I’ve long asked how “architectural conservation” and protection of the built environment can meet the latest technologies. How can a tradition‑rooted, engineering‑oriented field—one that carries historical responsibility—engage today’s tools and today’s questions? During my internship at the Second Surveying and Mapping Institute, I began to find practical, working answers that connect disciplinary heritage with contemporary methods.
Much of Hunan Province is hills, lake districts, and plains. Agriculture continually reshapes surface vegetation, settlement patterns, and the landscapes we see. Cultivation not only gives form to “landscape,” it also binds local livelihoods. Effective protection must therefore be place‑based and sustain a coupled system of ecological change from farming, visible land cover, and human activity. My internship became an exploration of agricultural landscapes and a study of crop distribution on the ground using GIS and computational techniques drawn from remote sensing and computer vision.
Knowing where crops are planted matters: it lets us track year‑to‑year land‑use change and shifts in landscape condition. Crop patterns also reflect farmers’ income prospects and the sustainability of rural economies. The difficult part is measuring consistently and at scale. GIS links data to geography—but where do we obtain current, synoptic vegetation information? From satellites. Time‑series imagery provides spectral bands and indices (e.g., NDVI, NDWI) that are strong indicators of land cover and crop types. I learned to access open datasets through platforms such as the Copernicus Open Access Hub, USGS/NASA portals, and Google Earth Engine’s public catalog.
In practice, I visited farming areas in the hills and around the plains‑and‑lakes and—together with researchers—mapped rice phenology (early, mid, and late seasons; early rice mainly March–May). Using Google Earth Engine and ArcGIS Pro with ESA Sentinel‑2 (10 m) data, we extracted NDVI/NDWI, applied cloud–shadow masking, and built area‑wide models to identify likely rice pixels. We validated outputs against field observations and very‑high‑resolution imagery and iterated thresholds, so the temporal signatures matched known planting calendars.
For village‑scale studies, I trained convolutional neural‑network (CNN) models on Google Colab. I also designed a lightweight labeling workflow to sharpen boundaries around paddy fields and reduce confusion with bare soil or shadows.
This internship deepened what I learned in the Digital Media course: programming with GIS and analyzing geospatial data pipelines end‑to‑end. It also sharpened my understanding of “landscape” and “environment”: regional research depends heavily on aerial and satellite imagery. “Survey and documentation,” a constant theme in Documentation class, which could extend to satellite data, computer vision–based object detection, and point‑cloud computation.
Equally important, I met scholars working across environmental and conservation fields and gained a clearer view of my own direction: to study how seasonal cultivation interacts with settlement fabric and heritage landscapes in hilly southern China, using reproducible, open remote‑sensing workflows. Under colleagues’ guidance, I practiced drafting research proposals and structuring a study—from framing questions and methods to dataset selection, timelines, risk management, and evaluation.
Agricultural landscapes are often left out of history because they lack buildings. Yet farming systems are essential to urban planning and landscape conservation. In China, rural landscapes are not background—they help anchor community memory and identity.
View Slideshow
