Agricompas

Agricultural Data Analytics: Knowledge and decision support

Technology

Automatic satellite image processing for farm intelligence

by

By Simon Dzurjak, Data Analyst

Satellites orbiting the Earth allow us to gather insights about farms and the crops they are producing at unprecedented spatial and temporal resolutions. This is one of the key features of Agricompas’ approach to digitising agricultural value chains and providing value to the sector.

Remote Sensing and Big Data

Remote sensing and Earth observation have not been left behind in the current big data revolution, with a large and ever-growing number of both open and commercial datasets providing years, sometimes decades, of historical data, ready to explore using big data techniques.

At Agricompas, we are using imagery from satellites such as Sentinel-2 to monitor farms and develop models to predict the crop phenology, health and yields. This is in combination with ground-sourced data such as weather stations and soil tests as part of our EcoProMIS platform.

Processing Satellite Images

Whilst satellite imagery in its raw form is intriguing to look at, it does not tell us enough to be immediately helpful, therefore it must be refined further into various indices and biophysical parameters.

Crops are complex, living, plant organisms which interact with various wavelengths of the electromagnetic (EM) spectrum in a variety of ways. When monitoring crops using satellite imagery, we can use the various reflectance and absorption characteristics of plants in the visible and infra-red portion of the EM spectrum to our advantage and use mathematical methods to derive vegetation indices, each of which highlights a certain property of the crop.

Satellite image and analysis layer for palm farm in Colombia

The vegetation index data derived from satellite images can then be correlated with various biophysical parameters of the crops and insights about crop health and potential yields can then be modelled and monitored, giving farmers the insights they need.

Real-time Data for Farmers

Over the last few weeks, a large pipeline to mine valuable data from satellite images has been under development at Agricompas. The products derived by this pipeline will soon be used in machine learning models and crop modelling to develop accurate predictive models.

We are now in the final stages of putting finishing touches on the pipeline and testing it, ensuring that everything runs smoothly. Soon, the pipeline will be deployed to our cloud environment, where it will run automatically, and continually process the satellite images.

This will allow us to process an image as soon as it becomes available to us, letting us always have up to date and quick results about the farms we are monitoring.

Additionally, we are making sure that the latest imagery of the farms we are monitoring is available to the farmers via our knowledge services application.

We are extracting individual farm data from the large satellite images, so that the farmers can visualise the farms they are interested in, and do not have to manually locate their farm in the satellite image scene.

Practical Application

Farmers and other users will be able to visualise both true-colour imagery, as well as derived indices such as NDVI, giving a bird’s eye view of farms, highlighting issues such as unhealthy crops and allowing for quick responses before these issues turn into problems.

Automating the mining of the satellite imagery for valuable data has been a very exciting project to work on and brings a lot of additional value to the EcoProMIS platform. We are continuing to make sure that all of the data we work with is accurate, with advanced models producing reliable results, and supporting our users in undertaking the best decisions.

Agricompas and Information Workers working in a big data project for the Agro-Industry

by

by Carlos Torres, Senior Consultant, IWCO

Information Workers, or IWCO, is an important technical partner working on the EcoProMIS project led by Agricompas.

A Colombian company with 10 years of experience, IWCO is focused on helping its clients get the most and the best value from their data by turning it into information that allows them to make decisions in the short, medium, and long term.

This value comes from four fundamental areas in the process of data harnessing: exploration, extraction, refinement, and consumption of data.

Data Exploration
This phase is designed to understand the types of data, their characteristics, where they are located and their potential value, additionally, the identification of business needs that can be solved with data.

Data Extraction
During the extraction process, we help our clients (such as the EcoProMIS project) to take their data, put it into a structured format that can be simple to use and that enables our clients to immediately create on-demand queries to get answers to their business questions.

Data Refinement
Once the data is in a structured format, we help our clients design and create models to predict, understand, and even extract hidden patterns in the data.

