to each other in a 'simulation tool' (Figure 1) which could be used to maximize the efficiency of water irrigation at both district and farm levels.


On this basis, this paper focuses on a methodological framework for using remote sensing and numerical simulation methods for the improvement of water management strategies in an irrigation district. The proposed methodology is under development with reference to a pressurized on-demand irrigation system in southern Italy (Consorzio d'Irrigazione Sinistra Sele, in the province of Salerno; 40°25 N-15°5 E) with a total irrigated area of 11 km2. The main crops are forages and artichokes during the winter-spring period, while in summer there are maize, fruit trees and industrial vegetables. Aiming to improve the irrigation efficiency in a situation of increasing shortage of water resources, the old concrete channel network is progressively being replaced by pressurized pipelines; in 1994, the area served by the pressurized network was approximately 3.5 km2. Although dramatic drought conditions have not been faced yet, the actual equilibrium in the irrigation system may collapse if fast changes in cropping practices or increased water resource scarcity occurs.

The experimental campaign carried out in 1994 was planned to test the operational implementation and feasibility of remote sensing techniques and simulation models in irrigation water management. The integration of these tools in a GIS-based software package can be routinely used by irrigation managers for the improvement of water allocation strategies at scheme level by taking into account different aspects, such as:

· spatial and temporal variations of actual evapotranspiration and net crop water requirements;

· irrigation practices at farm level: scheduling preferences, average duration, pressure head requirements, soil type, climate;

· hydraulic network features and operational constraints, including water resource availability.

In the following the conceptual basis of the procedure under development will be described.


The commonly accepted 'combination' equation for estimating evapotranspiration, according to the schematization of Monteith (Monteith and Unsworth, 1990), can be used for defining the crop coefficients (Doorenbos and Pruitt, 1977) as a function of climate data (temperature, T; humidity, RH%; solar radiation, Rs; and wind speed, U) and crop parameters such as the surface albedo, a, the leaf area index (LAI) and the crop height (related to the surface aerodynamic roughness, z0):

, (1)

where a *, LAI* and hc* values are referred to the standard crop assumed for determining the reference evapotranspiration (Smith, 1990). Remote sensing techniques can be used for monitoring these vegetation characteristics; analytical elaboration performed on Landsat reflectance values evidenced the possibility of retrieving the surface albedo (Brest and Goward, 1987), the leaf area index (Price, 1992) and the canopy roughness (Moran and Jackson, 1991). On the other hand, recent work (Stanghellini et al., 1990; Choudhury et al., 1994) has shown that crop coefficients are greatly influenced by canopy development and vegetation coverage. Since these parameters directly affect the reflectance of cropped areas, it has been demonstrated that is possible to establish a correlation between multispectral measurements of canopy reflectance and the corresponding Kc values (Heilman et al., 1982; Bausch and Neale, 1987). Therefore, remotely sensed reflectance values can be used in combination with other detailed information about land uses for mapping crop coefficients in irrigated areas. In this case study, the required crop parameters, a , LAI, hc have been derived either directly from canopy reflectance or can be estimated by means of empirical relationships with vegetation indices, with the support ground reference information contemporary to different satellite passes (D'Urso and Menenti, 1995). An example of LAI estimates based on a Landsat vegetation index for the study area is given in Figure 2. The SAVI vegetation index was calculated from Landsat TM bands 3 and 4 as follows:

Once the relationship between reflectance and canopy parameters has been defined, Landsat images of the Sele area are processed for the retrieval of monthly Kc maps, according to Eq.(1).


