big data for rice systems in latin america
TRANSCRIPT
Big Data for Climate Smart Agriculture
Enhancing & Sustaining Rice Systems for Latin America and the World
Introduction
Rice is the most important food crop in the world providing more energy to humanity than any other food
source. Latin America and the Caribbean (LAC) is a net importer of rice mainly because LAC farmers lack
adequate knowledge and current information to adapt their cropping systems to increasingly variable climate.
Recent climate change analytical studies by the WBG revealed that LAC could benefit from a more suitable
future climate for rice. New approaches are now urgently required to provide farmers with updated, relevant,
and near real time climate and cropping system information to support them in rice cropping decision making
and to enhance their resilience to climate variability and eventual climate change.
Figure1: Rice as a supplier of energy in the human diet and Global Rice Yields over time.
Problem and Opportunity
Climate change is not only about long term temperature and rainfall trends. It has already altered the patterns
of climate variables, especially in terms of variability, that were reasonably well understood and predicted by
farmers and communities worldwide. For example, in Colombia rainfall and daily maximum and minimum
temperature patterns, have changed in each region and the climate is less and less predictable.
Agriculture is greatly influenced by climate. Any perturbation of the climate directly affects production if
farmers are not able to adapt their cropping system in time. In Colombia, national average rice yields have
dropped from 6 to 5 t/ha in less than five years without changes in crop location (soil types) or in management
(which is actually improving). Traditionally, farmers across the world have used calendar references to make
decisions on when and what to sow. Nowadays, due to the increased climate variability, however, this
traditional knowledge is increasingly ineffective and farmers lack information to make appropriate decisions in
a new fast moving environment.
The International Center for Tropical Research (CIAT) in Cali, Colombia has recently tested an innovative ‘Big
Data’ approach to create a dynamic Decision Support System for rice farmers in Colombia that involves:
(i) A two way information system based on a web platform and an android app to support field data
capture,
(ii) An artificial neural network (ANN),
(iii) Clustering of data techniques for rapid analyses of harvest monitoring data
(iv) A coupled platform for accessing climate information at daily intervals, and
(v) A cloud-hosted, relational (SQL) database.
How the Big Data Concept was Operationalized for Rice in Colombia
Nowadays smartphones, cloud-computing and other ICTs make it possible to capture, analyze and share large
amount of information very quickly. In Colombia,
a) FEDEARROZ have been collecting information on commercial harvests for almost 20 years in the main
rice producing areas of Colombia, mainly for economic-studies and productivity monitoring purposes.
On the other hand, IDEAM (Colombian national meteorology institute) have also been capturing
climate data through a nationwide network of weather stations.
b) We combined those two databases, relating each individual harvest event to its corresponding ~120
days climate sequence between sowing and harvest for five main climate variables: minimum
temperature, maximum temperature, precipitation, relative humidity and solar radiation at daily
intervals.
c) We analyzed the data using machine learning techniques such as Artificial Neural Networks (ANNs),
Random Forest, Clustering so as to unravel underlying correlations patterns between climate factors
and yield variability that could help us identifying the combination of factor that result in high yields.
d) The analysis of the commercial data coupled with weather data generated valuable insights for rice
producers. The identification of the main limiting factors in each region allows the farmers to
understand why they got high or low yields in past years. The characterization of
favorable/unfavorable climate sequences and the match with seasonal forecasts allows learning from
past experiences to anticipate what is coming and to give advice to farmers on what variety should
work better, and when is the optimum sowing date.
e) The tool has the potential to compensate the loss of traditional calendar-based cropping system
interventions by providing the farmers with relevant data-driven information for decision making on
what, when, where to sow and how to manage the crop.
In pilot tests with Colombian rice farmers in two provinces, the methodology generated valuable insights
about the local rice cropping system by using available commercial data at two locations in Colombia. To the
surprise of local farmers, the new data system generated advice against planting rice in that particular season
due to projected adverse climate impacts. Farmers that followed the advice saved seed, labor, and fertilizer
and water inputs. Farmers that did not heed the advice, harvested nothing and lost all their inputs! Based on
this pilot outcome, the CIAT team and local rice association partner (FEDEARROZ) were awarded a UN Big Data
award. [http://www.theguardian.com/global-development/2014/sep/30/colombia-rice-growers-climate-change]
Figure2: Transforming Latin America from a Rice Importing to a Rice Exporting Region
Scaling Out and Up the Pilot Big Data for Rice Approach Developed in Colombia
The next step is to include soil and crop management factors and to scale out and up, the Colombian Big Data
pilot system for rice farmers in other countries in Latin America. We propose to target the big rice producers
Argentina, Brazil, and Uruguay via the establishment of an integrated partnership with the major regional
‘end-user’ association that has extensive networks with rice producers and the associated supply chains
throughout Latin America - the Fund for irrigated rice in Latin America (FLAR). For more than 19 years, FLAR
has gathered rice production and processing related data from all main rice producing countries in LAC and
their respective supporting organizations.
Next Steps: To take the approach to a new country or a new crop, requires the following main steps:
Involve from the very beginning the community of targeted end users of the tool in the design
and conception. It provides valuable feedback and facilitates the final adoption.
Undertake an initial diagnostic of available data and needs for complementary data capture on
climate, soil management and yields.
Set up an analysis team able to handle the data and analyze it to generate relevant information
for crop optimization.
Carry out a pilot case-study to demonstrate the value of the approach.
Accompany the dissemination of the method to all potential users
Expected outcomes:
Foster a data capture culture in agriculture, using ICTs
Adoption of the approach in routine work of farmers and rural advisory services to move
towards data-driven agronomy and climate smart agriculture.
Expected impacts:
Reduce yield losses due to uninformed decision making,
Increased adaptive capacity for agriculture to climate change
More efficient crops (bridge the yield gap) for more food with less land and inputs
Revolutionize agricultural advisory and extension models by transitioning to a robust, near real
time data-driven agronomy, and more site-specific recommendations.
Adapt the system to other crops (e.g. maize, beans, and horticultural crops).
Explore pilots for African rice and fruit growers (West and East Africa)
Contacts:
CIAT Team (Cali, Colombia): Andy Jarvis ([email protected]) and Daniel Jimenez, Sylvain Delerce,
Armando Muñoz, Hugo Dorado, Juan Felipe Rodriguez, Victor Hugo Patiño.
World Bank Team (Washington, DC & Brazil): Erick Fernandes ([email protected]) and
Renato Nardello, Holger Kray, Diego Arias Carballo