Data Consumption
Business users such as Agricompas and Pixalytics are extremely important, that is why we care about helping them consume their data models – information that can now be prepared for analysis. Through knowledge transfer processes we enable our clients to obtain answers to their strategic questions.

Information Workers is a Microsoft, AWS (Amazon), and Google Partner. With broad experience and a process focused on culture, we have an emphasis on helping organisations to operate in the world of self-service.

Information Workers (IWCO) have a multidisciplinary team

We have a multidisciplinary team that includes Mathematicians, Statisticians, Economists, Software Engineers, Systems Engineers, and even Petroleum Engineers (!). This diversity means that our team is able to bring our customers’ different professional perspectives, something that has helped us contribute value in the EcoProMIS consortium of seven international partners.

Together with Agricompas, we created a plan designed to resolve some of the challenges for the EcoProMIS project, namely how we acquire, store, manage, and secure the agricultural analytics data.

After developing this plan, it was essential to choose a platform aligned with these necessities pragmatically, and allow us to obtain results in a short period. Another important step was to identify the sources of big data for the project, for example publicly available data (e.g. earth observation images from UKSA), streaming data from different sensors like weather stations, and data from legacy databases.

Big Data
As we engaged with all of this data, it became obvious that we could not process this multivariate data using traditional methods. The data from the EcoProMIS project is clearly in the realm of big data, complying with the three ‘V’s:

Volume: EcoProMIS has collected a lot of data from different sources.
Velocity: The various data streams are handled within different timeframes from the different sensors.
Variety: All types of formats – from structured to unstructured.

Adding Value
Another key step in our contribution to the project was to find a way to access, manage, and store the data. We needed to find a data platform that can support the storage and the capabilities of analysing petabyte-size files and trillions of objects.

Equally important was prioritising security and data protection, making sure the database is GDPR-compliant, and that data is securely stored. Linked to this is our role in providing data auditing and ongoing support.

Agro-Industry
Big data in the agro-industry plays an important role. With such a large amount of information out there, the data needs to be shaped or tested in a way that adds value to the agro-industry.

By doing so, the agro-industry can better identify problems and reach the goal of sustainable optimisation that is at the heart of the EcoProMIS project.

Innovation, growth and data analytics

by

By Nicolás González, Business Development Manager

The world is changing. Every day we see how new technologies are being developed and many of us may have come to dread the idea of being displaced by a machine. But that is not the way of things. Technology is meant to make our lives easier, and we believe that the coexistence between traditional methods and newly developed ones is possible, if not meant to be.

In recent years, agricultural data analytics has become one of the top edge tendencies in terms of sustainable development globally. This means that there is a lot of research and a lot of projects currently trying to understand agricultural dynamics and how to use information in order to optimize processes and achieve sustainable objectives.

Although this is a very beautiful statement, the reality growers are facing on a daily basis, and how this information is to be gathered, processed and used, represents a huge challenge not only for farmers but for the whole agricultural value-chain.

Agricompas drones collect images of farms for innovative data analysis

Agriculture in LMIC countries still relies on manual labor. The culture of innovation, technology, insurance, good financial practices and data analytics is still in a juvenile stage.

Nevertheless, governments and private capital are incentivizing fast growth through technology and new services are becoming more common every day.

Innovation and Market Growth

For us as an agro data analytics company, innovation is the only way to break through such barriers and use our knowledge to evolve the growing market. This is our main drive, our oxygen and our compass. As Harvard Business Review’s article: Breaking down the barriers to innovation states:

To us, innovation doesn’t mean mere inventiveness. In our work we define it as: something different that creates value.

As we work with agricultural big-data, we face the challenge of gathering, processing and delivering useful information in the context of market needs and opportunities. By doing this, we look to make our EcoProMIS platform a value generator for growers, governments, and financial and insurance institutions.

Analytics for Decision Support

In order to capture what really drives the agricultural market in terms of financial services, risk management and productivity optimization, our innovation process aims to understand the market and develop tailored solutions that make the decision-making process more efficient, thus giving business intelligence the recognition and merit it deserves for the immense toolkit it provides us with.

