A sixth-century groundwater manual tested against India's well network: plant indicators and depth claims in Varāhamihira's Bṛhatsaṃhitā

Aneesh Karavadi 

The Dakārgala chapter of Varāhamihira's Bṛhatsaṃhitā (sixth century CE) is one of the oldest detailed manuals for finding groundwater. It names trees, often together with anthills, as signs of water and gives the depth at which water will be found. It is often cited as early hydrological knowledge, but its claims have not been tested systematically. We coded the chapter's plant indicators and 48 stated depths and compared them with India's groundwater monitoring network: 11,530 dug wells from a quality-controlled release of Central Ground Water Board data, aggregated to 2,385 grid cells of 0.25°. We combined these with GBIF plant-occurrence records and WorldClim climate data, following an analysis plan written before the plant and climate data were obtained. Four results stand out. (1) The text's lists of trees indicating near water and far water separate in the predicted direction: near-water species make up a larger share of plant records where the water table is shallow, after controlling for climate (difference in mean coefficient −0.40, permutation p = 0.020, spatial block bootstrap 95% CI −0.61 to −0.11). The effect kept its sign in all six analysis variants but was significant in only three, and it is not unusual compared with randomly chosen pairs of species lists in most variants. (2) The broader set of trees used as signs is no more associated with shallow water than common Indian plants (p = 0.43). (3) The text's rule that the same signs mean water at 3, 5 and 7 man-lengths in wet, dry and desert country is right in direction. The modern desert-to-wet ratio of 1.99 (95% CI 1.68–2.41) brackets the text's 2.33, but the dry-to-wet ratio of 1.12 is far below the text's 1.67. (4) Depths stated for wet and dry country (about 3–17 m) match modern dug-well depths. By contrast, a third of the depths in the desert sequence exceed the deepest plant root ever recorded, and the sequence is ordered almost perfectly by verse (ρ = 0.996), more like a constructed catalogue than field observations. The chapter reads as a practical, partly empirical tradition: plausible at the scale of ordinary wells, schematic at its extremes, and untested at the micro-site scale on which most of its claims are made. We describe a field design that could test those claims.