We aim to have the most complete, highest quality and best maintained global supply chain database in the business, as one of the critical components of our service.

It took us many years to gather, curate and/or accurately digitise all data for our global database of commodity production unit boundary and ownership datasets. We actively invest in gathering supply chain data through:

1. Our own field teams and cooperation with local GIS consultants and mapping companies, which includes visiting local government offices to collect and digitise paper maps, validate data, and ground checks;

2. Partnerships with local and international NGOs, auditors and other partners, Solidaridad, Rainforest Alliance, IDH, Grepalma, Fedepalma, etc;

3. Our clients; growers, traders, consumer goods companies and their partners. Note that data can be under Non Disclosure Agreement (NDA). We broker agreement between supply chain partners to share data or derived information securely;

4. Cooperation (as of lately) with forward thinking government agencies.

The main challenge to overcome has been to establish best practice procedures for quality control and maintenance of this data.

Piecing together all data is based on trust and a very time consuming and costly effort. Note that the data has to continuously be maintained and updated where required.

Satelligence uses a wide range of sources to achieve traceability to plantations and beyond. The most important condition for achieving full traceability is not technical but human: building trust and interaction between supply chain partners. This takes time.

We continuously update our information with the latest updated information. Our approach includes:

• Insights from our network of local experts (clients and partners);
• Data from public sourcing (mill and refinery) lists;
• Continuous data ingestion from online sources including campaigner reports (eg Chain Reaction Research, Mighty Earth).

On a case-by-case basis we have been implementing the following sources:
• Anonymous mobile geolocation data (i.e. machine learning approach finding exact matches between signals of the same telephone at a plantation and at a mill within 24 hours). Remark: working with mobile service leaders in Asia we find this provides 8-20% of linkages at most. As such, it should not be considered a silver bullet, but all bits of information may help if collected responsibly;
• Local truck movement analysis by local partners;
• Bills of lading and other customs datasets.

Yes! We actively work with Solidaridad, Fairtrade International, Rainforest Alliance and others to enable smallholder inclusion on international markets. Our 3-7m detail coverage worldwide is perfectly suitable for monitoring smallholder farms. We monitor 3.5 million smallholder farmers.

Let’s work together to support the other 36.5 million!

Satelligence is in continuous contact with campaigning NGOs and peers to discuss the validity of new grievances.

Secondly, we work in an exciting partnership with Ulula to offer combined environmental, labor and social monitoring.

For social risk monitoring we can now integrate multichannel and multilingual stakeholder engagement technology to amplify local voices (crowdsourcing information from workers) and provide organizations with social, labor and community impact data.

Data and insights are accessed securely via the Satelligence platform, or can be easily integrated into your existing software systems and workflows using our high quality data feed (API).

We provide a dedicated Platform with enterprise level security, connecting to your organization’s SSO server. Single sign-on (SSO) is a session and user authentication service aligned with security guidelines ISO 27001, NEN 7510, BIO, SOC 2 and GDPR/AVG.

We offer flexibility for integration with your other supply chain management (ERP) software such as SAP, Oracle, Accenture, and other.

Our software and infrastructure have been meticulously assessed and stress tested by some of the world’s biggest corporations, including Cargill, Rabobank and Unilever. As part of our ISAE 3000 audit process and operational best practice, we regularly assess security status.

We consider both industry consensus and the national forest definitions to guide definitions of forest and deforestation. The area parameter is important, for example: Cote d’Ivoire >0.05 ha, Indonesia >0.5 ha, Suriname >1 ha just to name a few examples. One-size-fits-all monitoring is not correct.

Moreover, we categorize our landscape baselines into e.g. primary and secondary forest following criteria in certification standards (e.g. identifying ‘primary’ forest, as Greenpeace IFL promotes). We do this by analyzing the full archive of imagery since 1980s, so we know which areas have been clearcut and regrown, or otherwise degraded historically.

Our system labels deforestation or fire or other risk, based on any previous land cover type or HCS carbon density, calculating statistics automatically for any administrative region from continental-country-state-district-group/coop up to individual farm.

Yes. We use both curated and corrected open data, government data and data developed in-house. What matters is to strive for the best quality, while remaining pragmatic and getting all stakeholders aligned.