The crop-soil system

The spatial distribution of crop water requirements can be assessed by combining the information contained in the satellite-derived potential evapotranspiration maps with numerical simulations of soil water balance. Remote sensing data and a digitalized soil map have been used as main input data for the application of a pixel-based, one-dimensional simulation of water flow in the crop-soil system. In recent years there has been quite extensive production of software for the simulation of hydrological processes and a large number of models are currently available. Nevertheless, since consistent efforts are needed for the schematization of each component of the whole hydrological system, surface and groundwater, soils, land uses, hydraulic network, the suitability of modelling algorithms to each case study is of particular concern. Several simplifications are introduced in most simulation models applied at regional scale in order to obtain accurate enough outputs without requiring too much input data and computational time. If a satisfactory description of an unsaturated zone is given, a reliable estimate of the soil moisture content in the root zone is output from the simulation run. The information concerning the soil water status allows the actual evapotranspiration to be estimated from its corresponding potential value (derived from remotely sensed data).

Simulating the farmer's irrigation criteria

For simulating correctly the hydrological conditions, irrigation deliveries must be included in water balance calculations. The detection of water stress conditions does not imply necessarily the application of irrigation. At farm level, irrigation could be decided according to completely different criteria, that may not include the soil water status. Often, in the decision-making process, the internal organization of the farm prevails on the actual hydrological conditions. Then the water demand at farm level has to face the structural constraint of the hydraulic network, mostly in terms of available flow rate and pressure head.

FIGURE 2 - Correlation between LAI and SAVI for different types of crops in the Sele area, 1994

In modelling terms, this means that irrigation occurs in a certain plot if a number of conditions are satisfied. These conditions reflect the actual preferences of farmers in the application of irrigation water (time of day, flow rate, required pressure for the delivery system, etc.). In principle, on-demand irrigation systems are regulated by two key factors:

· the depletion of water in the soil root zone;
· the farmer's water application criteria and preferences.

It appears obvious that these two elements mutually influence each other. Nevertheless, the second factor, the farmer's behaviour, seldom appears in simulation studies concerning on-demand irrigation systems, although its relevance to network performance and operation is much greater than water balance considerations. Traditionally, probabilistic criteria have been used for determining flow rates in the design stage of pressurized networks. These methods, still widely applied in the design stage, do not concern the farmer's actual needs and cannot be used for a correct monitoring of the network performance. One main reason for this neglect could be found in the difficulties in expressing farmers' needs and preferences in an analytical form. However, recent studies (Baars and van Logchem, 1993; Menenti et al., 1994) have evidenced the possibility of defining farmers' preferences in the management of irrigation water in precise quantitative terms by applying marketing-derived techniques.

In a more simplified manner, the use of questionnaires directed to farmers is able to provide useful indications for defining the rotational intervals and duration. This information can be combined with soil water budget calculations for adequately simulating the operating of on-demand networks. The following aspects should be investigated:

· factors influencing the starting time of irrigation;

· amount of irrigation water applied in relation to crop and soil type;

· flow rate and duration of each irrigation, in relation to farm facilities and available workers;

· possibility and/or willingness to improve irrigation efficiency by reducing volumes applied (e.g. with a maximum amount per season).

From the information obtained from a representative sample of farmers, the relationships between irrigation scheduling parameters and farm characteristics can be defined, also with the support of land use maps resulting from the classification of Landsat images and the digitalized maps of plot boundaries. The farmers' data and the georeferenced archive build up a customized Geographic Information System, which will be indicated in the following with the acronym FIRSA (Farm IRrigation Scheduling Archive). Essentially, the FIRSA database contains the following information for each j-th combination of crop type, extension and/or other farm characteristics (machinery, number of workers, etc.):

· the conditions to be satisfied for application of irrigation;
· the average water volumes applied;
· time and duration of applications.

From these data, it is possible to define a function qj (t), representing the average hourly water flow rate during an irrigation application for the generic plot jth. The FIRSA database is currently under definition for the case study area; for its validation, actual consumptive use of irrigation water was monitored during the 1994 and 1995 irrigation seasons at several nodes of the pipeline network.

