Work

sakshimaske303-commits/THE-COTTON-LINE: THE COTTON LINE : Testing the Persistence of an 1853 Colonial Land-Tenure Boundary in Present-Day Vidarbha

Year: 2026
Type: software
Abstract: This project checks if an old 1853 British border still leaves a physical mark on Vidarbha's soil today. To the west sits Berar (which ran on the private-plot Ryotwari cotton setup), while the Central Provinces sat to the east using the landlord-based Malguzari system. Even though the government officially threw out these laws back in 1956, I wanted to see if modern satellite greenness patterns, farm labor numbers, and land heat still change right at that old boundary line. I built the modern border layer by dissolving 120 taluka shapefiles across 11 districts, which gave me an exact 384.7 km shared boundary edge. Next, I used a separate archival check to verify that this modern line runs where the old historical division line used to sit, tracking it as far as documented district paperwork from 1903 to 1956 can show. I then ran an OLS spatial-discontinuity model to test my four main dataset outcomes across two different geographic scales: a tight 32-taluka near-boundary corridor buffer (±25 km) and a much bigger 66-to-68-taluka full satellite footprint. I found absolutely no statistically significant level jump right at the boundary line for any metric once I blocked out natural elevation and rainfall trends. Instead, a real regional gradient shows up when you look at the drop in greenness from the kharif monsoon peak to the winter rabi harvest. My full-footprint model showed a 0.070-point cotton-side excess in the first year, and my data independently caught a 0.074-point gap when I re-ran it on the next year's crop cycle. This same side × distance interaction model turned out significant for ground heat metrics (p=0.0105) alongside my NDVI vegetation checks (p=0.0348). Agrarian labor structures gave me a very weird, scale-sensitive result: it was significant at the narrow corridor scale (p=0.0361), and while the coefficient direction held up across a 32-round leave-one-out jackknife test, the actual statistical significance dropped out in some runs with the p-value going as high as 0.138. The model then flipped completely into a level jump when I used the big full-footprint scene (p=0.014), which I traced directly to how the satellite tracks cut off certain districts unevenly on the map. To cross-validate my satellite NDVI signal against real crop conditions, I checked seven distinct field boundaries on the Central Provinces side within Karanja Ghadge and Sawardoh villages using Sentinel-2 layers and dated ground photos. If you want to check my complete methodology layout, raw data providers, or coding history, you can find them right on the dashboard pages and inside my raw CL_Development_Log.md, CL_Research_Paper.md, and CL_Executive_Summary.md files.
Author: Sakshi Maske
Language: English
Subfield: Soil Science
Open Access status: green
OpenAlex ID: W7204257875
Cites:
Cited by:
Related to: 10