We include official government data after careful curation. Experience shows some government forest and land cover layers are hand drawn and should be used with caution.

We use a combination of publicly available radar and optical sensors: Sentinel-1, Sentinel-2, Landsat and PlanetScope. At least 10-200 different satellites. We use in-house algorithms to standardise the pixel size at 10m consistently across the full time and space.

We do not use commercial imagery as our core dataset at scale, because it is not necessary to use even higher resolution or revisit time. We also prefer to keep the system affordable for our many clients covering millions of hectares.

Yes. We’ve built advanced algorithms that detect deforestation across the globe based on satellite images. We use Machine Learning techniques like Random Forest and Gradient Boosting to examine information about different land cover types, like forests and agriculture, and make predictions. Deep Learning models help us remove unwanted noise and distortions from radar images, making them more precise and accurate. We also utilize additional statistical methods like Bayesian Iterative Updating to manage and keep track of land use change over time.

We use computationally intensive AI only where and when it makes sense. Such as for reaching higher accuracy of counting palms, or advanced noise filtering. In many cases, however, pragmatic machine learning solutions perform even better and are much less computationally intensive. I.e. lower cost, lower power use and therefore lower carbon intensity of our services.

Satellite data is NOT a substitute for ground data and knowledge. We use a global ground survey database with almost 200 million ground survey locations to train our algorithms and validate the results.

Given different ecoregions, one automated approach for all forest types around the world is a recipe for failure. Our system is locally calibrated and validated for specific ecoregions, whether lowland rainforest, montane forest, chaco, cerrado woodlands or other.

More details here: [https://satelligence.com/news/2020/3/24/why-you-dont-need-very-high-resolution-data-to-detect-deforestation]

We assess accuracy for each region based on:
A statistically valid sampling procedure using verification with very high resolution reference satellite imagery. Our method is in compliance with the best practice guidance of GOFC GOLD (2016). [http://www.gofcgold.wur.nl/redd/sourcebook/GOFC-GOLD_Sourcebook.pdf]
Feedback loop with clients (growers, traders) doing checks and reporting back;
Partners with boots on the ground, NGOs and sector organizations doing field verification checks as part of their daily activities.

Satelligence initiated and supports the RADD initiative to get alignment of leading companies behind a consistent methodological basis. However, the public WRI version is meant to be a global awareness resource and NOT specific to company compliance use.

Public RADD monitors ‘disturbance’, not deforestation. It exaggerates degradation (e.g. any 5x5m crown removed becomes 40x40km pixels) which is nice for rough logging alerting but not for consistent area change statistics on commodities.

Most importantly: deforested pixels on their own are not useful: they always need to be put in the context of baseline layers, supply chain data, and drivers of deforestation need to be assessed. That’s where we come in.

First of all we spent almost a decade to perfect large scale preprocessing of input imagery. This is critical to avoiding inconsistencies and errors in the final classified products, in particular from widespread artefacts of persistent clouds, haze, hilly terrain etc.

Secondly, we optimize forest and other ecosystem baseline generation per biome. Our proprietary commodity layers are crucial: open data always confuses perennial crops with forest, leading to an overload of false alerts.

An unfair advantage we have is access to a mind blowing number of perennial crop plantation boundaries to train our classification algorithms. No one else in the world has such a complete overview of actual distribution of tropical farms.

Thirdly, we integrate data from 10 sensors with 1 single approach, unlike GFW which mixes together multiple data sources generated with wildly different methods. Open data both over and underestimate deforestation. Contact us for more information.

No! Do not buy satellite data products and fancy dashboards sold as automated push button miracle solutions.

We believe that thorough local expertise and field data should be at the basis of any map production. Also, it is best practice to do occasional ground checks to verify and report accuracy. The benefit of satellite data is that such expensive and time-consuming field visits are needed way less. Done properly, decision-ready satellite data brings massive efficiency gains.

Yes, if you invest in seasoned staff with decades of experience and advanced technology you could, perhaps, but making risk-free monitoring look easy is very hard. Secondly, self-reporting is a thing of the past.

Why would consumers, buyers and investors trust companies grading their own homework?

