Philip Hampsheir

April 2026

Abstract

Large-scale de-desertification projects — including China’s afforestation programmes on the Loess Plateau and in Inner Mongolia — operate on a foundational assumption: that vegetation plus water will eventually restore regional rainfall dynamics. This paper asks whether that assumption is structurally complete. The bioprecipitation cycle, through which fungal and bacterial ice-nucleating proteins seed cloud formation, is an established driver of regional precipitation that is absent from degraded desert soils. The question that has not been formally posed is whether reforestation without deliberate restoration of this microbial infrastructure can reliably trigger the self-sustaining rainfall feedback loops that such projects depend upon — and whether the absence of this question from current project specifications may help explain outcomes that have proved harder to sustain than anticipated. We propose this as a question worth investigating, outline what a pilot study to test it might look like, and identify the primary empirical unknown on which the answer turns.

The Bioprecipitation Mechanism

Rain does not form from water vapour alone. Cloud droplets require a nucleation seed — a particle around which water molecules can crystallise at temperatures above −40°C. For most of Earth’s geological history, the most efficient providers of these seeds have not been dust or sea salt, but living organisms.

The mechanism is now well-established. Certain bacteria — most prominently Pseudomonas syringae, Pantoea, and Xanthomonas — produce ice-nucleating proteins (INpros) that trigger water crystallisation at temperatures as high as −2°C, far warmer than inorganic particles can achieve. These bacteria live on plant surfaces, are swept into the upper atmosphere by wind, and there function as highly efficient cloud seeds. Precipitation falls, watering the vegetation, propagating the bacteria, completing the cycle. This feedback loop — bioprecipitation — is not a peripheral quirk of atmospheric biology. It is, the evidence suggests, a significant driver of regional rainfall over vegetated land masses.

A 2026 study published in Science Advances substantially extended this picture. Eufemio et al. identified a previously unrecognised class of ice-nucleating proteins in fungi of the Mortierellaceae family — species including Mortierella alpina, Entomortierella parvispora, and Podila clonocystis. Unlike bacterial INpros, which are membrane-bound, fungal INpros are secreted directly into surrounding soil. They are water-soluble, smaller than their bacterial counterparts, extraordinarily stable across extreme pH ranges and high temperatures, and remain active in the environment long after the producing organism has moved on. These properties make fungal INpros more mobile, more durable, and potentially more significant in atmospheric ice nucleation than bacterial equivalents.

Phylogenetic analysis confirmed that fungi did not evolve this capacity independently. The gene encoding fungal INpros — an ortholog of the bacterial InaZ gene — was acquired through horizontal gene transfer from a Gammaproteobacterial ancestor at some point in deep evolutionary history. The gene was then modified over time, producing a secreted form more effective in soil environments than the original membrane-bound bacterial version.

Fusarium species, known since the early 1990s to possess ice-nucleation activity, have been detected not only in soil and on plant surfaces but in atmospheric samples and cloud water. The implication — that fungal INpros actively participate in cloud formation above vegetated landscapes — is no longer speculative. It is the parsimonious reading of the available data.

What Desertification Actually Destroys

Morris et al. (2014), in a landmark review in Global Change Biology, formalised the bioprecipitation feedback cycle as a measurable earth-system process linking vegetation cover, microbial aerosol emission, ice nucleation, precipitation, and plant growth.7 Their analysis is clear on one point: the land surface is the primary source of biogenic ice-nucleating particles in the atmosphere. Vegetated land produces them. Bare land does not.

Desertification, on this reading, does not merely remove trees and ground cover. It removes the organisms that generate regional rainfall. The loss of vegetation eliminates the hosts from which ice-nucleating bacteria are swept into clouds. The degradation of soil eliminates the fungal networks whose secreted INpros are among the most effective atmospheric ice nucleators yet characterised. What remains is land that can receive rain but may be less able to seed it — a weakening of the precipitation feedback loop that those who planted the vegetation were counting on to sustain it.