The delivery hydrograph

The simulation of non-steady water flow in a network requires that incoming and outgoing fluxes be correctly introduced as boundary conditions. In an irrigation system functioning on demand, outgoing fluxes depend on the farmer uptake and can vary greatly with time. The water flux derived in one of the delivery nodes of the irrigation network during a generic day, d, can be expressed by means of a function Qi (td), where Q is the flow rate (m3/s) and t is the time (hours) on the day in question; this function is teremed here delivery hydrograph. The agricultural area Ai served by a generic delivery node i is composed by a set of N, plots with different FIRSA characteristics (e.g., crop type and extension aj):

, (2)

From the results obtained from soil water balance calculations on day d-1, it is possible to identify which plots should be potentially irrigated on day d:

, (3)

where Kj (=0;1) is a flag indicating whether or not water deficit conditions occur. A further selection is made on the basis of the conditions for irrigation, found also in FIRSA. If these conditions are met, we are able to identify the total area that should actually be irrigated by means of a second flag indicator, Fj (0;1):

, (4)

The delivery hydrograph is the sum of the single water demands qj from the different plots concerning A, that present water deficit conditions (K=1) and meet irrigation conditions (F=1), according to farmer preferences. The single contributions qj (t) (t = 1,..., 24) related to the crop type and the size of plot are contained in the database FIRSA. Therefore, the delivery hydrograph in the generic node i at time td of day d is given by the equation:

, (5)

In this process, it is possible to assign to each plot a priority IPj (e.g., according to the resulting water deficit or the number of conditions satisfied) related to the possibility of delaying the irrigation application.

An interface model for coupling on-farm water demand and distribution system capabilities

The previously defined distribution of outgoing water fluxes Qi (td) should correspond to a perfect on-demand irrigation system, where every farmer's need is duly satisfied. The extent of the actual realization of this task depends on the structural capabilities of the irrigation network and on the availability of water resources. These two variables are the questions that irrigation managers have to deal with every day. The interface model is an iterative procedure where the water flow distribution is progressively adjusted to the actual capability of the irrigation network.

In the Sele case study, irrigation is ensured by means of a network of pipelines. Stationary and non-stationary water flow simulation techniques have been extensively investigated in the past and they can be applied for this purpose. In this case, the equilibrium piezometric value following the distribution Qi (td) in each ith node of the pipeline network can be computed by means of iterative algorithms. Flow rate and direction in each pipeline are iteratively adjusted in order to satisfy the mass conservation equation in each node (total in-going flow equal to total out-going). Generally, a minimum value of the piezometric head at the delivery node is required for a correct operation of farm irrigation machinery. If the resulting piezometric values at certain nodes are too low, it is necessary to modify the input distribution Qi (td) in correspondence of critical nodes. For this purpose, the irrigation priorities IPj are used to screen out those parcels where the irrigation can be delayed; so doing, the distribution Qi (td) is corrected and tested again.

Once the hydraulic verification of the scheduled day has been successfully completed, the water balance calculation for the considered date is run assuming the final distribution Qi (td). The model structure is depicted in Figure 3. This interface model enables the detection of 'weak' points and also indicates the intervention that should be undertaken.

FIGURE 3 - Basic flow chart of the interface model


Since the procedure described here is still under development, it is not possible to provide preliminary results of simulation and comparison with field data. Nevertheless, it is possible to outline some considerations on the actual possibilities of these advanced techniques in irrigation water management. Although user-friendly interfaces can be built routinely using such a package, its practical implementation requires a large effort and the whole procedure is computationally heavy. The preliminary acquisition of a large mass of data input (soil type, FIRSA database) is rather complex and requires specialized intervention. Furthermore, real-time simulations are hampered by the time required for image processing from the moment of satellite pass. When the preliminary set-up of the whole procedure has been completed, the package can be run by non-specialist irrigation managers for seasonal evaluations concerning the optimal distribution of water volumes, according to the operational features of the hydraulic network and water resources availability.

The main attractiveness in using simulation techniques lies in the possibility of tuning one or more variables and evaluating the ensuing results. In this sense, the tool described in this paper may be of great value for the simulation of different scenarios according to alternative water resource management criteria. By so doing, it is possible to identify which strategy maximises the effectiveness of irrigation and the crop production. A map of spatial distribution of such indicators may be derived (D'Urso et al., 1992). Statistics about the performance of irrigation over the entire district enable the identification of critical areas where corrective interventions should be undertaken.


