Sandworm Territorial Spacing and Coriolis Storm Disruption in Eight Deep-Desert Basins of Arrakis, 10228–10239 AG

1. Introduction

That Shai-Hulud, the great sandworm of Arrakis, holds territory is among the oldest observations in Fremen desert lore. Straka (10230 AG) assembled the Fremen and Imperial field observations and concluded that mature makers keep a broadly regular spacing across open sand. She attributed that regularity to mutual avoidance between neighbours, so that spacing is set by how many animals a basin holds. Regional estimates of thumper response rest on the same reasoning.

Sietch observers have nevertheless recorded, season after season, that worms seem to range more widely and stand farther apart when the great storms are running. A pure competition account cannot easily explain this. Worm numbers in a basin change slowly, through the death of adults and the growth of young worms from the sandtrout stage, a maturation that the planetological account measures in years (Office of the Imperial Planetologist, Arrakis, 10152–10190 AG). Storm conditions change within weeks. Coriolis storms, with winds strong enough to scour exposed flesh, rework the upper sand of whole basins, moving dune crests and burying or exposing surfaces over hundreds of square kilometres.

Testing the pattern depends on a storm record kept on the ground. The Spacing Guild keeps no weather satellites over Arrakis, an arrangement the Fremen secured by paying it in spice, so weather knowledge of the deep desert has rested on sietch observers. After the death of Liet-Kynes in 10191 AG the Fremen Planetological Survey continued his registers. From 10228 AG it standardised the storm and sighting logs of its deep-desert observation posts, and these logs are the basis of the present study.

We ask whether the extent of storm reworking in a basin-season is associated with worm spacing once density is held constant, and whether it accounts for the storm-season widening. Writing from Sietch Tabr in 10241 AG, we treat the competition-only model as the baseline and test whether a disruption term adds explanatory value. The mechanism advanced in the Discussion is our own proposal and is not attested in Fremen or Imperial observation.

2. Methods

Study basins. The study covers eight deep-desert basins south of the Shield Wall, each bounded by rock outcrops or by the sand margins used in the Survey's own cartography. Basin areas ranged from about 41,000 to 118,000 km², and all eight lie well outside the sietch-country planting domain of the greening programme and its collector-altered surface moisture. Each basin was watched by two to five sietch observation posts.

Sighting records. The Survey's worm logs (Fremen Planetological Survey, 10228–10239 AG, Series WS) contain 6,318 georeferenced sightings of mature makers across the eight basins, between 480 and 1,150 per basin over the twelve years. Each fixes a surface wave or emergence by bearing and range from a surveyed post. Because individual animals cannot be identified, spacing was estimated from concurrent sightings, defined as sightings in the same basin logged by different posts within 36 hours. For each such sighting we measured the distance to its nearest concurrent neighbour and took the basin-season mean as the spacing estimate (Threll, 10236 AG). A basin-season entered the analysis only if it yielded at least 12 concurrent pairs.

Worm density, expressed as mature worms per 1,000 km², was derived from sighting rates corrected by a distance-based detection function. That function was calibrated on 1,214 thumper-response trials in which both the planted thumper and the worm's first surface wave were fixed from surveyed posts, so the response distance was known (Ferrach, 10231 AG). Density was estimated separately for each basin-season.

Storm records. The storm registers (Fremen Planetological Survey, 10228–10239 AG, Series CS) record each storm passage by date and track. After each season, survey parties resurveyed fixed dune transects. From these we took disruption as the percentage of basin surface showing more than 0.5 m of wind-driven vertical dune change, the threshold Kesteven (10233 AG) adopted as the point at which reworked sand ceases to match earlier terrain survey. Intensity grading changed between posts and years, so intensity was not modelled. Each year was divided into a storm season and a calm season, the two halves of the local year with the most and fewest recorded passages.

Statistical analysis. Twelve years and two seasons in eight basins give 192 possible basin-seasons. Eleven failed the concurrent-pair threshold, which left 181 (87 in the storm season and 94 in the calm season). Spacing is a continuous measurement, so we used linear regression with basin fixed effects, making each coefficient a within-basin association. Model 1, the competition-only baseline, regressed spacing on density and a storm-season indicator. Model 2 added disruption, and the two were compared by likelihood-ratio test and Akaike's information criterion (AIC). Model 3 replaced disruption with the number of storm passages in the season. We also refitted Model 2 within each season alone, so that disruption contrasts heavy and light storm years in the same season.

Two checks addressed change over time, since these were years of ecological transformation and of a maturing observer network. Each model, including the season-specific fits, was refitted first with year fixed effects and then with a linear year term. To allow for dependence among seasons within a basin, we recomputed the main disruption interval by a wild-cluster bootstrap over basins (9,999 replicates).

3. Results

Mean nearest-neighbour spacing across the 181 basin-seasons was 58.1 km (SD 7.3; range 39.5–79.8): 55.1 km (SD 6.2) in the calm season and 61.4 km (SD 7.0) in the storm season. Storm-season spacing exceeded calm-season spacing in all eight basins, by a mean of 6.3 km (95% CI 4.0–8.6; paired t(7) = 6.47, p < .001). Estimated density ranged from 0.15 to 0.61 worms per 1,000 km² (SD 0.095) and scarcely differed between seasons (calm 0.38, SD 0.09; storm 0.36, SD 0.10). Disruption averaged 11.2% of basin surface (SD 4.0) in the calm season and 35.5% (SD 8.6; range 15.4–63.1) in the storm season, with an overall SD of 13.8 points. The registers recorded a mean of 1.2 passages per calm season and 6.2 per storm season.