This connection was noted, with some directness, in an earlier formulation of the bioprecipitation hypothesis: the destruction of plants and their epiphytic microflora through overgrazing would reduce downwind nucleation of precipitation and accelerate desertification.8 The mechanism was identified. The engineering question was not drawn from it.

The Question That Has Not Been Asked

China’s afforestation programmes in arid and semi-arid regions represent the largest tree-planting effort in human history. On the Loess Plateau and across Inner Mongolia, hundreds of millions of trees have been planted over several decades with the ambition of reversing desertification, stabilising soils, and restoring regional hydrology. The outcomes have been mixed in ways that remain incompletely explained.

A 2025 study in Global Change Biology found that planted forests in China show significantly higher drought risk than natural forests — a finding whose mechanistic explanation remains contested.9 Contributing factors identified in the literature include inappropriate species selection, insufficient soil preparation, water table depletion by fast-growing monocultures, and mismatch between planted species and local precipitation regimes.10 What is absent from this diagnostic literature is any systematic consideration of whether the biogenic ice-nucleation infrastructure required to seed the desired rainfall feedback had been lost from target soils — and whether its restoration was ever part of the intervention design.

The question, to be precise, is this: if planted forests in China are showing unexpectedly high drought stress, and if regional rainfall feedback depends in part on the biogenic ice-nucleating capacity of soil fungal communities, is it possible that some proportion of the drought stress reflects the absence of a restored bioprecipitation cycle rather than — or in addition to — the factors currently under investigation? And if so, what would it cost to find out?

We are not claiming that bioprecipitation is the dominant rainfall mechanism in any of these regions. Moisture recycling, orographic effects, and large-scale circulation patterns all play roles that vary by geography. We are asking whether biogenic ice-nucleating capacity has been considered as a component of the restoration specification, and suggesting that the answer appears to be no.

Mycorrhizal Inoculation: The Existing Practice and Its Atmospheric Dimension

The restoration biology literature does not ignore fungi. Arbuscular mycorrhizal fungi (AMF) inoculation is an established and increasingly well-evidenced technique for improving seedling survival and establishment in degraded drylands. A 2013 meta-analysis of field experiments across 14 countries found that mycorrhizal inoculation was among the most effective ecotechnological interventions for dryland restoration, improving both seedling survival and growth.11

The justification for AMF inoculation is, however, entirely plant-centric. The literature frames it as a tool for improving nutrient uptake, drought resistance, and root establishment in degraded soils where native mycorrhizal networks have been depleted.12 A 2001 study in Mediterranean desertified ecosystems demonstrated that dual inoculation with AMF and nitrogen-fixing rhizobia not only improved plant establishment but enhanced soil fertility, organic matter, and aggregate stability.13 The case for mycorrhizal inoculation in restoration is well-made and well-supported.

What the restoration literature does not do is connect soil fungal communities to the bioprecipitation cycle. The ice-nucleating capacity of soil fungi — their role as atmospheric cloud seeds — is entirely absent from restoration engineering practice. Two bodies of knowledge exist in parallel without intersection: one asking how fungi help trees survive, the other asking how fungi help rain form. The question of whether restoring the former automatically restores the latter has not been asked.

The relevant fungi are not exclusively mycorrhizal partners. Fusarium and Mortierellaceae species are cosmopolitan soil inhabitants whose secreted INpros represent a direct atmospheric service. A restoration project that inoculates for plant health but has no specification for atmospheric ice-nucleation capacity is, at minimum, missing a question it should be asking.

The Central Empirical Unknown

The question on which everything else turns is one that has not been measured in any restoration context: what is the flux of biogenic ice-nucleating particles from restored soil communities into the atmosphere, and is it sufficient to materially influence cloud-altitude INP concentrations downwind?

Fungal INpros are durable in soil and water — stable across extreme pH and temperature ranges, persistent after the producing organism has moved on. But their pathway from secreted-into-soil to suspended-in-cloud involves uplift mechanisms — splash dispersal, dust events, adsorption onto mineral aerosol — that have not been measured in degraded or recovering dryland ecosystems. The atmospheric flux from restored soil communities, and the timescale on which that flux becomes ecologically meaningful, is genuinely unknown.

