# Paleoecological Data Integration Reduces Uncertainty in Nutrient Reduction Targets for Cyanobacterial Bloom Management

## Metadata

- **SPORE ID**: SPR-2026-F50A
- **Domaines**: Paleoecology x Aquatic Ecosystems and Phytoplankton Dynamics
- **Date de generation**: 2026-08-23
- **Panel consensus score**: 6.4/10
- **Novelty score**: 0.7/1.0
- **Panel verdict**: publish_brief

## Abstract

If paleoecological reconstructions of pre-disturbance nutrient baselines and historical bloom thresholds are integrated as Bayesian priors into a lake eutrophication model (PCLake), then the uncertainty (coefficient of variation) of the predicted total phosphorus reduction target required to achieve a 20% reduction in cyanobacterial biomass will decrease by at least 20% compared to a model calibrated on contemporary monitoring data alone, because the paleo data provide information on long-term system states and resilience thresholds not captured by short-term monitoring.

The proposed mechanism involves 8 causal steps: (1) Sediment cores are collected from 10-15 lakes, and proxies (diatoms, pigments, s -> (2) The reconstructed time series are used to identify the pre-disturbance TP baseli -> (3) These paleo-derived baseline and threshold values are incorporated into the PCLa -> (4) The PCLake model is calibrated using (a) contemporary monitoring data only, (b) .

Literature grounding on 11 verified references yields a novelty score of 0.7 (novel). A 3-phase experimental protocol (budget: €50k-150k, timeline: 12-18 months) is proposed, starting with in silico validation. A panel of 5 expert reviewers reached a consensus score of 6.4/10.

## 1. Hypothese et mecanisme propose

### 1.1 Formulation formelle

If paleoecological reconstructions of pre-disturbance nutrient baselines and historical bloom thresholds are integrated as Bayesian priors into a lake eutrophication model (PCLake), then the uncertainty (coefficient of variation) of the predicted total phosphorus reduction target required to achieve a 20% reduction in cyanobacterial biomass will decrease by at least 20% compared to a model calibrated on contemporary monitoring data alone, because the paleo data provide information on long-term system states and resilience thresholds not captured by short-term monitoring.

### 1.2 Variables

**Variables independantes :**

| Variable | Type | Plage | Unite |
|----------|------|-------|-------|
| Data integration strategy | categorical | Contemporary-only, Paleo-informed (Bayesian prior), Contemporary-augmented (synthetic 20-year) | N/A |
| Proxy error inflation factor | continuous | 1.0 - 2.0 | dimensionless (multiplicative factor) |
| Lake type | categorical | Subalpine (low climate change), Lowland (high climate change) | N/A |

**Variables dependantes :**

| Variable | Type | Direction attendue | Unite |
|----------|------|-------------------|-------|
| Coefficient of variation (CV) of predicted TP reduction target | continuous | decrease | % |
| Posterior distribution width of the TP threshold for cyanobacterial blooms | continuous | decrease | log10(µg P/L) |
| Accuracy of predicted cyanobacterial biomass under historical conditions | continuous | increase | RMSE (mg/L) |

### 1.3 Chaine causale

1. Step 1: Sediment cores are collected from 10-15 lakes, and proxies (diatoms, pigments, sedaDNA) are analyzed to reconstruct historical TP concentrations (via diatom transfer functions with RMSEP ±0.2-0.4 log10 µg/L) and cyanobacterial biomass (via pigments and sedaDNA) over the last 200 years.
1. Step 2: The reconstructed time series are used to identify the pre-disturbance TP baseline (e.g., median TP from 1800-1850) and the historical TP threshold at which cyanobacterial blooms first appeared (e.g., first sustained increase in pigment concentrations above a baseline).
1. Step 3: These paleo-derived baseline and threshold values are incorporated into the PCLake model as Bayesian priors for the parameters governing the TP-cyanobacteria relationship, specifically the critical TP loading and the internal phosphorus loading coefficient.
1. Step 4: The PCLake model is calibrated using (a) contemporary monitoring data only, (b) contemporary data plus paleo-informed priors, and (c) contemporary data augmented with 20 years of synthetic data.
1. Step 5: For each calibration strategy, the model is used to predict the TP reduction target required to achieve a 20% reduction in cyanobacterial biomass, and the uncertainty (CV) of this target is calculated via Monte Carlo simulations.
1. Step 6: The CV of the TP reduction target from the paleo-informed model is compared to the CV from the contemporary-only and contemporary-augmented models to quantify the information gain from the paleo data.
1. Step 7: A power analysis is conducted to determine the number of lakes needed to detect a 20% reduction in CV with a paired t-test (alpha=0.05, power=0.80), accounting for inter-lake variance.
1. Step 8: For lakes with significant climate warming, a temperature-dependent TP threshold displacement term is added to the model to account for non-stationarity, and the analysis is repeated to test the robustness of the results.

**Hypotheses cles :**

- Diatom transfer functions for TP are valid for the target lakes, with RMSEP within the specified range (±0.2-0.4 log10 µg/L).
- Sedimentary pigments and sedaDNA provide semi-quantitative but reliable reconstructions of past cyanobacterial biomass trends.
- The PCLake model structure is sufficiently accurate to represent the TP-cyanobacteria relationship in the target lakes, and the paleo data can be integrated without violating structural assumptions (e.g., no equifinality issues).
- The pre-disturbance period (1800-1850) represents a stable, low-nutrient baseline for the target lakes.
- The relationship between TP and cyanobacterial biomass is stationary over the calibration period, except for the explicitly modeled temperature and legacy P effects.

**Inconnues identifiees :**

- The exact magnitude of taphonomic bias in sedaDNA and pigment records for the target lakes.
- The degree to which internal phosphorus loading (legacy P) affects the TP-cyanobacteria relationship in each lake.
- The extent to which climate change has already shifted the TP threshold for cyanobacterial blooms in the target lakes.
- The optimal method for integrating paleo data into PCLake (e.g., priors vs. soft calibration targets) without introducing bias.

### 1.4 Conditions aux limites

- **Lakes must have identifiable sediment records spanning at least 150 years, with minimal sediment mixing (e.g., varved or well-dated cores).** — Without a clear stratigraphy, the temporal resolution of the paleo reconstructions is insufficient to identify pre-disturbance baselines and historical thresholds.
- **The analysis is restricted to lakes where the diatom transfer function for TP has been validated or can be validated with local training sets.** — Diatom-based TP reconstructions are only reliable within the range of the calibration dataset; extrapolation beyond this range introduces unacceptable uncertainty.
- **The PCLake model must be calibrated for the target lakes with a Nash-Sutcliffe efficiency > 0.5 for contemporary TP and cyanobacterial biomass.** — If the model cannot reproduce contemporary dynamics, it is unlikely to provide reliable predictions under alternative calibration strategies.
- **For lakes with significant climate warming (>0.5°C over the last 50 years), a temperature-dependent TP threshold displacement term must be included in the model.** — Climate change can alter the TP-cyanobacteria relationship, violating the stationarity assumption and biasing the paleo-informed predictions.
- **The analysis is limited to lakes with no major hydrological modifications (e.g., dam construction, water diversion) that would fundamentally alter the nutrient loading regime.** — Such modifications create discontinuities in the sediment record that cannot be adequately captured by the paleo-proxies or the model.

### 1.5 Cadre theorique

Theoretical framework: Resilience theory and regime shift ecology, applied to lake eutrophication. The hypothesis posits that paleoecological data can reveal the historical stability landscape of lake ecosystems, including the position of critical nutrient thresholds (tipping points) and the resilience of the clear-water state. This information is used to parameterize process-based models (PCLake) to reduce uncertainty in management targets, aligning with the concept of 'reference conditions' in restoration ecology.