In Model 1, spacing narrowed as density rose: each additional worm per 1,000 km² was associated with spacing 39.3 km closer (95% CI −54.6 to −23.9; p < .001), and the storm season with spacing 5.7 km wider (95% CI 4.1–7.4; p < .001). With density nearly constant across seasons, only the season term could carry the seasonal widening.

Adding disruption in Model 2 improved fit (likelihood-ratio χ²(1) = 12.95, p < .001; ΔAIC = 11.0). Each percentage point of reworked surface was associated with 0.22 km wider spacing (95% CI 0.10–0.34; t(170) = 3.55, p < .001), while density was essentially unchanged (Table 1). The season term fell to 0.4 km (95% CI −3.0 to 3.7; p = .83). The 24.3-point seasonal difference in mean disruption corresponds to about 5.3 km of spacing, close to Model 1's season term. Scaled by one standard deviation of each predictor, density corresponds to about 4.0 km of spacing and disruption to about 3.0 km, associations of similar magnitude.

Storm frequency was a weaker measure than surface reworking. In Model 3 each passage was associated with 0.49 km wider spacing (95% CI 0.07–0.90; p = .02), but the fit was poorer than that of Model 2 (AIC 7.4 units higher). Within the storm season alone, disruption was associated with 0.28 km wider spacing per percentage point (95% CI 0.14–0.41; p < .001), so heavy storm years showed wider spacing than light ones in the same season. Within the calm season the estimate was 0.10 km (95% CI −0.18 to 0.38; p = .47). With disruption confined to a narrow range in that season, the interval cannot separate an absent association from an unresolved one.

Adjustment for time left these findings intact. With year fixed effects the disruption coefficient was 0.20 km (95% CI 0.07–0.33; p = .003), the likelihood-ratio test for disruption gave χ²(1) = 10.6 (p = .001), and the year effects themselves did not improve AIC. The wild-cluster bootstrap widened the interval to 0.05–0.36 (p = .02). A linear year term (0.18 km per year; 95% CI −0.09 to 0.45; p = .19) left disruption at 0.21 km (95% CI 0.09–0.33). In the storm-season fit with year effects, disruption remained at 0.26 km (95% CI 0.11–0.41; p = .001); in the calm season it was 0.09 km (95% CI −0.21 to 0.39; p = .56). The storm-passage coefficient, by contrast, weakened to 0.44 km (95% CI −0.01 to 0.89; p = .06).

4. Discussion

Density remains a strong predictor of how far apart worms stand, and nothing here displaces the competition account of Straka (10230 AG). Storm reworking of the surface nonetheless adds explanatory value that density cannot supply, with an association of comparable size per standard deviation. It also absorbs most of the apparent seasonal effect, leaving the season term near zero. The within-season contrast between heavy and light storm years holds the calendar constant and so narrows the room for other seasonal explanations. That the association survives year effects argues against a shared drift in climate or observer practice as its source.

Our proposed mechanism is a hypothesis. The field record establishes that worms are territorial but is silent on how a worm recognises that ground is already held. We suggest that an established animal leaves a persistent signature in the upper sand, perhaps a trace of spice-bearing excretion along its habitual paths, or a pattern of compaction that carries its movement vibrations in a characteristic way. Heavy reworking would bury or scatter it, and neighbours would keep a wider margin until it is re-established. The hypothesis predicts that spacing should narrow again over the weeks after a storm season, faster in basins with light reworking, and that post-storm surveys should find worms re-crossing boundaries they avoided before the season. Both can be tested from the existing logs.

For practical desert work, the results matter because estimates of thumper-response time, and of where a harvester is likely to meet a worm, are usually built from density alone. If spacing widens after heavy storms, the chance that a thumper falls within an animal's hearing should fall correspondingly in the weeks after a severe season. These conclusions are confined to the deep desert. Whether the same relation holds near the planting domain, where the transformation is altering surface conditions, is a question for further survey.

5. Limitations

Detection is uneven: posts sit near sietches, and basins with more posts contribute more concurrent pairs. The detection function corrects for distance from a post but not for ground that no post overlooks. Storms also reduce visibility, and if they suppress sightings unevenly the concurrent-pair method may overstate spacing where disruption is high. The positive disruption coefficient within the storm season, when visibility is poor in all years, argues against this as the whole explanation but cannot exclude it.

Successive seasons in one basin share observers, posts and animals. The bootstrap over basins addresses this dependence, but with only eight basins even clustered inference rests on few independent landscapes, and its interval should be read as approximate. Year effects remove drift common to all basins, not trends confined to particular basins. Density is itself an estimate, and error in it will tend to bias its coefficient toward zero. Disruption and season remain strongly correlated, so the season-specific fits are the more direct evidence. All associations are observational, and the territorial-signature mechanism has not been observed.