Survey & Beyond: The Data Collection Podcast

The episode contains genuinely useful technical workflow detail about offline geospatial data collection (MBTiles, polygon capture, Python/QGIS integration) but is pitched at a beginner audience and spends significant time on basic definitions of EUDR and GPS concepts. There is little that a practitioner in geospatial or agricultural data would not already know.

“First, what I actually do is process the data using Python and goodies, which is extremely useful because it allows you to organize and visualize all the farm layers in a very practical way”

“we found a building function inside services deal that allow you to calculate the area of the polygon that you just draw”

The episode is almost entirely descriptive and explanatory — walking through established tools, standard compliance workflows, and widely known regulatory requirements. There are no contrarian claims, no first-principles reframings, and the one forward-looking observation about ML boundary detection is vague and unattributed.

“Also something that has caught my attention lately is that there are some advances in satellite mapping. Using machine learning models developed by Google to analyze large volumes of earth observation data”

“Sustainability standards and traceability requirements continue to grow. Because they will definitely continue”

Nicole Linares is a genuine field practitioner who has executed real geospatial data collection projects in rural Peru for Rainforest Alliance — not a thought-leader or career conference speaker. However, she is a mid-level research analyst, not a senior operator who has scaled these systems at enterprise level, which limits the ceiling of depth she can speak to.

“I currently work at letter 8 with this research data firm that supports organizations working in areas like agriculture, supply chain and environmental sustainability”

“Many of the projects we work on involve collecting both survey data and geospatial data. For example, this case we have with Rainforest alliance, it was all about mapping farm boundaries and understanding land use patterns”

The episode names real organizations (Rainforest Alliance), a specific regulation (EUDR), geographic locations (Peru, San Martin), and a concrete toolchain (MBTiles, QGIS, Survey CTO, Google Earth Engine, Sentinel), which is a reasonable level of specificity. However, there are virtually no quantitative data points — no farm counts, no compliance rates, no timeline details, no cost or scale figures — leaving the claims largely unverifiable.

“In places like San Martin, where farms are often surrounded by forests and access can be challenging”

“it usually is collected annually, and that allows field teams to be prepared for this data collection”

The host structures the conversation clearly and occasionally adds useful bridging context for listeners (e.g., the Google Maps analogy for offline maps), but questions are almost exclusively descriptive prompts ('could you explain…', 'what do you think…') with no meaningful pushback, no probing of failure cases, and no challenge to any claim Nicole makes.

“And out of curiosity, so that you know that no deforestation has happened, like over time, is this something that you need to collect longitudinally?”

“So like a single GPS point is not good enough now. Right. You really need to map the whole land of a farmer”