Azzali, S., Bastiaanssen, W.G.M., Menenti, M., de Brouwer, H., Meijerink, A.M.J., Maza, J. and Zevenbergen, A.W. 1991. Remote sensing and watershed modelling. Report 48, DLO-the W. Staring Centre, Wageningen

Baars, E. and van Logchem, B.P. 1993. Regional water resources management in the Andean region with numerical models and satellite remote sensing. DLO-Staring Centrum Report 75.1. 119 p.

Bausch, W.C. and Neale, C.M.U. 1987. Crop coefficients derived from reflected canopy radiation: a concept. Transactions American Soc. Agric. Engin. 30 (3): 703-709.

Brest, C.L. and Goward, S.N. 1987. Deriving surface albedo measurements from narrow band satellite data. Int. J. Remote Sensing 8 (3): 351-367.

Choudhury, B.J., Ahmed, N.U., Idso, S.B., Reginato, R.J. and Daughtry, C.S.T. 1994. Relations between evaporation coefficients and vegetation indices studied by model simulations. Remote Sens. Environ. 50: 1-17.

Doorenbos, J. and Pruitt, W.O. 1977. Guidelines for predicting crop water requirements. FAO Irrigation and Drainage Paper 24. pp. 40-41.

D'Urso, G., Querner, E.P. and Morabito, J.A. 1992. Integration of hydrological simulation models with remotely sensed data: an application to irrigation management. In: Advances in Planning, Design and Management of Irrigation Systems as Related to Sustainable Land Use. Vol. 2. J. Feyen et al. (eds.). Leuven, Belgium. pp. 463-472.

D'Urso, G. and Menenti, M. 1995. Mapping crop coefficients in irrigated areas from Landsat TM images. Proceed. European Symposium on Satellite Remote Sensing II, Europto, Paris; SPIE Vol. 2585.

Feddes, R.A., Kabat, P., van Bakel, P.J.T., Bronswijk, J.J.B. and Halbertsma, J. 1988. Modelling soil water dynamics in the unsaturated zone. State of the art. J. Hydrology 100: 69-111.

Heilman, J.L., Heilman, W.E. and Moore, D.G. 1982. Evaluating the crop coefficient using spectral reflectance. Agron. 74: 967-871.

Menenti, M., Azzali, S. and D'Urso, G. 1994. Adapting numerical models and remote sensing applications to the needs of irrigation agencies and farmers. Proceed. XII World Congress on Agricultural Engin., Milan, Vol. I. pp. 450-459.

Menenti, M., Visser, T.N.M., Morabito, J.A. and Drovandi, A. 1989. Appraisal of irrigation performance with satellite data and geo-referenced information. In: Irrigation Theory and Practice. J.R. Rydzewsky and K. Ward (eds.). London. pp. 785-801

Monteith, J.L. and Unsworth, M.H. 1990. Principles of Environmental Physics. E. Arnold Ed., London. 290 p.

Moran, M.S. and Jackson, R.D. 1991. Assessing the spatial distribution of evapotranspiration using remotely sensed inputs. J. Environm. Qual. 20: 725-737.

Price, J.C. 1992. Estimating Leaf Area Index from Remotely Sensed Data. Proc. IGARSS '92 (Houston). Vol. 1. pp. 1500-1502.

Santini, A. 1992. Modelling water dynamics in the soil-plant-atmosphere system for irrigation problems. Excerpta no. 6. 1991/92.

Smith, M. 1990. Expert consultation on revision of FAO methodologies for crop water requirements. FAO Report, Land and Water Development Div., Rome. 54 p.

Stanghellini, C., Bosma, A.H., Gabriels, P.C.J. and Werkhoven, C. 1990. The water consumption of agricultural crops: how crop coefficients are affected by crop geometry and microclimate. Acta Horticulturae 278: 509-515.

Previous Page Top of Page Next Page