## 2. Etat de l'art et positionnement

### 2.1 Travaux les plus proches

- **[2023] Early human impact on lake cyanobacteria revealed by a Holocene record of sedimentary ancient DNA** — [10.1038/s42003-023-04430-z](https://doi.org/10.1038/s42003-023-04430-z)
  - Similarite: related
  - Difference cle: This paper demonstrates the use of sedaDNA to reconstruct cyanobacteria history over the Holocene, but it does not explicitly link these reconstructions to the design of nutrient reduction strategies for HAB management. The SPORE hypothesis extends this by proposing a direct application of such paleoecological data to parameterize models and identify baselines for contemporary management.
- **[2021] Using a lake sediment record to infer the long-term history of cyanobacteria and the recent rise of an anatoxin producing Dolichospermum sp.** — [10.1016/j.hal.2020.101971](https://doi.org/10.1016/j.hal.2020.101971)
  - Similarite: related
  - Difference cle: This paper uses a sediment core to reconstruct cyanobacteria history and links it to nutrient inputs, but it focuses on a single lake's history rather than proposing a general framework for using paleoecology to inform nutrient reduction strategies across different systems.
- **[2022] Molecular and Pigment Analyses Provide Comparative Results When Reconstructing Historic Cyanobacterial Abundances from Lake Sediment Cores** — [10.3390/microorganisms10020279](https://doi.org/10.3390/microorganisms10020279)
  - Similarite: related
  - Difference cle: This paper validates the methodological tools (molecular vs. pigment) for reconstructing cyanobacteria from sediments, which is a core component of the SPORE hypothesis. However, it does not directly address the application of these tools to inform nutrient management strategies.

### 2.2 Base de preuves

- **[2023] Early human impact on lake cyanobacteria revealed by a Holocene record of sedimentary ancient DNA** — [10.1038/s42003-023-04430-z](https://doi.org/10.1038/s42003-023-04430-z)
  - Type: direct | Citations: 17
  - Directly supports the hypothesis by demonstrating that sedaDNA can be used to reconstruct long-term cyanobacteria dynamics and link them to human activity, providing a historical baseline for HABs.
- **[2021] Using a lake sediment record to infer the long-term history of cyanobacteria and the recent rise of an anatoxin producing Dolichospermum sp.** — [10.1016/j.hal.2020.101971](https://doi.org/10.1016/j.hal.2020.101971)
  - Type: direct | Citations: 17
  - Directly supports the hypothesis by using a sediment core to reconstruct the history of cyanobacteria and linking it to nutrient inputs, showing how paleoecology can reveal the legacy of nutrient loading and the development of toxic blooms.
- **[2022] Molecular and Pigment Analyses Provide Comparative Results When Reconstructing Historic Cyanobacterial Abundances from Lake Sediment Cores** — [10.3390/microorganisms10020279](https://doi.org/10.3390/microorganisms10020279)
  - Type: direct | Citations: 17
  - Directly supports the methodological bridge of the hypothesis by validating that molecular methods (e.g., sedaDNA) can yield results similar to traditional paleolimnological proxies (pigments) for reconstructing cyanobacteria, making them reliable tools for management.
- **[2021] From Water into Sediment—Tracing Freshwater Cyanobacteria via DNA Analyses** — [10.3390/microorganisms9081778](https://doi.org/10.3390/microorganisms9081778)
  - Type: indirect | Citations: 22
  - Indirectly supports the hypothesis by demonstrating the feasibility of tracing cyanobacteria from water into sediments using DNA analyses, which is a key step in using sediment archives for reconstructing past bloom intensity.
- **[2022] Properties of sediment dissolved organic matter respond to eutrophication and interact with bacterial communities in a plateau lake.** — [10.1016/j.envpol.2022.118996](https://doi.org/10.1016/j.envpol.2022.118996)
  - Type: indirect | Citations: 54
  - Indirectly supports the hypothesis by showing that sediment cores can record the historical response of a lake to eutrophication, providing a long-term perspective that is not captured by contemporary monitoring.
- **[2020] Paleolimnological response to climate variability during Late Glacial and Holocene times: A record from Lake Arturo, located in the Fuegian steppe, southern Argentina** — [10.1016/j.palaeo.2020.109737](https://doi.org/10.1016/j.palaeo.2020.109737)
  - Type: analogous | Citations: 18
  - Analogous support: demonstrates that paleolimnological proxies (diatoms, ostracods) can reconstruct long-term hydrological and climate-driven variability, which is the type of natural experiment the SPORE hypothesis proposes to use for understanding ecosystem responses.
- **[2021] Phosphorus internal loading and sediment diagenesis in a large eutrophic lake (Lake Chaohu, China).** — [10.1016/j.envpol.2021.118471](https://doi.org/10.1016/j.envpol.2021.118471)
  - Type: indirect | Citations: 74
  - Indirectly supports the hypothesis by highlighting the importance of legacy phosphorus in sediments, which is a key reason why historical baselines and thresholds from paleoecology are needed for effective management.
- **[2019] Consideration of species-specific diatom indicators of anthropogenic stress in the Great Lakes** — [10.1371/journal.pone.0210927](https://doi.org/10.1371/journal.pone.0210927)
  - Type: indirect | Citations: 19
  - Indirectly supports the hypothesis by validating the use of diatom transfer functions, a key paleoecological tool for reconstructing past nutrient concentrations, while also cautioning about their limitations.
- **[2021] Ecological restoration guided by historical reference conditions can increase resilience to climate change of southwestern U.S. Ponderosa pine forests** — [10.1016/J.FORECO.2021.119256](https://doi.org/10.1016/J.FORECO.2021.119256)
  - Type: analogous | Citations: 39
  - Analogous support: demonstrates the general principle that using historical reference conditions can guide restoration and increase resilience to climate change, which is the core idea of the SPORE hypothesis applied to aquatic ecosystems.
- **[2023] Regarding reference state to identify priority areas for ecological restoration in a karst region.** — [10.1016/j.jenvman.2023.119214](https://doi.org/10.1016/j.jenvman.2023.119214)
  - Type: analogous | Citations: 39
  - Analogous support: reinforces the importance of defining reference states for ecological restoration, a concept central to the SPORE hypothesis for setting nutrient baselines.
- **[2023] Ecological quality assessment: A framework to report ecosystems quality and their dynamics from reference conditions** — [10.1002/ecs2.4726](https://doi.org/10.1002/ecs2.4726)
  - Type: analogous | Citations: 25
  - Analogous support: provides a framework for assessing ecological quality relative to reference conditions, which is the conceptual basis for using paleoecological baselines in the SPORE hypothesis.

### 2.3 Contre-preuves et limitations connues

- **[addressable] From Water into Sediment—Tracing Freshwater Cyanobacteria via DNA Analyses**
  - The study found that taphonomic processes can skew the taxonomic abundances of cyanobacteria in sediment records, which could complicate the interpretation of past bloom intensity and community composition from sedaDNA.
- **[addressable] Consideration of species-specific diatom indicators of anthropogenic stress in the Great Lakes**
  - The paper cautions that many diatom species have little or no indicator value or have confusing relationships with multiple environmental variables, which could limit the accuracy of diatom transfer functions for reconstructing past nutrient concentrations.

### 2.4 Evaluation de nouveaute

- **Score**: 0.7/1.0
- **Verdict**: novel

## 3. Predictions falsifiables

| # | Prediction | Borne quantitative | Methode | H0 | Test statistique |
|---|-----------|-------------------|---------|-----|-----------------|
| 1 | Integrating paleoecological reconstructions as Bayesian priors into PCLake will reduce the CV of the predicted TP reduction target by at least 20% compared to a model calibrated on contemporary monitoring data alone. | The CV of the TP reduction target will be 20-40% lower (e.g., from 30% to 18-24%) when paleo priors are used, even when proxy error is inflated by a factor of 2. | Monte Carlo simulations with PCLake, using posterior distributions of the TP reduction target from three calibration strategies (contemporary-only, paleo-informed, contemporary-augmented). | H0: The CV of the TP reduction target from the paleo-informed model is not significantly different from the CV from the contemporary-only model (i.e., reduction ≤ 0%). | Paired t-test on log-transformed CV values from 1000 bootstrap resamples, alpha=0.05, one-tailed. |
| 2 | The posterior distribution of the TP threshold for cyanobacterial blooms will be narrower (lower variance) when paleo data are included, compared to the prior distribution. | The standard deviation of the posterior TP threshold will be at least 30% narrower than the prior, even with a 2x inflation of proxy error (e.g., from ±0.5 to ±0.35 log10 µg/L). | Bayesian calibration of PCLake, comparing prior and posterior distributions of the TP threshold parameter. | H0: The posterior standard deviation is ≥ 70% of the prior standard deviation. | F-test for equality of variances, alpha=0.05. |
| 3 | The reduction in CV from paleo-informed calibration will be significantly greater than the reduction from simply adding 20 years of synthetic contemporary monitoring data. | The CV reduction from paleo-informed calibration will be at least 10 percentage points greater than the CV reduction from contemporary-augmented calibration (e.g., 25% vs. 15% reduction). | Comparison of CV values from Monte Carlo simulations of the three calibration strategies. | H0: The CV reduction from paleo-informed calibration is ≤ the CV reduction from contemporary-augmented calibration. | Paired t-test, alpha=0.05, one-tailed. |
| 4 | The power analysis will show that 10-15 lakes are sufficient to detect a 20% reduction in CV with a paired t-test, given an inter-lake CV of 15-25%. | Power ≥ 0.80 with alpha=0.05 for a one-tailed paired t-test, assuming an effect size of 0.5-0.7 (Cohen's d) for the CV reduction. | Analytical power calculation using the expected inter-lake variance from the literature and pilot data. | H0: Power < 0.80 for the specified effect size and sample size. | Power analysis (e.g., using G*Power or R package 'pwr'). |

