Forecasting Land Art Under Climate Scenarios
Forecasts Spiral Jetty's state in 2030/2050 using complexity-signature regression and climate-conditioned diffusion.
Key Findings
Methodology
The paper proposes a two-stage forecasting pipeline. Stage 1 uses IPCC AR6 SSP scenarios to predict regional temperature, lake elevation, and cumulative CO2. Stage 2 forecasts image complexity features using linear and Random Forest regressions. It also employs Stable Diffusion XL for climate-conditioned image generation, incorporating LoRA fine-tuning and ControlNet conditioning.
Key Results
- In all scenarios, Spiral Jetty is classified as dry playa in 2030 and 2050, with lake elevation below 4,193 ft.
- The Random Forest model saturates at the post-2015 dry-playa regime, while the linear model diverges under extreme conditions.
- Generated images exhibit visual complexity similar to the post-2015 state.
Significance
This study demonstrates how generative models can predict the future state of cultural heritage under climate change, filling a gap in research on climate impacts on cultural heritage. The method can be applied to other land art predictions, offering significant academic and practical value.
Technical Contribution
The paper uniquely combines complexity-signature regression with climate-conditioned diffusion models for cultural heritage forecasting, providing new theoretical frameworks and engineering possibilities. Personalization and conditioning of generative models are achieved through LoRA and ControlNet.
Novelty
This is the first application of climate-conditioned diffusion models to cultural heritage forecasting, combining complexity features and climate data for a unique predictive perspective.
Limitations
- The model's predictions are less accurate when extrapolating beyond historical data, especially under extreme climate conditions.
- Other factors affecting lake elevation, such as upstream water use, are not considered.
Future Work
Future work could apply this method to other land art and incorporate multi-factor hydrological models to improve prediction accuracy.
AI Executive Summary
Predicting the future state of cultural heritage under climate change is increasingly important. This paper uses Robert Smithson's Spiral Jetty as a case study to propose an innovative two-stage forecasting method. Stage 1 uses IPCC AR6 SSP scenarios to predict regional climate variables such as temperature and lake elevation. Stage 2 forecasts image visual complexity through complexity-signature regression and climate-conditioned diffusion models.
Experimental results show that under all climate scenarios, Spiral Jetty will be in a dry playa state in 2030 and 2050, with lake elevation below 4,193 feet. This finding suggests that even under the most optimistic climate scenarios, the artwork will remain exposed. Using Stable Diffusion XL and LoRA fine-tuning, the generated images exhibit visual complexity similar to the post-2015 state.
This study not only fills a gap in research on climate impacts on cultural heritage but also provides a new method for predicting other land art. Future work can extend to other cultural heritage sites and incorporate more complex hydrological models to improve prediction accuracy.
Deep Analysis
Background
With the intensification of climate change, predicting the future state of cultural heritage becomes crucial. Robert Smithson's Spiral Jetty, located in Utah's Great Salt Lake, serves as a fixed remote-sensing target whose visual complexity reflects hydroclimatic conditions. Previous studies have established relationships between image complexity and lake elevation, regional temperature, and cumulative CO2, providing a foundation for prediction.
Core Problem
The core problem is predicting the future state of cultural heritage under climate change scenarios. Existing methods lack the ability to predict cultural heritage under extreme climate conditions, particularly when extrapolating beyond historical data.
Innovation
This paper innovatively combines complexity-signature regression with climate-conditioned diffusion models for cultural heritage forecasting. By using Stable Diffusion XL and LoRA fine-tuning, it achieves personalization and conditioning of generative models, offering a new predictive perspective.
Methodology
- �� Stage 1: Use IPCC AR6 SSP scenarios to predict regional climate variables.
- �� Stage 2: Forecast image complexity features using linear and Random Forest regressions.
- �� Employ Stable Diffusion XL for climate-conditioned image generation, incorporating LoRA fine-tuning and ControlNet conditioning.
Experiments
The experimental design includes using Landsat 4-9 and Sentinel-2 datasets, training on a 42-year record. Linear and Random Forest regressions are used for complexity feature prediction, and Stable Diffusion XL is used for image generation.
Results
Experimental results show that under all climate scenarios, Spiral Jetty will be in a dry playa state in 2030 and 2050, with lake elevation below 4,193 feet. The Random Forest model saturates at the post-2015 dry-playa regime, while the linear model diverges under extreme conditions.
