Decision support system for nitrogen fertilization using fuzzy theory A. Papadopoulos a , D. Kalivas b,⇑ , T. Hatzichristos a a National Technical University of Athens, 9, Iroon Polytehniou Str., 15773 Zografou, Athens, Greece b Agricultural University of Athens, 75, Iera Odos Str., 11855 Votanikos, Athens, Greece article info Article history: Received 10 November 2010 Received in revised form 10 June 2011 Accepted 11 June 2011 Keywords: Decision support Site specific management Fuzzy logic Fertilization Nitrogen Cotton abstract During the last three decades there has been great concern about the impact of agriculture on the envi- ronment and its resources. Conventional agriculture is based on whole field and mostly empirical approaches to defining and applying agrochemical inputs, which poses certain limitations regarding the management of existing variability in agricultural land. In this paper, the design and application of a fuzzy decision support system, concerning site specific nitrogen fertilization, is described. The system uses an easy but efficient way of solving the nitrogen equation under agricultural conditions and is based on knowledge elicitation and fuzzy logic methodologies. More specifically, the system is composed of two parts; a knowledge base and an analytical modular part which simulates nitrogen balance. The analytical part is built in a four level structure which consists of eleven fuzzy systems. The evaluation of the system presupposes the availability of 14 state variables that can be easily collected and refer to characteristics of the soil, weather and farming practices. The incorporated knowledge and the formulation of fuzzy rules were based on interviews with experts and on annotating scientific and technical bibliographic resources. A sensitivity analysis of the developed system was carried out in order to evaluate its robustness against errors or uncertainty in the state variables and further to assess and highlight the important variables. The application of the system using a set of point data, drawn from cotton fields in central Greece and stored in a Geographical Information System, is described in brief and the results show considerable var- iability in the recommended amount of nitrogen fertilizer among the reference sites. Ó 2011 Elsevier B.V. All rights reserved. 1. Introduction Various negative impacts of current fertilizing management such as pollution of ground and surface water bodies have become apparent and there is great concern about effective methods of addressing them. Research has so far demonstrated that significant mistakes have been made in the use of nitrogen (N) fertilizers, resulting in low nitrogen use efficiency (Torbett et al., 2008). Much of the excess N is lost from the system, through leaching, denitrifi- cation and volatilization. The ability to estimate the economic and environmental optimum rates of N fertiliser application is there- fore essential for efficient fertiliser management. In agricultural management, the constitution of a fertilizing prescription that is adapted to the special natural and anthropo- genic conditions of a specific area is of great importance. Site-spe- cific crop management (SSCM), as opposed to conventional agriculture, offers a better and more justified approach to applying agronomic inputs (Plant, 2001). In Europe, the application of site specific crop management includes well-developed sites in the UK, Eastern Germany and Denmark. In Greece, the potentials for fully integrating the dimensions of site specific crop management, as met in the strict ‘‘sub-field’’ context of ‘‘precision agriculture’’, are limited due to the small size of fields and limited revenues (Gemtos et al., 2002). Naturally, this raises the question of whether farming in this form is really needed or if there are important rea- sons for such restrained use. However, fertilizing management on a regional scale can and should be a realistic objective for practising agriculture under Greek farming conditions. The lack of development of appropriate decision support sys- tems (DSS) is thought to have hindered the full adoption of SSCM (McBratney et al., 2005). A DSS can be regarded in general as an interactive, flexible and adaptable computer based information system especially developed for supporting the recognition and solution of a complex, poorly structured or unstructured, strategic management problem for improved decision-making (Keen and Scott-Morton, 1978). Farmers are engaged in adaptive manage- ment in a highly variable and unpredictable environment and therefore no farm (or farmer) is the same. For that reason, a suc- cessful DSS in agriculture should focus on specific strategies, main- taining its ability to integrate them in a holistic managerial scheme. Agricultural DSS of many kinds have been described in the scientific literature (Antonopoulou, 2003). The majority of these DSS include a core model based on simulation models or optimization methods in combination with probabilistic methods. Either way, the reported DSS’s structure adopts discrete values and is built under crisp theory. 0168-1699/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.compag.2011.06.007 ⇑ Corresponding author. E-mail address: kalivas@aua.gr (D. Kalivas). Computers and Electronics in Agriculture 78 (2011) 130–139 Contents lists available at ScienceDirect Computers and Electronics in Agriculture journal homepage: www.elsevier.com/locate/compag