## 4. Protocole experimental

**Timeline globale**: 12-18 months
**Budget global**: €50k-150k

### 4.1 Phase 1 — In Silico Validation (Virtual Twin & Sensitivity Analysis)

**Objectif**: To falsify the core hypothesis using existing datasets and a virtual lake environment, before any new sediment cores are collected. This phase tests whether the proposed mechanism (Bayesian prior integration) can theoretically produce the predicted 20% CV reduction, and identifies the most sensitive parameters and data requirements.

**Methodologie**: 1. **Data Assembly**: Compile a training dataset of 20-30 lakes from open-access sources (e.g., LAGOS-US, NLA, GLEON) with: (a) ≥10 years of contemporary TP and cyanobacterial biomass (or chlorophyll-a as proxy) monitoring data, (b) published or publicly available diatom-inferred TP reconstructions (or a validated regional transfer function), (c) sediment core data (pigments, sedaDNA) if available. 2. **Virtual Lake Setup**: Use a representative subset (n=5) of these lakes to create 'virtual twins' in PCLake (v1.2 or newer). Calibrate the model to contemporary data to achieve NSE > 0.5. 3. **Paleo-Data Simulation**: For each virtual twin, simulate a synthetic 'paleo-record' by running the model with a known historical TP loading scenario (e.g., pre-disturbance baseline from 1800-1850, then gradual increase). Add realistic noise to this simulated record to mimic proxy error (RMSEP ±0.2-0.4 log10 µg/L). 4. **Bayesian Calibration**: Implement three calibration strategies using the R package 'rjags' or 'Stan' (via 'brms'): (a) Contemporary-only (prior: vague), (b) Paleo-informed (prior: informative, based on the simulated paleo-TP baseline and threshold), (c) Contemporary-augmented (add 20 years of synthetic data generated by resampling contemporary data with noise). 5. **Monte Carlo Uncertainty Propagation**: For each strategy, run 1000 Monte Carlo simulations to generate posterior distributions of the TP reduction target (required to achieve 20% cyanobacterial biomass reduction). Calculate the CV of this target. 6. **Comparative Analysis**: Perform paired t-tests on log-transformed CV values across the three strategies. Also, test the effect of proxy error inflation (factor 1.0, 1.5, 2.0) on the CV reduction. 7. **Power Analysis**: Use the 'pwr' package in R to calculate the required sample size (lakes) to detect the observed effect size (Cohen's d) with power = 0.80 and alpha = 0.05.

- Cout: €1,500 - €3,000 (mostly computational time & personnel)
- Duree: 6-8 weeks
- Equipement: High-performance computing cluster (or cloud VM with 16+ cores)
- Logiciels: PCLake (Fortran or R wrapper), R (v4.2+), R packages: rjags, brms, pwr, deSolve, sensitivity, JAGS (v4.3+)

**Criteres de succes :**

- CV reduction from paleo-informed vs. contemporary-only calibration: ≥ 20% reduction (e.g., from 30% to ≤24%) in at least 4 out of 5 virtual lakes, even with proxy error inflation factor = 2.0
- Posterior narrowing of TP threshold parameter: ≥ 30% reduction in standard deviation of the posterior TP threshold distribution compared to the prior
- Superiority over synthetic data augmentation: CV reduction from paleo-informed is ≥ 10 percentage points greater than from contemporary-augmented (e.g., 25% vs. 15%)
- Power analysis: Power ≥ 0.80 to detect a 20% CV reduction with n ≤ 15 lakes

- **GO**: All four success criteria are met in the virtual environment. The mechanism is theoretically sound and the required sample size is feasible.
- **NO-GO**: The CV reduction is < 10% in the majority of virtual lakes, or the required sample size exceeds 25 lakes. The hypothesis is likely false or the effect is too small to be practically significant.
- **PIVOT**: The CV reduction is 10-20%, or only met with proxy error inflation factor = 1.0. Pivot: Re-evaluate the proxy error assumptions, refine the prior construction method (e.g., use a different parameterization of the prior), or focus on a subset of lake types (e.g., only lowland lakes) where the effect is stronger.

### 4.2 Phase 2 — Minimal Experimental Validation (Pilot Lake Study)

**Objectif**: To validate the core mechanism with a minimal physical test: collect and analyze sediment cores from 2-3 lakes to confirm that the paleo-proxies (diatoms, pigments) can reliably reconstruct historical TP and cyanobacterial biomass trends, and that these reconstructions can be integrated into PCLake to produce a measurable reduction in prediction uncertainty.

**Methodologie**: 1. **Lake Selection**: Select 2-3 lakes from the Phase 1 virtual twins that meet boundary conditions (varved or well-dated sediments, no major hydrological modifications). Prioritize lakes with existing monitoring data and some preliminary paleo data. 2. **Sediment Coring & Dating**: Collect one short core (50-100 cm) from the deepest point of each lake using a UWITEC corer. Date the cores using 210Pb and 137Cs (via gamma spectrometry) to establish a chronology for the last 200 years. 3. **Proxy Analysis**: (a) Diatoms: Prepare slides and count ~400 valves per sample at 2-5 cm intervals. Apply a published or locally validated diatom-TP transfer function (e.g., from EDDI) to reconstruct historical TP. (b) Pigments: Analyze for cyanobacterial pigments (echinenone, canthaxanthin, zeaxanthin) using HPLC. (c) sedaDNA: Extract DNA and use qPCR or metabarcoding (16S rRNA gene) to quantify cyanobacterial abundance. 4. **Data Integration & Modeling**: Use the reconstructed TP baseline (median 1800-1850) and the historical TP threshold (first sustained increase in pigments) to construct informative priors for PCLake. Calibrate PCLake with (a) contemporary data only and (b) contemporary + paleo priors. 5. **Uncertainty Quantification**: Run Monte Carlo simulations to calculate the CV of the TP reduction target for each calibration strategy. 6. **Comparison**: Compare the CV reduction from the pilot lakes to the Phase 1 predictions.

- Cout: €8,000 - €15,000 (coring, dating, lab analysis, personnel)
- Duree: 2-3 months
- Equipement: UWITEC corer (or similar), Gamma spectrometer (for 210Pb/137Cs), HPLC system (for pigments), PCR thermocycler & qPCR machine (for sedaDNA), Microscope with oil immersion (1000x) for diatoms
- Logiciels: R (for statistical analysis & modeling), PCLake, RJAGS/Stan, C2 or Tilia (for stratigraphic diagrams)

**Criteres de succes :**

- Proxy reliability: Diatom-inferred TP reconstructions have RMSEP within ±0.4 log10 µg/L of the validation set, and pigment/sedaDNA trends correlate with known historical eutrophication events (e.g., from archival data).
- CV reduction in pilot lakes: At least 2 out of 3 lakes show a CV reduction ≥ 15% (a slightly lower bar than Phase 1, acknowledging real-world noise).
- Model performance: PCLake calibration with paleo priors maintains NSE > 0.5 for contemporary TP and cyanobacteria.

- **GO**: At least 2 out of 3 lakes show a CV reduction ≥ 15%, and the proxy reconstructions are validated. The mechanism works in practice.
- **NO-GO**: No lake shows a CV reduction ≥ 10%, or the proxy reconstructions are unreliable (e.g., poor dating, low pigment preservation). The hypothesis is not supported by empirical evidence.
- **PIVOT**: Only 1 lake shows a CV reduction ≥ 15%, or the reduction is inconsistent. Pivot: Investigate why the mechanism fails in specific lakes (e.g., legacy P effects, climate change). Refine the selection criteria for Phase 3 to exclude such lakes, or adjust the prior construction method.

