It seems that both "Acre" and "CropCultLand" refer to the same concept. The primary difference lies in the units of measurement. "Acre" is clearly measured in acres, a specific unit of land area. In contrast, "CropCultLand" does not specify any particular unit in its description, suggesting it might represent the land area in a different unit of measurement.

While it's tempting to assume they convey the same information, the true relationship can only be determined through data analysis. The histograms for both variables exhibit similar shapes, leading one to hypothesize that "Acre" might simply be "CropCultLand" scaled by a factor.

If this hypothesis is correct, the two should exhibit a strong linear relationship. However, this isn't the case. It's possible that "Acre" is a non-linear transformation of "CropCultLand" rather than a scaled version. Another possibility is that they originate from different data sources. Even if they're intended to represent the same information, discrepancies can occur due to the varied origins.

The Spearman correlation score provides insight into non-linear relationships, and in this case, it's relatively high at 0.636.

Ultimately, this is a machine learning exercise, and experimentation is essential. It's up to you to determine if both features are relevant for your model.

Fantastic point on the non-linearity @yanteixeira

When it comes to the conversion, I think the main issue is the non-standardization of land measuring units across India, like Dhurki, Katha & Bigha, where each has a different number of acres per unit value. I suspect "CropCultLand" uses any of these metrics with no real standardization, so there isn't a single conversion factor that can be applied universally to convert them to acres. Instead, I'd imagine a more cumbersome approach of converting for each District (or even Block) individually to demonstrate equivalency with corresponding Acre values. The good thing though is that Acre seems to be a standardized metric!