Soil Geochemistry and Pre-Spice Mass Formation: A Comparative Survey and Trace-Element Enrichment Score from Thirty-Seven Open-Erg Sites Cored in 10238–10244 AG

1. Introduction

The planetological tradition of Arrakis gives a short causal account of the pre-spice mass. Sandtrout, the little makers that form the juvenile stage of Shai-Hulud, encapsulate free water and so hold the planet's surface dry. When water reaches their excretions deep under the sand, the resulting fungoid growth, the pre-spice mass, ferments until it erupts as a spice blow, which carries deep material to the surface. In this account water is the only condition named. Our companion time-to-event study of archived aggregation sequences found that the interval from aggregation to blow is gated by colony mass, with light aggregates slow to reach the threshold and several never seen to blow. It treated site conditions as unmeasured confounders.

Field records complicate the simple account in the same direction. The Fremen Planetological Survey and the older sietch recitations describe places where little makers were seen to gather and where water was known to have been lost into the sand, yet no blow followed. Such failures could reflect quantity, timing or colony size, or some property of the surrounding sand that gates the process in addition to water. The sandtrout membrane takes up water by an active pump working against the osmotic pressure of the pore water, and the ionic composition of that pore water gives a plausible point at which mineral chemistry could matter.

This paper asks whether the trace-element composition and moisture-retention behaviour of the sand differ between sites that produced a blow and sites that did not, and whether any difference survives the objection that a blow itself reworks the soil it leaves behind. Our fieldwork falls in 10238–10244 AG. The planetological programme of that period has altered moisture regimes in the planted basins, so we restricted sampling to open erg well outside them.

2. Methods

Sites were drawn from three sources. Blow sites (n = 18) were located from the harvest location registers of the CHOAM Directorate for 10180–10240 AG, which record the position and approximate date of each blow worked by a factory crawler. We accepted only blows placed to within about half a kilometre and confirmed by an entry in the Survey station logbooks, which run continuously from 10110 to 10240 AG. Failed sites (n = 7) came from the logbooks and baseline transects of 10110–10191 AG and from the Sietch Tabr recitation series. To qualify, a failed site needed two independent attestations of little-maker activity at the place, at least one describing water lost into the sand there, and no blow recorded in either the logbooks or the harvest registers for the following three decades. Control sites (n = 12) were chosen at random from the same erg sectors, at least 5 km from any recorded blow or little-maker attestation.

A blow excavates and redeposits the sand above the mass. Survey excavation profiles of recent blow craters place the reworked layer at roughly 2–4 m. We therefore cored every site to 8 m and analysed two intervals: a surface interval (0–3 m), which at blow sites lies largely within the reworked layer, and a deep interval (5–8 m), which should lie below it. Sixteen sites received one core and 21 received two, giving 58 cores (median two per site). Where a site had two cores, interval values were averaged so that the site is the unit of analysis throughout.

Fourteen major and trace elements were measured by X-ray fluorescence, calibrated against sand-matrix standards. Moisture retention was measured on the surface interval only, as gravimetric water content (percent by mass) held after saturation and drainage under a standard suction.

Element concentrations were log-transformed. Each element was compared across the three groups with a linear model, and the blow-versus-failed contrast was tested with a Holm correction across the 14 elements. For classification we defined an enrichment score as the mean, over strontium, zirconium and lanthanum, of the base-2 logarithm of each concentration divided by the control-group geometric mean for the same interval. Surface scores are thus referenced to surface controls and deep scores to deep controls, so that controls average zero at each depth and any natural vertical gradient in background chemistry is removed. One unit corresponds to a doubling above background. Group differences in score were tested with Welch's t-test, and a linear model added surface moisture retention as a covariate. A single threshold on the score separated blow from failed sites. Since controls are never classified, the threshold is the only fitted quantity. We re-chose it within each training set under leave-one-out cross-validation, and Clopper–Pearson intervals were used for proportions.

Four blow sites had been cored during routine Survey baseline transects before their blows occurred. Those archived cores, stored at Sietch Tabr, reached only 3 m, so they supply a surface interval matching ours but no deep interval. They were reanalysed with the same protocol and scored against the surface control means.

3. Results

Of the 14 elements, only strontium, zirconium and lanthanum differed between blow and failed sites after Holm correction (Table 1). Cerium showed an unadjusted difference that did not survive correction. Failed sites were modestly above controls for each, and blow sites well above both.

Moisture retention differed across the groups (F(2, 34) = 20.3, p < .001). Blow sites held 3.4% (SD 0.9) and failed sites 3.1% (SD 0.8), against 1.6% (SD 0.5) at controls. The blow-versus-failed difference was 0.3 percentage points (95% CI −0.4 to 1.0, p = .39). Both active groups exceeded controls, by 1.8 points (1.2 to 2.4) and 1.5 points (0.8 to 2.3) respectively.