We flag this not as a reason to dismiss the question but as the precise reason to investigate it. Sands and Morris (1987) posed a structurally similar question — whether bacteria associated with plants were seeding rainfall — without being able to quantify the flux. That question eventually became Morris et al. (2014), became Eufemio et al. (2026). The trajectory from provocation to characterisation to application took decades, but it required someone to first formally ask whether the connection existed.14

The question we are posing is analogous: has the loss of ice-nucleating fungal communities from degraded soils measurably reduced biogenic INP flux in ways that impede rainfall recovery, and can deliberate inoculation restore that flux on timescales relevant to reforestation project viability? We do not know. We do not know because it has not been measured. It has not been measured because the question has not been formally posed in this context.

What a Pilot Study Might Look Like

The infrastructure for testing this question already partially exists. Mycorrhizal inoculation programmes are operational in Chinese dryland restoration. The incremental step would be to screen native soil fungal communities from analogous intact ecosystems for ice-nucleating activity, identify candidate Mortierella and Fusarium isolates adapted to the target environment, and layer them into existing inoculation protocols on 10–100 hectare test plots within afforestation areas showing documented drought stress.

Primary output metrics for such a pilot would not be vegetation survival alone — the standard current measure — but aerosol INP concentrations at altitude above treated and control plots, downwind precipitation anomalies over a multi-year monitoring window, and soil INP flux under varying wind and moisture conditions. These measurements would answer the empirical question directly.

The choice of native isolates is important. Evidence from dryland restoration suggests that mycorrhizal inoculation efficacy is site-specific, with native species consistently outperforming introduced ones.15 The same principle would apply to ice-nucleation-active species: the goal is not to introduce M. alpina from a European culture collection but to identify and culture the INP-active fungi that belong in the target ecosystem and are currently absent from its degraded soils.

The cost of such a pilot, layered onto existing inoculation programmes, would be modest relative to the overall investment in large-scale afforestation. The informational value — whether positive or negative — would be substantial. A null result clarifies the mechanism. A positive result opens a new lever in restoration engineering that currently does not exist in any project specification.

Caveats

Bioprecipitation is not the only or dominant rainfall mechanism in any of the regions discussed. Moisture recycling and large-scale atmospheric circulation are primary drivers of precipitation in arid continental interiors, and we make no claim that biogenic INP flux is the rate-limiting variable in any specific ecosystem. The argument is that it is a variable that has not been considered, not that it is the variable that explains everything.

The fungal species best characterised for ice-nucleating activity in the literature are not necessarily those most relevant to the target ecosystems discussed here. Mortierellaceae and Fusarium are cosmopolitan, but native INP-active species in Chinese dryland soils require dedicated screening rather than assumption from European or laboratory strains.

The relationship between bacterial and fungal INpros in the atmosphere is not fully characterised. Both are likely required for a functional bioprecipitation cycle, and an inoculation protocol targeting fungi alone may be insufficient. The bacterial component — plant-surface epiphytes swept into cloud — is itself dependent on restored vegetation, suggesting that the full cycle may only engage once both elements are present above threshold densities.

Conclusion

The bioprecipitation cycle — the feedback loop through which soil fungi and bacteria seed regional rainfall — is among the most significant atmospheric services provided by living ecosystems. Desertification destroys not only the visible vegetation of an ecosystem but the microbial infrastructure through which that ecosystem participates in generating its own precipitation.

Current de-desertification and afforestation engineering does not appear to account for this. Mycorrhizal inoculation, where it exists in restoration practice, is justified on plant-health grounds. The question of whether restored soil fungal communities also restore a component of the rainfall feedback loop that planted forests depend upon has not been formally posed in the project literature.

We are suggesting it should be. Sands and Morris asked a similar question in 1987 without being able to answer it quantitatively. The question eventually generated forty years of productive atmospheric microbiology. The question we are posing is cheaper to test, more practically bounded, and more immediately applicable to projects currently in the field. The pilots required to answer it are technically feasible with existing methods, at incremental cost within existing programmes.