We believe that by bringing sustainable agriculture into the digital era, better conditions for growers may be achieved. We believe that when real value is generated it can also be garnered.

We believe that corporate institutions can see a benefit as well, using high quality information and business intelligence, improving their margins, creating new delivery methods, enhancing R&D and, finally, increasing sales.

Our EcoProMIS platform sends farmers notifications using a suite of apps

With the richness of precise, accurate and relevant information, we enable an increase to the market size of agro insurance and provide the much-needed agronomic crop management data necessary for a new and creative product development ecosystem.

The world is changing. And so are we. Breaking through the barriers of convention, we have come to innovate and leave the world in better shape than how we found it. We believe that a new era for sustainable agriculture and analytics has come at last.

Connecting growers with an essential service in the modern world

by

By Richard Strange, Head of Engineering at Agricompas

As individuals we all grow in wisdom and capability when we take time to reflect on our actions and find lessons to apply to tomorrow’s challenges. We look at our achievements and the memories that are anchored around them. We use them to guide us in becoming better in both our professional and personal lives. Modern business is very much the same. In the modern world, when a business moves, the byproduct of their actions is data.

Agricultural Data Gap

Whether it is a financial officer’s log of transactions, the record of work hours from an employee’s timesheet, or the number of clicks a website receives each day. It is rare to find a part of a business that isn’t measured or collected, either directly or by proxy through other measures. Yet in agriculture, little information is available around many crucial farming practices that often mean the difference between a bumper crop or financial devastation for families and communities.

It is not enough to say that you have an employee, or a website, or an invoice. The crucial questions are if the employee is doing their work, if the website is drawing attention, if the invoice is correct. Yet farmers are not able to answer critical questions about their own farms. They have sown their seeds, yet cannot say how many are germinating. They apply fertiliser, yet cannot tell if it is cost-effective. By leaving agriculture behind in this wave of data-driven business, the world is abandoning millions of farmers in data poverty, and powerless to compete against their wealthier first-world counterparts.

EcoProMIS Collects Quality Data

In leading the EcoProMIS project, the aim of Agricompas is to make a difference by empowering farmers with the knowledge they need, from sensor to survey to satellite to weather to drone data. But with each additional source of data, the difficulty of pulling them together increases exponentially. I’m the Head of Engineering at Agricompas, and I’m responsible for all the data EcoProMIS gathers. My job is to work out how we pull all this information together, understand it and then provide the information to those that need it.

There are two approaches to tackling a challenge like ours. Firstly, you can manually handle the data, with a team of analysts pushing round files via email, shared folders and collaborative spreadsheets. This does come with the advantage of immediate productivity and visibility. But there’s little certainty over the quality and completeness of data, and no way to be sure what information is where. The second option is to invest time and effort into a fully-fledged platform for data. It must allow the scientists we work with and the farmers that we support to put in and take out the information they need effortlessly.

Advanced Data Platform Prevents Errors

Only recently, the failure of the first, manual approach was highlighted by the loss of the records of 16,000 positive COVID-19 cases by the UK government. Was it a catastrophic server failure? the act of a malicious hacker? The truth was far more mundane. An analyst had opened the spreadsheet holding the list of COVID-19 cases in an old version of Excel, slicing 16,000 rows of data off without ever realising their mistake. Suppose this approach cannot work reliably in the hands of a team as well-staffed as the Public Health England team. How can we trust our own information in a similar system? We owe our growers and our own team better than that.

The cost of getting your data platform wrong (https://www.bbc.co.uk/news/uk-54412581)

Over the last six months, the EcoProMIS team has been carefully creating a central platform that can look after farmer data responsibly and safely. A system of databases, redundant servers and security measures means that data doesn’t get forgotten, doesn’t get destroyed and doesn’t get leaked. Over the coming months, we are combining our suite of analytics, models and AI with new apps.