### 4.3 Phase 3 — Full Experimental Protocol (Multi-Lake Validation & Publication)

**Objectif**: To rigorously validate the hypothesis across a diverse set of 10-15 lakes, produce a publishable dataset, and generate robust, generalizable conclusions about the value of paleoecological data for nutrient management.

**Methodologie**: 1. **Lake Selection**: Select 10-15 lakes based on Phase 2 results and a formal power analysis (target: power ≥ 0.80). Ensure a balanced design across lake types (subalpine vs. lowland) and climate change scenarios. 2. **Field Campaign**: Collect sediment cores from all selected lakes. Use a consistent coring protocol (e.g., UWITEC corer, 2-3 replicate cores per lake). 3. **Laboratory Analyses**: (a) Dating: 210Pb, 137Cs, and 14C (where needed) for all cores. (b) Diatoms: Count and apply transfer function for TP reconstruction. (c) Pigments: HPLC analysis for cyanobacterial pigments. (d) sedaDNA: Metabarcoding or qPCR for cyanobacterial community composition and abundance. 4. **Modeling & Bayesian Calibration**: (a) For each lake, calibrate PCLake with three strategies: contemporary-only, paleo-informed, contemporary-augmented. (b) Use a standardized Bayesian framework (e.g., 'brms' with a PCLake emulator) to ensure consistency. (c) Account for climate change by adding a temperature-dependent TP threshold displacement term for lakes with >0.5°C warming. 5. **Uncertainty Quantification & Statistical Analysis**: (a) Calculate the CV of the TP reduction target for each lake and strategy. (b) Perform a paired t-test (or Wilcoxon signed-rank test if non-normal) on the CV reductions. (c) Conduct a mixed-effects model to test the effect of lake type and climate change scenario on the CV reduction. 6. **Publication & Dissemination**: Prepare manuscripts for submission to high-impact journals (e.g., *Limnology and Oceanography*, *Global Change Biology*).

- Cout: €50,000 - €120,000 (fieldwork, lab analysis, personnel, publication fees)
- Duree: 8-12 months
- Equipement: Full paleolimnological field equipment (coring, GPS, etc.), Analytical lab equipment (HPLC, GC-MS, PCR, etc.), High-performance computing cluster
- Logiciels: PCLake, R (with packages: brms, rjags, pwr, lme4, ggplot2), QGIS (for spatial analysis)

**Criteres de succes :**

- Primary hypothesis test: Statistically significant (p < 0.05, one-tailed) ≥ 20% reduction in CV of TP reduction target with paleo-informed calibration vs. contemporary-only, across the full lake set.
- Secondary hypothesis test: Statistically significant (p < 0.05) ≥ 30% narrowing of the posterior TP threshold distribution.
- Tertiary hypothesis test: CV reduction from paleo-informed is significantly greater (p < 0.05) than from contemporary-augmented, by at least 10 percentage points.
- Generalizability: The CV reduction is consistent (≥ 15%) in both subalpine and lowland lake types, and the effect is not entirely lost in lakes with significant climate warming.

- **GO**: All primary and secondary success criteria are met. The hypothesis is validated, and the results are suitable for publication.
- **NO-GO**: The primary hypothesis test fails (p > 0.05 or CV reduction < 20%). The approach does not provide the expected benefit.
- **PIVOT**: The primary hypothesis is met, but the effect is not generalizable (e.g., only works in lowland lakes). Pivot: Refine the scope of the conclusion and recommend the approach for specific lake types only.

### 4.4 Quick start : comment demarrer aujourd'hui

- **Peut demarrer maintenant**: Oui
- **Premiere action**: Download the LAGOS-US and NLA datasets, and install R, JAGS, and the 'rjags' and 'brms' packages. Begin assembling a table of candidate lakes with both contemporary and paleo data availability.
- **Outils**: R (v4.2+), RStudio, JAGS (v4.3+), PCLake (R wrapper or Fortran binary), Git (for version control)
- **Donnees ouvertes**: LAGOS-US (lagoslakes.org), US EPA National Lakes Assessment (NLA), EDDI (European Diatom Database), GLEON (gleon.org), NOAA NCEI Paleoclimatology (for sediment core data)

## 5. Analyse d'impact

### 5.1 Impact scientifique

Novelty score: 0.7/1.0 (novel)

### 5.2 Applications industrielles et marche

**Score industriel**: 5.8/10

**Forces :**
- Cible un besoin réglementaire payant : les agences de bassin (ex : Agences de l'Eau en France, RIVM aux Pays-Bas, US EPA) et les bureaux d'études (ex : Suez, Arcadis, WSP) doivent justifier des cibles de réduction de nutriments (DCE, Water Framework Directive) avec des marges d'incertitude réduites pour éviter des surcoûts de traitement (stations d'épuration) ou des amendes.
- L'approche 'paleo-informed' crée un avantage compétitif fort en transformant des données sédimentaires (souvent perçues comme académiques) en un actif de modélisation prédictif, permettant de breveter une méthode de calibration (Bayesian priors + PCLake) difficile à copier sans les jeux de données paléo historiques propriétaires.

**Faiblesses :**
- Barrière à l'entrée majeure : le marché est fragmenté et dominé par des acteurs établis (consultants en modélisation) qui utilisent déjà des approches standards (MOHID, Delft3D). Les gestionnaires de l'eau sont risk-averse et exigent des validations longues (5-10 ans) avant d'adopter une nouvelle méthode, surtout si elle repose sur des carottes sédimentaires (coûts de terrain et d'analyse non négligeables, ~50-100k€ par lac).
- Le ROI est incertain : le gain de 20% de réduction de CV sur une cible de TP ne se traduit pas directement en euros pour le client final. Le seul acheteur potentiel est le secteur public (agences de l'eau), qui a des budgets contraints et peut considérer cette approche comme une 'niche académique' plutôt qu'une nécessité opérationnelle, surtout si les modèles existants sont 'suffisamment bons' pour la réglementation.

**Recommandation**: Ne pas commercialiser en tant que produit autonome. Intégrer cette méthode comme un 'module de calibration avancée' dans une offre de conseil existante (ex : mise à jour des SDAGE ou des plans de gestion de bassin). Cibler les agences de l'eau en Europe du Nord (Pays-Bas, Danemark) où la pression sur les blooms cyanobactériens est forte et les budgets R&D existent. Déposer un brevet sur le 'système de génération de priors paléo pour modèles de qualité d'eau' et créer une base de données propriétaire de carottes sédimentaires numérisées pour verrouiller les clients. Le ROI est long (5-7 ans) et dépend de la capacité à standardiser le protocole (Phase 3) pour le rendre aussi simple qu'une analyse de routine.

### 5.3 Opportunites de financement

| Programme | Agence | Fit | Budget type | Taux succes |
|-----------|--------|-----|-------------|-------------|
| Biodiversa+ (Call 2024/2025) - BiodivRestore | European Commission / Partenariat europeen Biodiversa+ | 0.85 | 1.5 - 2.5 M€ (consortium international de 3 a 6 partenaires) | 15-20% (taux de succes moyen des appels Biodiversa) |
| ERC Starting Grant (StG) - 2026 | European Research Council | 0.65 | 1.5 M€ sur 5 ans | 12-15% (taux de succes moyen StG) |
| Appel a projets generique (AAPG) - CE50 (Sante, environnement, societes) ou CE53 (Ecosystemes) | Agence Nationale de la Recherche (ANR) | 0.75 | 200k - 500k€ (projet JCJC ou PRC) | 15-20% (taux de succes moyen ANR) |

- **Biodiversa+ (Call 2024/2025) - BiodivRestore** (European Commission / Partenariat europeen Biodiversa+): Ce programme cible explicitement les solutions fondees sur la nature et la restauration des ecosystemes. L'approche interdisciplinaire (paleoecologie + modelisation) et l'objectif de reduction de l'incertitude pour la gestion de l'eau repondent parfaitement aux axes 'restauration des ecosystemes d'eau douce' et 'outils de decision'. Le consortium devra etre elargi a 3-4 pays europeens, ce qui est coherent avec la validation multi-lacs de la phase 3.
- **ERC Starting Grant (StG) - 2026** (European Research Council): Le projet a un fort potentiel 'ground-breaking' (integration de proxies paleo dans des modeles mecanistiques). Cependant, le TRL est trop faible pour un ERC sans donnees preliminaires. Si le porteur est eligible (moins de 7 ans post-PhD), il pourrait postuler en proposant ce projet comme un 'work package' de demarrage d'un programme plus large sur les seuils de resilience des ecosystemes aquatiques, mais il faudrait ajouter une composante de developpement theorique plus ambitieuse (ex : developpement d'un nouveau formalisme bayesien pour les modeles de qualite d'eau).
- **Appel a projets generique (AAPG) - CE50 (Sante, environnement, societes) ou CE53 (Ecosystemes)** (Agence Nationale de la Recherche (ANR)): L'ANR finance des projets de recherche fondamentale avec un fort accent sur la methodologie. Le budget demande (50-150k€) est inferieur a l'enveloppe classique, mais le projet pourrait etre elargi pour inclure un volet de developpement d'un package R/Shiny pour la diffusion de la methode, ce qui justifierait une demande plus importante. Le comite CE53 est pertinent pour l'aspect ecosystemes aquatiques, et le caractere interdisciplinaire (SHS/ecologie) peut etre un atout si le consortium est renforce avec un laboratoire de statistique.