The forest does not merely receive rain. It participates in making it. Whether restoring a forest without restoring its rain-making microbiome is sufficient for long-term self-sustainability is a question worth formally asking. We do not believe it has been.

Notes and References

1. Cloud condensation nuclei and ice nucleating particles: Pruppacher, H.R. & Klett, J.D. (1997). Microphysics of Clouds and Precipitation. Springer.
2. Bacterial ice-nucleating proteins: Lindow, S.E., Arny, D.C. & Upper, C.D. (1982). Bacterial ice nucleation: a factor in frost injury to plants. Plant Physiology, 70(4), 1084–1089.
3. Bioprecipitation as an earth-system feedback: Sands, D.C., Langhans, V.E., Scharen, A.L. & de Smet, G. (1982). The association between bacteria and rain and possible resultant meteorological implications. Journal of the Hungarian Meteorological Service, 86, 148–152.
4. Fungal INpros and horizontal gene transfer: Eufemio, R.J., Rojas, M., Shaw, K., et al. (2026). A previously unrecognized class of fungal ice-nucleating proteins with bacterial ancestry. Science Advances, 12(11), eaed9652. DOI: 10.1126/sciadv.aed9652.
5. Ibid. Horizontal gene transfer of the InaZ ortholog from Gammaproteobacteria to a fungal ancestor of Mortierellaceae.
6. Fusarium in atmospheric samples: Kunert, A.T. et al. (2019). Macromolecular fungal ice nuclei in Fusarium: effects of physical and chemical processing. Biogeosciences, 16, 4647–4659. Also: Schwidetzky, R. et al. (2023). Functional aggregation of cell-free proteins enables fungal ice nucleation. Proceedings of the National Academy of Sciences, 120(46).
7. Bioprecipitation feedback cycle: Morris, C.E., Conen, F., Huffman, J.A., Phillips, V., Pöschl, U. & Sands, D.C. (2014). Bioprecipitation: a feedback cycle linking Earth history, ecosystem dynamics and land use through biological ice nucleators in the atmosphere. Global Change Biology, 20(2), 341–351.
8. Bioprecipitation and desertification: Sands, D.C. & Morris, C.E. (1987). A Hypothetical Bioprecipitation Cycle Involving Ice-Nucleating and Dew-Condensing Bacteria, Plants and Rainfall. In Biological Ice Nucleation and Its Applications, APS Press.
9. Drought risk in Chinese planted forests: Ma, L.L. et al. (2025). Planted forests in China have higher drought risk than natural forests. Global Change Biology, 31, e70055.
10. Chinese afforestation and desertification: Cao, S. (2008). Why large-scale afforestation efforts in China have failed to solve the desertification problem. Environmental Science & Technology, 42(5), 1826–1831.
11. Meta-analysis of ecotechnological restoration: Cortina, J. et al. (2013). Ecotechnology as a tool for restoring degraded drylands: a meta-analysis of field experiments. Forest Ecology and Management. doi:10.1016/j.foreco.2013.09.028.
12. AMF inoculation in restoration: Muleta, D. (2016). The Potential Role of Arbuscular Mycorrhizal Fungi in the Restoration of Degraded Lands. Frontiers in Microbiology. PMC4960231.
13. Mediterranean desertification and plant-microbe symbiosis: Requena, N. et al. (2001). Management of Indigenous Plant-Microbe Symbioses Aids Restoration of Desertified Ecosystems. Proceedings of the National Academy of Sciences, 98(2), 821–826.
14. The precedent provocation: Sands & Morris (1987), op. cit. The trajectory from that hypothesis to Morris et al. (2014) to Eufemio et al. (2026) illustrates the timescale on which formally posed questions in atmospheric microbiology become actionable.
15. Site-specificity of AMF inoculation: Poudel, M. et al. (2024). Arbuscular Mycorrhizal Fungi and Rhizobium Improve Nutrient Uptake and Microbial Diversity Relative to Dryland Site-Specific Soil Conditions. Microorganisms, 12(4), 667.

Philip Hampsheir writes on science, history, and their intersections. This paper was developed in April 2026.