These apps will allow farmers to provide and see their data about their farms and help them make the right agricultural decisions. We already have the first app in early tests, with a knowledge presentation app in the works for release by the new year.

Connected Growers

As we evolve our platform and grower apps through close feedback with early users, we will be able to put more power back into the hands of growers, irrespective of their literacy or agricultural experience.

Agricompas and the EcoProMIS project exist to level the playing field and make agriculture fairer for farmers in the most challenging economic, environmental and social settings. I am incredibly proud of what our technical team has achieved to make that happen.

Why the eddy covariance technique is an ally in the search of sustainable agriculture

by

By Agricompas Crop Model Team

In recent decades, agriculture has been under the scrutiny of society and the scientific community due to the negative impact that it generates on the environment. These impacts are of many types, including deforestation, eutrophication of water bodies, the reduction of biodiversity due to the intense use of pesticides, and the emission of greenhouse gases (GHG).

In relation to GHG emissions, these are released in the process of manufacturing inputs, such as, fertilizers. Also included are the GHGs released as a result of the transport process: first of inputs towards the production areas, and then of the product towards the consumption areas.

An eddy covariance system recording greenhouse gases emissions on a commercial rice field at Colombia for EcoProMIS project. (Agricompas)

A Complicated Task

But the most complicated task from a methodological point of view is to determine the GHGs that are released during the production stage. Among the GHGs released to the atmosphere during the production phase, the most important are carbon dioxide, methane (in systems where the soil is in anaerobic conditions), nitrous oxide, and ammonia.

Methodological difficulties are associated with the fact that these emissions are determined by dynamic factors such as climate, soil characteristics, and management practices, especially fertilization and irrigation.

Since it is impossible to survive without agriculture, efforts have focused on developing and implementing production systems able to maximize yields while reducing negative effects on the environment. A prerequisite for advancing in this direction is to measure the GHGs generated during agricultural production cycles.

Static Chambers

To understand better the methodological challenges involved in determining these gases under field conditions, let us take methane as an example. This gas is generated as a product of the decomposition of organic matter in the soil under non-oxygen conditions, typical of crops such as flooded rice.

Traditionally, static dark chambers have been used to collect samples that are later analysed by the gas chromatography technique in specialized laboratories.

This technique has a high sensitivity to determine low methane fluxes, is easy to handle, and has a low cost. But its main disadvantages are related to the low spatial representativeness and the inability to generate data at different time scales.

In other words, the measurements only represent the gas flux in a small area and at a specific time point, which leads to the question: can this technique generate data to represent what happens in inherently heterogeneous and dynamic agricultural systems?

Eddy Covariance

It is in this context that the technique of eddy covariance appears, as an alternative way to measure, among other variables, methane flows with greater spatial and temporal representativeness.

This technique employs a complex assembly of sensors arranged in a tower (which is why they are usually called eddy covariance towers) that records variables that ultimately allow the determination of the exchange of gases and energy between the crop and the atmosphere.

Although the foundations of the technique and data processing are complex, it provides useful information in the search for more sustainable agricultural systems.

This is because, in addition to determining GHG emissions, such as methane and carbon dioxide, the eddy covariance technique also provides information about the flow of energy between the soil, the plant, and the atmosphere. This means that information is also useful to improve the water use efficiency since the measurements allow the determination of water fluxes from the crops to the atmosphere (evapotranspiration).

All of this information is comparable in terms of accuracy with data obtained by reference instruments such as lysimeters. Therefore, the technique of eddy covariance is currently a powerful ally in the search for more sustainable agricultural systems.

Use with EcoProMIS

The EcoProMIS project has four eddy covariance towers in Colombia, two recording data on rice crops, and two on oil palm crops. The data collected by these stations are being processed to calibrate crop models that allow, in addition to predicting yields, to estimate GHG emissions.

Together with our partners (CIAT, Cenipalma, Fedearroz, IWCO, Pixalytics and Solidaridad), the final objective of the project is to generate “knowledge and decision support” to orient stakeholders towards sustainability.