**Recommandation financement**: Positionnez ce projet comme une preuve de concept methodologique a fort potentiel de rupture dans le domaine de la restauration ecologique. Ciblez en priorite les programmes de financement exploratoires et interdisciplinaires (ex : Biodiversa, appel pilote du Belmont Forum) ou les bourses de demarrage (ex : ERC Starting Grant si le porteur est a moins de 7 ans du doctorat, mais avec un dossier renforce par des donnees preliminaires). Pour augmenter la fundabilite, integrez un partenaire en statistique bayesienne des l'ecriture du projet et proposez une etude de cas reelle (ex : un lac peri-alpin) comme 'demonstrateur' dans la phase 2.

## 6. Panel Review Summary

| Reviewer | Score | Verdict | Point cle |
|----------|-------|---------|-----------|
| methodologist | 7.8/10 | accept | The three-phase design is exemplary: it de-risks the hypothesis through in silico validation before committing to expensive fieldwork, and the explicit GO/NO-GO/PIVOT criteria at each phase provide a clear, pre-registered decision framework that prevents confirmation bias and allows for adaptive management of the research plan. |
| domain_expert | 7.8/10 | accept | The hypothesis is theoretically well-grounded in the framework of alternative stable states and regime shifts in lake ecosystems. The explicit use of paleo-derived pre-disturbance baselines as proxies for the 'clear-water state' and historical thresholds as empirical anchors for critical nutrient loading is a sophisticated and coherent application of resilience theory to management. This directly addresses the known issue of shifting baselines in contemporary monitoring. |
| contrarian | 3.5/10 | weak_reject | La tentative d'utiliser des donnees paleoecologiques comme contrainte bayesienne pour reduire l'incertitude des modeles de gestion est une approche conceptuellement interessante et potentiellement novatrice, qui merite d'etre exploree. |
| industrialist | 5.8/10 | weak_accept | Cible un besoin réglementaire payant : les agences de bassin (ex : Agences de l'Eau en France, RIVM aux Pays-Bas, US EPA) et les bureaux d'études (ex : Suez, Arcadis, WSP) doivent justifier des cibles de réduction de nutriments (DCE, Water Framework Directive) avec des marges d'incertitude réduites pour éviter des surcoûts de traitement (stations d'épuration) ou des amendes. |
| funding_strategist | 6.8/10 | weak_accept | Originalite conceptuelle : l'integration de donnees paleoecologiques comme priors bayesiens dans un modele mecanistique de qualite d'eau (PCLake) est une approche novatrice qui repond a un verrou methodologique identifie dans la gestion des plans d'eau (incertitude des cibles de reduction de nutriments). |

**Consensus score**: 6.4/10
**Verdict final**: publish_brief

### 6.1 Consensus

- The hypothesis is theoretically well-grounded and addresses a critical gap in water quality management: the integration of paleoecological data as Bayesian priors into process-based models (PCLake) to reduce uncertainty in nutrient reduction targets is a novel, sophisticated, and methodologically sound approach.
- The three-phase design with explicit GO/NO-GO/PIVOT criteria is a major strength, providing a clear, pre-registered decision framework that prevents confirmation bias and allows for adaptive management of the research plan.
- The primary risk to the hypothesis's validity is the potential for biased central estimates (being confidently wrong) rather than the reduction of variance (CV) itself. All reviewers agree that the protocol must explicitly test for shifts in the mean predicted target, not just the CV, to ensure the paleo priors are constraining the model correctly.
- The issue of model equifinality and parameter identifiability in PCLake is a significant concern that must be addressed with a formal sensitivity analysis or multi-model comparison to ensure the observed CV reduction is not an artifact of the model's structure.
- The non-stationarity of the TP-cyanobacteria relationship (due to climate change, species shifts) is a critical challenge. The proposed 'temperature-dependent threshold displacement' is a step in the right direction, but a more nuanced treatment is needed to avoid introducing bias from historical baselines that may not be relevant to future conditions.

### 6.2 Points de desaccord

- The Contrarian argues that the paleo priors on equilibrium states will not constrain the transient dynamics parameters that drive the response to nutrient reduction, making the hypothesized CV reduction a 'petition de principe'. In contrast, the Methodologist and Domain Expert believe the approach is a pragmatic and statistically robust method to address equifinality, provided the protocol includes a formal identifiability analysis.
- The Contrarian and Domain Expert both highlight the risk of a biased central estimate, but the Contrarian concludes this makes the hypothesis 'speculation', while the Domain Expert sees it as a refinement issue, recommending acceptance with a focus on strengthening the validation framework against biased priors.
- The Industrialist and Funding Strategist diverge on the commercial viability: the Industrialist sees a high barrier to entry and uncertain ROI, recommending integration into existing consulting offers, while the Funding Strategist sees strong potential for exploratory grants and emphasizes the need for a real-world demonstrator and a stronger consortium to increase fundability.
- There is disagreement on the likely effect size and power of the study. The Contrarian argues the assumed effect size (Cohen's d = 0.5-0.7) and inter-lake variance (15-25%) are overly optimistic, leading to a high risk of a false negative. The Methodologist and Domain Expert, while acknowledging this, do not see it as a fatal flaw but recommend a pre-specified contingency plan for larger variance.

### 6.3 Critical path

The most critical factor for the success of this hypothesis is demonstrating that the reduction in uncertainty (CV) is not achieved at the expense of a biased central estimate. The protocol must explicitly test for shifts in the mean predicted TP reduction target and include a formal analysis of model equifinality (e.g., profile likelihoods or a multi-model ensemble) to prove the paleo priors are constraining the model to the correct region of the parameter space, rather than artificially narrowing it around a wrong answer.

**Recommandation finale**: The panel recommends publishing this brief to proceed with the hypothesis. The consensus is that the hypothesis is theoretically sound, methodologically ambitious, and addresses a critical gap in lake management. The three-phase design with pre-registered decision gates is a major strength. However, the protocol must be refined to explicitly address the risk of biased central estimates by reporting and comparing mean predictions alongside CVs, and to include a formal plan for handling model equifinality (e.g., a global sensitivity analysis or identifiability assessment). The panel also emphasizes the need for a more detailed strategy for handling the spatial-temporal data mismatch between sediment cores and monitoring data, and a more nuanced treatment of non-stationarity beyond a simple linear temperature correction. While the Contrarian's concerns about the fundamental value of the paleo priors are noted, they are not considered fatal if the above refinements are implemented, as the phased approach allows for early termination if the in silico phase does not demonstrate the expected value.

## 7. Gap Manifest residuel

### 7.1 Data gaps

- The transferability of paleoecological proxies (e.g., sedaDNA, pigments) to current HAB dynamics is still uncertain, particularly regarding taphonomic biases and the representation of cyanobacterial community composition.
- The integration of paleoecological data with contemporary nutrient load models requires the development of specialized statistical techniques to handle dating uncertainties and different data resolutions.