The input data (airborne and spaceborne GEDI LiDAR, Sentinel-1 and 2 satellite imagery), methods and algorithms used by all leading providers are more or less similar. Ignore words like ‘cutting-edge’, ’groundbreaking’, ‘AI/ML’, ‘unique’, ‘innovative’ or ‘state-of-the-art’ while researching providers.

Please consider this instead:

1. Most competitive pricing for the high quality results. We spent years optimizing our processing efficiency to enable affordable scaling to global supply chain coverage. Thanks to scaling years ahead of competitors and learnings from proven applications with a large number of clients, we are able to offer a lower price point.

2. Lowest carbon footprint. Spending years on improving cloud computing efficiency at scale also means we require less energy to deliver results covering entire global supply chains. Thanks to Google we are able to quantify and report the limited emissions associated with our service.

3. Best input data. Having worked in cloudy and hazy tropical areas for years, we were forced to build the best image preprocessing engine in the world. Creating the highest quality cloud and haze free input pixels. We don’t process noise to create noise. Example: 3m carbon data may look fantastic, but most of what you see is noise. There is a reason the best available data is available at 10-30m resolution. As with digital camera’s: don’t fall for the Megapixel myth: more Megapixels DOES NOT mean better quality photos.

4. Contextual intelligence. You benefit from 25 years of struggling in remote areas. Measuring trees. Surveying vegetation. We know what we are talking about. We know a forest in Indonesia is rather different from a forest in Ghana or a forest in Brazil. We know what a palm oil plantation is. We know how diverse cocoa management systems can be. We know how coffee is produced. Because we have been there. We saw it with our own eyes.

5. Carbon data is not what matters. What matters is reporting your Emission Factors. So you need integration of supply chain data, commodity layers and historical deforestation data. We offer a best in class solution integrating all these components combined. Globally. Temporally consistent. Coherent.

6. We are nice people

Other providers are limited to just carbon. Providing data, not insights. Satelligence offers crucial contextual intelligence to enable specific and relevant assessments:

Just knowing a specific patch of land stores 50 tonnes of carbon is insufficient: if it is not known whether it is associated to your supply chain, or associated with the commodity in focus, such information is worthless.

A cocoa plantation, young palm plantation, shrubland, and young regrowth forest can all have the same carbon stock. Satelligence’s exceptional traceability database and global commodity layers make carbon assessments superbly relevant.

Don’t risk overreporting of your emission reductions and removals using non-specific carbon data. And losing market share.

You may select emission factors from the literature. However, using such default values you will tremendously overestimate your climate impact! Commodity-driven deforestation can be uncovered and quantified precisely with explicit spatial data. Not with exaggerated default values from literature.

For example, if you produce or source cocoa from Ghana, why should your carbon footprint be inflated by massive emissions from nearby gold mining and logging? Using our distinctive cocoa maps we are able to limit quantification to cocoa.

Using more specific and more granular data gives precise estimates. Leading to better market access. Buyers increasingly choose low carbon density suppliers. Be one of them.

Satelligence helps you monitor and deliver. Keeping up with process innovations, new technologies, and new alliances is very demanding on your staff. Tap us to help enable skilled and knowledgeable procurement teams.

We guide you through best practice using monitoring for the implementation of sustainable business models. In order to support procurement decisions our system screens all suppliers in the industry: who are the top performers? Who need support to raise the bar?

We unburden you in engagement with suppliers, buyers and other stakeholders, saving time by automating tracking down grievances, reporting of actions taken and progress.

Our highly responsive development team is flexible adapting our service to your workflows. Our customer success team is standby providing a Helpdesk for tailor-made support.

Get strategic advice on how to best integrate monitoring in your work processes. We develop state of the art approaches with leading stakeholder initiatives. Satelligence works together with a large network of local and international partners. Providing strategic advice on forest and supply chain monitoring to CGF, POCG, WCF, RTRS, IDH, GPI, Norwegian Government, Netherlands Government etc.

Satelligence is uniquely positioned to ensure that a wide network of suppliers is united, as we are involved in setting the standard, we foster cooperation and efficiency, and our client-base has the critical mass to push our work as industry-consensus.