Digital Knowledge Ecosystem for Agribusiness: Transforming Agriculture
Two of the main causes of low efficiency in agricultural production are coordination failure and uncertainty due difficulties in obtaining right information at the right time. Thus a key condition for farmers to be included in successful value chains is that they have access to market information and possess the ability to translate it to market intelligence. Thus in order to improve efficiency agricultural sector needs to transform itself from fragmented production and marketing relationships towards an integrated market system.
Unlike in any other industry production lag associated with the seasonality of agricultural crops has created a unique problem for market actors. The synchronization of different activities carried out by different actors, from the time of cultivation to time of selling, has to be well coordinated. Failure to do so would result in either an oversupply or undersupply of commodity in the market.
The digital knowledge ecosystem (DKES) that we developed and deployed in Sri Lanka addresses these key problems that farmers and other stakeholders face: agriculture commodity over and under production resulting from an uncoordinated market and the root cause for this; lack of access to right information at the right time to derive the necessary market intelligence. This new system has evolved over the last five years and is now poised to transform agriculture by linking major stakeholders and providing them with relevant context-dependent actionable information at the right time to create a well-coordinated system.
This brief article outlines the evolution of the DKES and highlights its unique features.
In 2011, we embarked on a new project to help address the problem of overproduction of agricultural produce common in Sri Lanka. Soon we realised that overproduction is only a symptom of a much deeper issue: farmers as well as other stakeholders in the agricultural domain not getting right actionable information they need at the right time. To address these problems, following some in-depth research, which has now resulted in 4 PhDs and a Masters Honours theses, we developed a Mobile based Information System (MBIS) to enhance flow of information in the agriculture domain. Our research also highlighted that for information to be useful to market actors it needs to be provided as context-specific actionable information. The term actionable information refers to such information that will enable a stakeholder to act with least amount of further processing. The research challenge was to discover a way to generate context specific actionable information to support decision making throughout all stages of a farming cycle in a way to empower users and create a well-coordinated market.
To meet information needs of market actors we need to generate two types of actionable information; quasi static and dynamic. Quasi static actionable information can be derived from published domain literature to answer questions such as “what crops will grow in 'my farm'?", "which fertilizer to use for ‘my crops’ and when to apply?", and "which pest is destroying ‘my crop’?”. As we provide customised information to suite ‘my farm’, ‘my crop’ etc. farmers can immediately act on this information eliminating additional cognitive processing required if more general information was provided.
To answers questions such as “how much carrot has now been planted in ‘my district’?”, “in which market can I get the best price for ‘my beetroot’?”, “where can I buy fertilizer for ‘my crop’?”, “what is the best option to get a loan to buy fertilizer for ‘my crop’?” we need to capture data in-real time and generate dynamic actionable information.
To provide actionable information to such questions, first we developed a context model to represent ‘my farm’ in terms of soil, climatic and geographical parameters as well as past history of crops grown in the farm. We then developed an agriculture ontology to reorganise published crop knowledge, and based on this ontology, created a knowledgebase to deliver context specific actionable information through the MBIS. The following mobile interfaces illustrate steps involved in crop selection -- how a farmer can find what crops will grow in his farm.
After login, the farmer choses “Crop Selection” from the menu. Next, the farmer can select a farm already added or add a new farm. Adding a farm is easy and can be done by providing the location via set of drop- down lists or by identifying the location in a map. Based on the location information provided, the backend Geographical Information System (GIS) computes the corresponding agro-ecological zone and provides appropriate parameter values to specify the farm context. Based on the farm context, a list of crops that will grow in such an environment is displayed on the interface. Farmers can then get specific varieties for each crop type and other related information about the crop variety.
Besides advice on crop selection, farmers also want information such as current level of production for various crops, market price of produce, information to obtain finance, price of agricultural inputs and stores that supply them. This information has to be generated dynamically and updated regularly. To provide this feature, we developed a stakeholder-centric agriculture information flow model. After farmers get a list of crops that will grow in their farm, they would like to know how much it will cost them to grow the selected crops. An expense calculator that we created provides this information. Farmer needs to select a crop and provide the planned extent. The system then gets information on fertilizer and pesticide requirements and other costs from the ontological knowledge base, prices provided by input suppliers and computes an approximate cost of growing that crop. Based on this information a farmer decides what crop he would grow in the next season. Now the system in addition to ‘my farm’ also knows how to interpret ‘my crops’. This is essential to generate actionable information as it needs to be customised for each user. The figure below shows the interactions involved in estimating the cost of production.
The system then systematically guides the farmer throughout the growing phase. After a few farmers use the system, the system knows the types of crops and extent to which they are grown by a sample of farmers in a given region. Then, using a predictive algorithm, an estimate of the total production of a crop in a region (say, district) is computed. These aggregated anticipated production information is also provided to the farmers to answer the question “how much a crop (carrot) has now been planted in ‘my district’?”. This will assist a farmer who is in the process of selecting a crop to grow to avoid over-production as well as to the government to plan necessary agriculture products to be imported in the next few months to maintain food security in the country.
The system also computes the demand for various types of fertilizer and chemicals in the next few months on a district basis and provides this information to their suppliers to plan their supply chains. This seamless flow of information acts as an incentive for input suppliers to provide price of their supplies and location of their retailers to the system. The input suppliers also wanted a facility for farmers to order their inputs via the MBIS. The banks then got interested in providing micro finance through the system to help farmers that have some cash flow problems. Now a major super market chain is joining the system to establish forward contracts with farmers. Thus a well-coordinated system to manage the overall agriculture production evolved.
Based on a psychological empowerment framework, we designed the mobile front end that motivates stakeholders to act on the actionable information, capture the transaction data and generate new actionable information needed by all stakeholders.
The initial mobile based information system over time evolved into a Digital Knowledge Ecosystem for agriculture domain connecting all major stakeholders minimising the information asymmetry. This system was deployed as a full functional research prototype in Sri Lanka in April 2015.
The government of Sri Lanka has made this Digital Knowledge Ecosystem a national project in November 2015 and will be used to coordinate the overall agriculture production in the country. This will enable transformation of the uncoordinated agriculture practices into a coordinated Agri Business value chains. The diagram below shows at a conceptual level the way this coordination can take place. As explained earlier using the evolved system, extent of cultivation of a crop can be obtained in real time. Based on this information the Government can vary incentives such as guaranteed prices for harvest and available subsidies for fertiliser etc. dynamically through policy setting to obtain a better balance between supply and demand.
Same concept can be used by different stakeholders and businesses in the agriculture domain to create their own value chains. At present four such major stakeholders in Sri Lanka are working with us to integrate their specific value chains into the Digital Knowledge Ecosystem.