### 7.2 Competence gaps

- Phase 1: Process-based lake modeling (PCLake)
- Phase 1: Bayesian statistics & MCMC methods
- Phase 1: R programming
- Phase 1: Paleolimnological data interpretation
- Phase 2: Paleolimnology (coring, dating, diatom analysis)
- Phase 2: Analytical chemistry (HPLC)
- Phase 2: Molecular biology (DNA extraction, PCR)
- Phase 2: PCLake modeling & Bayesian statistics
- Phase 3: Experienced paleolimnologist
- Phase 3: Limnologist/ecosystem modeler
- Phase 3: Bayesian statistician
- Phase 3: Molecular biologist (for sedaDNA)
- Phase 3: Science communicator

### 7.3 Epistemic gaps

- The exact magnitude of taphonomic bias in sedaDNA and pigment records for the target lakes.
- The degree to which internal phosphorus loading (legacy P) affects the TP-cyanobacteria relationship in each lake.
- The extent to which climate change has already shifted the TP threshold for cyanobacterial blooms in the target lakes.
- The optimal method for integrating paleo data into PCLake (e.g., priors vs. soft calibration targets) without introducing bias.

## References

[1] E. C. Nwosu, A. Brauer, et al. (2023). *Early human impact on lake cyanobacteria revealed by a Holocene record of sedimentary ancient DNA*. DOI: [10.1038/s42003-023-04430-z](https://doi.org/10.1038/s42003-023-04430-z)
[2] W. Hobbs, T. Dreher, et al. (2021). *Using a lake sediment record to infer the long-term history of cyanobacteria and the recent rise of an anatoxin producing Dolichospermum sp.*. DOI: [10.1016/j.hal.2020.101971](https://doi.org/10.1016/j.hal.2020.101971)
[3] M. Picard, S. Wood, et al. (2022). *Molecular and Pigment Analyses Provide Comparative Results When Reconstructing Historic Cyanobacterial Abundances from Lake Sediment Cores*. DOI: [10.3390/microorganisms10020279](https://doi.org/10.3390/microorganisms10020279)
[4] E. C. Nwosu, P. Roeser, et al. (2021). *From Water into Sediment—Tracing Freshwater Cyanobacteria via DNA Analyses*. DOI: [10.3390/microorganisms9081778](https://doi.org/10.3390/microorganisms9081778)
[5] Shuaidong Li, Jie Fang, et al. (2022). *Properties of sediment dissolved organic matter respond to eutrophication and interact with bacterial communities in a plateau lake.*. DOI: [10.1016/j.envpol.2022.118996](https://doi.org/10.1016/j.envpol.2022.118996)
[6] Marilén Fernández, J. F. Ponce, et al. (2020). *Paleolimnological response to climate variability during Late Glacial and Holocene times: A record from Lake Arturo, located in the Fuegian steppe, southern Argentina*. DOI: [10.1016/j.palaeo.2020.109737](https://doi.org/10.1016/j.palaeo.2020.109737)
[7] Chunhui Yang, Jiying Li, et al. (2021). *Phosphorus internal loading and sediment diagenesis in a large eutrophic lake (Lake Chaohu, China).*. DOI: [10.1016/j.envpol.2021.118471](https://doi.org/10.1016/j.envpol.2021.118471)
[8] E. Reavie, Meijun Cai (2019). *Consideration of species-specific diatom indicators of anthropogenic stress in the Great Lakes*. DOI: [10.1371/journal.pone.0210927](https://doi.org/10.1371/journal.pone.0210927)
[9] M. Stoddard, J. P. Roccaforte, et al. (2021). *Ecological restoration guided by historical reference conditions can increase resilience to climate change of southwestern U.S. Ponderosa pine forests*. DOI: [10.1016/J.FORECO.2021.119256](https://doi.org/10.1016/J.FORECO.2021.119256)
[10] Jian Peng, Hui Tang, et al. (2023). *Regarding reference state to identify priority areas for ecological restoration in a karst region.*. DOI: [10.1016/j.jenvman.2023.119214](https://doi.org/10.1016/j.jenvman.2023.119214)
[11] A. Sturbois, M. de Caceres, et al. (2023). *Ecological quality assessment: A framework to report ecosystems quality and their dynamics from reference conditions*. DOI: [10.1002/ecs2.4726](https://doi.org/10.1002/ecs2.4726)

## Annexes

### A. Detailed Reviewer Reports

#### Methodologist

- **Score**: 7.8/10 | **Verdict**: accept | **Confidence**: 0.9
- **Strengths**: The three-phase design is exemplary: it de-risks the hypothesis through in silico validation before committing to expensive fieldwork, and the explicit GO/NO-GO/PIVOT criteria at each phase provide a clear, pre-registered decision framework that prevents confirmation bias and allows for adaptive management of the research plan.; The use of a 'contemporary-augmented' (synthetic data) control arm is a methodologically sophisticated and powerful addition. It directly isolates the value of the *information content* of paleo data versus merely having a longer time series, which is a far more stringent and insightful test than a simple comparison against a contemporary-only baseline.; The explicit and quantitative treatment of proxy error (inflation factor 1.0-2.0) and the inclusion of a power analysis with a priori effect sizes (Cohen's d) demonstrate a strong grasp of the key sources of uncertainty and the practical constraints of detecting the hypothesized effect across a realistic number of lakes.
- **Weaknesses**: The protocol relies heavily on the assumption that the PCLake model structure is capable of capturing the long-term system states and resilience thresholds that paleo data are supposed to inform. The risk of 'equifinality' (multiple parameter sets yielding the same fit) is acknowledged, but the protocol lacks a specific plan for addressing it, such as a formal identifiability analysis or a comparison of model structures (e.g., using a simpler alternative model) to ensure the observed CV reduction is not an artifact of the chosen model's rigidity.; The primary endpoint, the CV of the TP reduction target, is a second-order statistic derived from a model output. The protocol does not explicitly address the potential for bias in this CV estimate due to the non-linearity of the model. A small error in the mean predicted target could lead to a large, spurious change in the CV, especially with a limited number of Monte Carlo simulations. A more robust primary endpoint might be the width of the posterior predictive interval for the target, or a direct comparison of the full posterior distributions.; The protocol for Phase 2 and 3 lacks a clear strategy for handling the inherent spatial and temporal mismatch between sediment core data (which integrates over a large area and time) and the point-based, instantaneous measurements used for contemporary monitoring. This 'commensurability' issue is a major source of potential bias in paleo-ecological model calibration and is not explicitly addressed in the data integration methodology.
- **Questions**: How will you formally assess and mitigate the risk of model equifinality in PCLake? Specifically, do you plan to use a multi-model ensemble or a parameter identifiability analysis (e.g., profile likelihoods) to ensure that the reduction in prediction uncertainty is a real consequence of the paleo priors and not simply a result of the priors constraining the model to a particular, but incorrect, region of the parameter space?; Given that the CV is a normalized measure of dispersion, how will you ensure that the predicted TP reduction target is not biased (e.g., systematically over- or under-estimated) across the three calibration strategies? A biased mean with a smaller CV could be misinterpreted as a success, when in fact the model is confidently wrong. Will you report and compare the mean predictions alongside the CVs?; The power analysis for Phase 3 is based on the inter-lake CV of the *CV reduction*. This is a meta-uncertainty that is rarely known a priori. What is your contingency plan if the observed inter-lake variance in the CV reduction is significantly larger than the 15-25% assumed in your power calculation, leading to a post-hoc power of less than 0.80? Will you pre-specify a 'fail-safe' sample size or a Bayesian adaptive design to address this?; How will you standardize the construction of the 'informative prior' from the paleo data across different lakes to ensure it is genuinely 'paleo-informed' and not inadvertently incorporating contemporary information? For example, if the diatom-inferred TP baseline is calibrated using a transfer function that was itself developed with data from lakes that include some of the same contemporary monitoring data, this could lead to circularity and an overly optimistic assessment of the prior's value.
- **Recommendation**: This is a well-designed, ambitious, and methodologically sound proposal that addresses a critical gap in water quality management. The phased approach with clear decision gates is a major strength. My recommendation is to accept, but with a strong request to address the issue of model equifinality and to explicitly report and compare the mean predictions alongside the CVs to avoid the pitfall of confidently wrong estimates. The protocol should also include a more detailed plan for handling the spatial-temporal data mismatch between sediment cores and monitoring data.