Applications
This method can be applied to other land art predictions, such as Sun Tunnels and Lightning Field, helping assess climate change impacts on cultural heritage.
Limitations & Outlook
The model's predictions are less accurate when extrapolating beyond historical data, especially under extreme climate conditions. Other factors affecting lake elevation, such as upstream water use, are not considered.
Plain Language Accessible to non-experts
Imagine building a sandcastle on the beach. As the tides change, the shape of the sandcastle changes. This method is like a high-tech crystal ball that predicts future tide changes and the sandcastle's shape. By analyzing past decades of tide data and sandcastle changes, we can predict how future tides will affect the sandcastle. It's like using a complex computer program to simulate future weather and tides, helping us understand what the sandcastle will look like in the future.
ELI14 Explained like you're 14
Imagine you're playing a game about the future, and you need to predict what an ancient artwork will look like. Scientists used a high-tech tool called a 'diffusion model,' like a super-smart crystal ball, to predict future weather and lake changes. They trained this model using decades of past data, like leveling up your character in a game. The results show that this artwork will stay exposed in the future, like unlocking a hidden level in the game!
Glossary
Diffusion Model
A machine learning model used for generating images by progressively denoising to produce high-quality images.
Used for generating future images of Spiral Jetty.
Random Forest
An ensemble learning method that makes predictions by combining multiple decision trees.
Used for predicting image complexity features.
LoRA Fine-tuning
A lightweight model fine-tuning method that inserts low-rank matrices to achieve personalization.
Used to personalize the Stable Diffusion XL model.
ControlNet
A network structure for conditional generation by controlling the generation process with additional inputs.
Used to input climate conditions into the generative model.
Complexity Feature
Statistical features describing image complexity, such as entropy and mean intensity.
Used to predict the visual state of Spiral Jetty.
Open Questions Unanswered questions from this research
- 1 How to improve model prediction accuracy under extreme climate conditions? Current methods perform poorly when extrapolating beyond historical data, requiring new algorithms to address this issue.
- 2 How to incorporate other influencing factors into the prediction model, such as upstream water use and precipitation changes?
Applications
Immediate Applications
Cultural Heritage Preservation
Helps cultural heritage managers develop preservation strategies by predicting future states. Requires high-precision climate and image data.
Long-term Vision
Climate Change Impact Assessment
Provides scientific basis for policymakers on climate change impacts on cultural heritage, promoting broader environmental protection measures.
Abstract
Robert Smithson's 1970 land artwork Spiral Jetty, located in the north arm of Utah's Great Salt Lake, provides a fixed remote-sensing target whose visual complexity reflects hydroclimatic conditions. A companion study analyzed 1,744 co-registered Landsat 4-9 and Sentinel-2 image chips spanning every year and month from 1984 to 2025. It found robust relationships between coarse-scale permutation entropy, mean intensity, and the third principal component of ResNet50 avg-pool embeddings, and lake elevation, regional temperature, and cumulative CO2. It also showed that image complexity leads lake stage by about three years and that the long-term trend is non-monotonic, with decline after 2015. Building on those findings, this paper develops a two-stage forecasting pipeline. Stage 1 applies IPCC AR6 SSP1-2.6, SSP2-4.5, and SSP5-8.5 temperature changes to forecast regional temperature, north- and south-arm lake elevation, salinity, and cumulative CO2 for 2030 and 2050. Stage 2a projects 14 image-complexity features using linear and Random Forest regressions trained on the 42-year record. All six scenario-year combinations move the inputs outside the 1984-2025 training distribution. The Random Forest model therefore saturates near the post-2015 dry-playa regime, while a threshold-based hydrological interpretation classifies Spiral Jetty as fully exposed in every scenario. Stage 2b specifies a climate-conditioned latent-diffusion framework using Stable Diffusion XL fine-tuned with LoRA on the 1,744 chips, ControlNet conditioning on the climate vector, and a hydrological physics mask for ensemble image synthesis. Code is provided, and validation results will be released after diffusion training. We conclude by discussing ethical implications of generative speculation on cultural heritage and outlining a roadmap for forecasting at Sun Tunnels, Double Negative, Lightning Field, and Roden Crater.