The mean enrichment score was 1.61 (SD 0.41, range 0.71–2.38) at blow sites and 0.28 (SD 0.38, range −0.22 to 0.97) at failed sites. The difference of 1.33 score units (95% CI 0.95–1.71; Welch t = 7.69, df ≈ 11.8, p < .001) became 1.29 units (95% CI 0.90–1.68) with moisture retention as a covariate. The best in-sample threshold, 0.90, classified 23 of 25 active sites correctly (92.0%, 95% CI 74.0–99.0%). One blow site fell below it and one failed site above it, giving a sensitivity of 17 of 18 and a specificity of 6 of 7 (Youden index 0.80). Under leave-one-out cross-validation, 22 of 25 sites were classified correctly (88.0%, 95% CI 68.8–97.5%), with a sensitivity of 16 of 18 (88.9%, 65.3–98.6%) and a specificity of 6 of 7 (85.7%, 42.1–99.6%).

Control geometric means in the deep interval were close to their surface values: strontium 305 ppm, zirconium 121 ppm and lanthanum 69 ppm, against 310, 118 and 71 ppm at the surface. Background chemistry thus showed no appreciable vertical gradient. Deep scores averaged 0.98 (SD 0.44) at blow sites and 0.23 (SD 0.40) at failed sites. The blow-versus-failed contrast of 0.75 units (95% CI 0.35–1.15, p = .002) was about 56% of its surface size. Failed sites changed little between intervals (0.28 at the surface, 0.23 at depth), whereas blow sites fell from 1.61 to 0.98.

The four archived pre-blow cores had surface scores of 0.81, 1.04, 1.35 and 1.52 (mean 1.18). Three lay above the 0.90 threshold, and all four were above the failed-site mean. The post-blow surface scores at the same four sites were 1.22, 1.58, 1.71 and 1.89 (mean 1.60). Every site rose after its blow, by 0.36 to 0.54 units (mean 0.42). With four pairs we report these differences descriptively.

4. Discussion

Two findings define the result. First, moisture retention separated active ground from background but not blows from failures, so the failures in the record are hard to attribute to sand that simply could not hold water. Second, a small group of trace elements was markedly more concentrated at sites that went on to blow, and a one-parameter score built from them recovered most of the outcome under cross-validation. The data do not show that it is necessary, because one blow site scored below the threshold, and they cannot show that it acts independently of colony size or organism condition, because neither was measured.

The principal threat to interpretation is reverse causation. A blow throws deep material onto the surface, so enrichment in the surface interval at blow sites could be the blow's deposit and not a condition that preceded it. Yet the contrast persisted below the reworked layer, where the blow adds nothing, though at a reduced size. Because control chemistry barely changes with depth and failed sites held their scores across intervals, the attenuation at blow sites is compatible with part of the surface signal being blow-derived. The paired pre-blow cores point the same way: each site was already enriched above the failed-site mean before its blow and rose by roughly four-tenths of a unit afterwards, an increment of the same order as the surface-to-deep gap. Four sites cannot establish the direction of the association, however, and the deep interval at blow sites lies next to where the mass itself grew, so the mass may have altered its surroundings chemically.

Aggregate mass is a competing explanation that we could not measure. The companion time-to-event analysis found that light aggregates approach the blow slowly and may never reach it within any practical observation window. Some of our failed sites may therefore be places where the colony was simply too small. If colony size and trace-element enrichment covary, for example because richer ground supports larger aggregations, the score could partly be a proxy for mass.

Any mechanism is speculative. The companion study examined the heat that exothermic water sequestration could release and found it far too small to account for the thermal anomalies at mass sites, so we do not build on that step. A more promising point of contact is the membrane pump. Sandtrout take up water actively against the osmotic pressure of the pore water, and divalent and trivalent cations in that water could plausibly affect the pump or the membrane surface. We have no measurement linking strontium, zirconium or lanthanum to little-maker physiology, and the hypothesis needs laboratory uptake trials in sand of controlled composition.

For survey practice, the score is best regarded as a candidate screen. Its specificity interval reaches down to about 42%. A prospective test, in which the deep interval is cored before any blow, aggregate extent is recorded, and outcomes are then followed, would settle its predictive value, the direction of the association and its independence from colony mass.

5. Limitations

The failed-site group is small (n = 7) and was defined from records that are uneven in quality. Recitations favour memorable events, so the verifiable failed sites may be unrepresentative, perhaps skewed toward sietch routes. Their attestations date from 10110–10191 AG, between about five decades and more than a century before coring, and the location of some is uncertain to within a few hundred metres.

Time is a second confound. Blow sites were cored between a few years and several decades after their blows, and failed sites long after the events that define them. Aeolian reworking and the moisture changes of the greening era may have shifted surface chemistry differently across groups.

Moisture retention was measured only on the surface interval, which at blow sites is largely reworked blow material. The same reverse-causation concern that applies to the elements therefore applies here, and the null blow-versus-failed contrast in moisture, as well as the covariate adjustment of the score, rests on a measurement the blow may itself have altered.

Finally, the threshold and the choice of three elements were both derived from the same 25 sites. Cross-validation guards against overfitting the threshold, but the three elements were selected by the full-sample analysis, so the cross-validated accuracy is itself likely to be somewhat optimistic.