#### Domain Expert

- **Score**: 7.8/10 | **Verdict**: accept | **Confidence**: 0.88
- **Strengths**: The hypothesis is theoretically well-grounded in the framework of alternative stable states and regime shifts in lake ecosystems. The explicit use of paleo-derived pre-disturbance baselines as proxies for the 'clear-water state' and historical thresholds as empirical anchors for critical nutrient loading is a sophisticated and coherent application of resilience theory to management. This directly addresses the known issue of shifting baselines in contemporary monitoring.; The proposed mechanism is physically and chemically plausible. The causal chain is logical: sediment cores provide a time-integrated record of past nutrient conditions (via diatom transfer functions) and biological response (via pigments/sedaDNA), which can be used to constrain the parameter space of a process-based model like PCLake. This approach is a pragmatic and statistically robust method to address equifinality and reduce parametric uncertainty, which is a major limitation of calibrating complex models with short-term data alone.; The hypothesis is well-positioned relative to the state of the art. It builds directly upon recent advances in sedimentary DNA (sedaDNA) and pigment analyses that have demonstrated the ability to reconstruct past cyanobacterial dynamics (e.g., Nwosu et al., 2023; Hobbs et al., 2021). It also correctly identifies the critical challenge of legacy phosphorus and non-stationarity due to climate change, proposing a concrete methodological extension (temperature-dependent threshold displacement) to address these issues, which demonstrates a deep understanding of current research frontiers.
- **Weaknesses**: The hypothesis relies on a strong and potentially problematic assumption of stationarity in the TP-cyanobacteria relationship over the calibration period. While the authors acknowledge this, the proposed 'temperature-dependent TP threshold displacement term' is a simplistic linear fix for a complex, non-linear interaction involving multiple factors (e.g., light, grazing, CO2, species composition). This could introduce significant model structural error and bias, potentially undermining the claimed reduction in uncertainty.; The formulation is overly focused on the statistical reduction of uncertainty (CV) as the primary outcome. The more critical and scientifically challenging question is whether the paleo-informed priors lead to a *different* and *more accurate* central estimate of the TP reduction target. A model can have a lower CV but be confidently wrong if the priors are biased (e.g., due to taphonomic biases in the proxies or a non-representative pre-disturbance baseline period). The hypothesis should explicitly test for a shift in the mean target, not just the variance.; The proposed mechanism implicitly assumes that the pre-disturbance baseline (1800-1850) is a relevant management target for all lakes. However, many lakes may have naturally high background nutrient levels or have been impacted by non-nutrient stressors (e.g., hydrological alteration, species invasions) that make a return to this state impossible or undesirable. The hypothesis does not adequately address how this potential mismatch between the paleo-baseline and the desired ecological state is reconciled in the model's objective function.
- **Questions**: How will the authors distinguish between a reduction in uncertainty (CV) that comes from correctly constraining the model's parameter space versus a reduction that comes from artificially narrowing the parameter distribution around a biased central estimate, given that the true TP reduction target is unknown?; What is the proposed method for quantifying and propagating the substantial uncertainty associated with the paleo-proxies themselves (e.g., RMSEP of diatom transfer functions, taphonomic bias in sedaDNA) into the Bayesian priors, and how will this be separated from the model's parametric uncertainty?; Given the known issue of non-stationarity, how will the model's temperature-dependent TP threshold displacement term be parameterized and validated, especially for lakes where the historical record does not contain a period of warming comparable to future projections?
- **Recommendation**: This is a strong and highly relevant hypothesis that bridges paleoecology and ecosystem management in a novel way. The theoretical framing is excellent and the proposed methodology is a logical next step in the field. However, the hypothesis should be refined to explicitly address the potential for bias in the central estimate, not just the variance, and to provide a more nuanced treatment of non-stationarity. I recommend accepting this hypothesis for further development, with a focus on strengthening the validation framework against biased priors and expanding the analysis to explicitly test for changes in the mean predicted target, not just its coefficient of variation.

#### Contrarian

- **Score**: 3.5/10 | **Verdict**: weak_reject | **Confidence**: 0.85
- **Strengths**: La tentative d'utiliser des donnees paleoecologiques comme contrainte bayesienne pour reduire l'incertitude des modeles de gestion est une approche conceptuellement interessante et potentiellement novatrice, qui merite d'etre exploree.; La formulation de predictions quantitatives et testables (reduction de CV de 20%, puissance statistique) est une qualite rare et louable, montrant une volonte de rigueur methodologique.
- **Weaknesses**: FAIL REASON #1: L'equifinalite et le probleme d'identifiabilite des parametres dans PCLake sont ignores. Le modele est complexe et de nombreux jeux de parametres peuvent produire des sorties similaires. L'ajout de priors paleo, qui contraignent des etats d'equilibre (baseline pre-impact), ne reduit pas necessairement l'incertitude sur les parametres de dynamique transitoire (coefficient de recyclage interne) qui pilotent la reponse a une reduction de charge. Le CV de la cible de reduction de TP pourrait etre domine par cette incertitude dynamique, que les priors paleo ne contraignent pas. L'hypothese que les priors sur les seuils historiques reduiront l'incertitude sur la reponse future est une hypothese forte et non demontree.; FAIL REASON #2: Le probleme de non-stationnarite et de changement de regime est un angle mort critique. L'hypothese postule une relation TP-cyanobacteries stationnaire, a l'exception d'un terme de deplacement thermique. Or, les lacs eutrophes subissent souvent des changements de regime (basculement vers un etat turbide) avec des hysteresis et des seuils de resilience qui ne sont pas captures par une simple relation lineaire ou log-lineaire. Les donnees paleo de la periode 1800-1850 representent un etat de reference pre-industriel, mais l'application de ce seuil historique a un lac de 2024, avec un climat different, une composition d'especes differente et des pressions de selection differentes, est une extrapolation incertaine. Le modele PCLake, meme avec un terme thermique, peut ne pas capturer ces dynamiques non-lineaires, rendant la reduction d'incertitude annoncee illusoire.; FAIL REASON #3: L'effet de taille probable est surestime et le scenario de detection est fragile. L'erreur des fonctions de transfert (RMSEP ±0.2-0.4 log10 µg/L) est importante. Si on l'inflate par 2, l'incertitude sur le prior paleo devient enorme. L'information apportee par le prior est alors trop faible pour contraindre significativement les parametres de PCLake, qui a deja ses propres incertitudes structurelles. De plus, la prediction 4 (10-15 lacs suffisent) est basee sur un effet de taille (Cohen's d = 0.5-0.7) qui est optimiste. La variance inter-lacs de la reduction de CV est probablement bien superieure a 15-25%, car elle depend de la qualite des carottes, de l'historique de chaque lac, et de la structure du modele. Le test de puissance est donc probablement sous-estime, et l'etude risque de ne pas detecter d'effet significatif, menant a un faux negatif ou a une conclusion erronee sur l'absence d'utilite des donnees paleo.
- **Questions**: Comment demontrez-vous que l'incertitude sur la cible de reduction de TP est dominee par l'incertitude sur les parametres d'etat d'equilibre (que les priors paleo contraignent) et non par l'incertitude sur les parametres de dynamique transitoire (qui ne sont pas contraints par des donnees paleo d'etat stable) ? Sans une analyse de sensibilite globale (variance-based) de PCLake sur ce point, votre reduction de CV annoncee est une peticion de principe.; Comment votre modele gere-t-il le probleme de l'hysteresis et de la non-stationnarite des seuils de resilience ? Un seuil historique de bloom (ex: 50 µg/L de TP en 1850) est-il encore pertinent pour un lac en 2024 avec un climat plus chaud et une communaute biologique differente ? Si le seuil a change, votre prior est biaise et la reduction d'incertitude est obtenue au prix d'un biais accru, ce qui est pire que l'incertitude initiale.
- **Recommendation**: Pour que cette hypothese soit credible, il faut d'abord demontrer sur des donnees synthetiques (ou des lacs experimentaux) que l'ajout de priors sur les etats d'equilibre reduit specifiquement l'incertitude sur les parametres de dynamique transitoire. Ensuite, il faut fournir une analyse de sensibilite globale de PCLake pour identifier les parametres dominants de l'incertitude de la cible de reduction. Enfin, il faut simuler explicitement des scenarios de changement de regime et de non-stationnarite pour montrer que le prior paleo n'introduit pas de biais. Sans ces demonstrations, l'hypothese reste une speculation interessante mais non fondee.

#### Industrialist

- **Score**: 5.8/10 | **Verdict**: weak_accept | **Confidence**: 0.65
- **Strengths**: Cible un besoin réglementaire payant : les agences de bassin (ex : Agences de l'Eau en France, RIVM aux Pays-Bas, US EPA) et les bureaux d'études (ex : Suez, Arcadis, WSP) doivent justifier des cibles de réduction de nutriments (DCE, Water Framework Directive) avec des marges d'incertitude réduites pour éviter des surcoûts de traitement (stations d'épuration) ou des amendes.; L'approche 'paleo-informed' crée un avantage compétitif fort en transformant des données sédimentaires (souvent perçues comme académiques) en un actif de modélisation prédictif, permettant de breveter une méthode de calibration (Bayesian priors + PCLake) difficile à copier sans les jeux de données paléo historiques propriétaires.
- **Weaknesses**: Barrière à l'entrée majeure : le marché est fragmenté et dominé par des acteurs établis (consultants en modélisation) qui utilisent déjà des approches standards (MOHID, Delft3D). Les gestionnaires de l'eau sont risk-averse et exigent des validations longues (5-10 ans) avant d'adopter une nouvelle méthode, surtout si elle repose sur des carottes sédimentaires (coûts de terrain et d'analyse non négligeables, ~50-100k€ par lac).; Le ROI est incertain : le gain de 20% de réduction de CV sur une cible de TP ne se traduit pas directement en euros pour le client final. Le seul acheteur potentiel est le secteur public (agences de l'eau), qui a des budgets contraints et peut considérer cette approche comme une 'niche académique' plutôt qu'une nécessité opérationnelle, surtout si les modèles existants sont 'suffisamment bons' pour la réglementation.
- **Questions**: Quel est le coût complet d'une étude 'paleo-informed' par lac (carottage, datation, analyse diatomées/pigments, intégration PCLake) et comment se compare-t-il au budget typique d'une étude de modélisation DCE (souvent < 50k€ par lac) ? Peut-on réduire ce coût à moins de 20k€ par lac pour être compétitif ?; Quelle est la propriété intellectuelle réelle ? Un brevet sur la méthode de calibration (priors bayésiens) est-il défendable, ou les données paléo (carottes) sont-elles la seule barrière à l'entrée, ce qui rendrait le modèle économique dépendant de la possession de banques de sédiments (ex : dépôts universitaires) ?; Quel est le cycle de vente réel ? Les agences de bassin ont-elles un mandat explicite pour réduire l'incertitude des cibles de nutriments (ex : DCE 2027) ou est-ce un 'nice-to-have' ? Un pilote réussi dans 2-3 lacs suffira-t-il à convaincre un comité de bassin de changer sa méthodologie standard ?
- **Recommendation**: Ne pas commercialiser en tant que produit autonome. Intégrer cette méthode comme un 'module de calibration avancée' dans une offre de conseil existante (ex : mise à jour des SDAGE ou des plans de gestion de bassin). Cibler les agences de l'eau en Europe du Nord (Pays-Bas, Danemark) où la pression sur les blooms cyanobactériens est forte et les budgets R&D existent. Déposer un brevet sur le 'système de génération de priors paléo pour modèles de qualité d'eau' et créer une base de données propriétaire de carottes sédimentaires numérisées pour verrouiller les clients. Le ROI est long (5-7 ans) et dépend de la capacité à standardiser le protocole (Phase 3) pour le rendre aussi simple qu'une analyse de routine.

#### Funding Strategist

- **Score**: 6.8/10 | **Verdict**: weak_accept | **Confidence**: 0.72
- **Strengths**: Originalite conceptuelle : l'integration de donnees paleoecologiques comme priors bayesiens dans un modele mecanistique de qualite d'eau (PCLake) est une approche novatrice qui repond a un verrou methodologique identifie dans la gestion des plans d'eau (incertitude des cibles de reduction de nutriments).; Protocole de validation rigoureux et falsifiable : la structure en 3 phases avec des criteres GO/NO-GO quantitatifs (CV reduction ≥ 20%) demontre une maturite scientifique et une honnetete intellectuelle qui rassurent les evaluateurs sur la faisabilite et la portee du projet.; Taille et duree adaptees aux appels a projets exploratoires : le budget de 50-150k€ et le calendrier de 12-18 mois correspondent parfaitement aux enveloppes des programmes de science ouverte ou de semences (pilot grants), permettant une prise de risque acceptable pour les bailleurs.
- **Weaknesses**: TRL initial faible (TRL 2-3) : le projet est essentiellement une preuve de concept methodologique. L'absence de donnees preliminaires solides (ex : une etude de cas reelle montrant une reduction d'incertitude) rend le dossier moins competitif pour les appels a projets plus importants (ERC, Horizon Europe Collaboration) qui exigent un TRL plus eleve ou des resultats pilotes.; Consortium limite et mono-disciplinaire : l'hypothese est portee par une approche couplee (paleoecologie + modelisation). Un consortium ideal devrait inclure un partenaire en ecologie theorique/statistique bayesienne et un gestionnaire de bassin versant (autorite de l'eau) pour garantir l'operationalite, ce qui n'est pas mentionne dans le protocole.; Dependance a la qualite des proxies et des donnees historiques : le risque scientifique est majore par la disponibilite et la resolution des carottes sedimentaires (fiabilite de la datation, preservation des pigments) et la qualite des series temporelles de monitoring contemporain, ce qui n'est pas suffisamment adresse dans la phase 1 (in silico).
- **Questions**: Avez-vous identifie un jeu de donnees existant (ex : base de donnees de l'EPA, reseau de lacs suisses ou suedois) pour valider la phase 1 in silico, ou comptez-vous sur des donnees synthetiques ? La disponibilite de donnees reelles est un critere de faisabilite majeur pour les evaluateurs.; Quel est le plan de gestion des donnees et des incertitudes pour les proxies paleo (ex : erreur de datation, variance intra-lac) ? Comment ces incertitudes seront-elles propagees dans le modele bayesien pour ne pas biaiser les priors ?; Quel est le partenaire ou l'utilisateur final (end-user) implique dans le projet (ex : agence de l'eau, ONG) pour garantir que les resultats sur les cibles de reduction de nutriments seront effectivement utilises dans un processus de decision publique ?
- **Recommendation**: Positionnez ce projet comme une preuve de concept methodologique a fort potentiel de rupture dans le domaine de la restauration ecologique. Ciblez en priorite les programmes de financement exploratoires et interdisciplinaires (ex : Biodiversa, appel pilote du Belmont Forum) ou les bourses de demarrage (ex : ERC Starting Grant si le porteur est a moins de 7 ans du doctorat, mais avec un dossier renforce par des donnees preliminaires). Pour augmenter la fundabilite, integrez un partenaire en statistique bayesienne des l'ecriture du projet et proposez une etude de cas reelle (ex : un lac peri-alpin) comme 'demonstrateur' dans la phase 2.

### B. Semantic Scholar Search Queries

- [novelty] `paleolimnology eutrophication cyanobacteria sediment cores` — Direct search for paleolimnological reconstructions of cyanobacterial blooms and eutrophication history
- [novelty] `diatom transfer functions historical nutrient baselines lakes` — Targets the specific methodological bridge between diatom-based nutrient reconstructions and baseline identification
- [novelty] `paleoecology harmful algal bloom management restoration` — Searches for existing applications of paleoecological data to HAB management strategies
- [novelty] `sedimentary ancient DNA cyanobacteria bloom reconstruction` — Tests whether sedaDNA-based cyanobacteria reconstructions have been explicitly linked to management thresholds
- [evidence] `lake sediment pigments historical eutrophication trajectory` — Seeks empirical evidence of fossil pigment records tracking past bloom intensity and nutrient shifts
- [evidence] `charcoal pollen land use lake nutrient loading` — Targets causal evidence linking catchment vegetation/fire history to aquatic nutrient loading changes
- [evidence] `legacy phosphorus internal loading paleolimnology` — Finds evidence for historical phosphorus accumulation and its role in current bloom dynamics
- [evidence] `climate variability Holocene lake productivity thresholds` — Seeks evidence of past climate-driven ecosystem state shifts and resilience thresholds in lakes
- [cross_domain] `paleoecology water quality management decision support` — Finds interdisciplinary precedents where paleo data inform contemporary water management decisions
- [cross_domain] `reference conditions ecological restoration lakes estuaries` — Explores the broader restoration ecology literature on using historical baselines for target setting
- [cross_domain] `paleolimnology nutrient criteria establishment` — Targets regulatory science linking sediment reconstructions to numeric nutrient criteria development

---

*Generated by SPORE (Systeme de Production d'Opportunites de Recherche par Exploration) on 2026-08-23*