Decomposing DIAs. Analysis of Environmental Impact Statements for solar installations

XII National Congress for Environmental Impact Assessment CONEIA 2024

Keywords

#Agroambientales #CompensaciónSuperficie #InteligenciaArtificial #Mortalidad #Resolución #SeguimientoAmbiental #SolarFotovoltaica

introduction

The collaboration between Ideas MedioAmbiental and Taidy (https://taidy.cloud/es/) has resulted in an advanced procedure for the automated analysis of Environmental Impact Statements (DIAs) using Artificial Intelligence (AI) technology. This synergy between the environmental impact assessment experts at Ideas MedioAmbiental and the AI specialists at Taidy has been fundamental to its development. The hard manual labeling work carried out by Ideas experts stands out, which has enriched Taidy's AI models.

The final system developed has proven to be highly effective in automatically evaluating DIAs, providing accurate and valuable analyses that support environmental research and informed decision-making. In addition, this system is presented as a complementary tool for professionals in the field, optimizing their work without replacing human judgment. This achievement underscores the ability of AI-powered technology to generate exceptional results and offer new perspectives in protecting the environment.

Una Environmental Impact Statement It is a mandatory and decisive report issued by the competent environmental body with which the environmental impact assessment ordinary, which evaluates the integration of environmental aspects into the project and determines the conditions that must be established for the adequate protection of the environment and natural resources during the execution and exploitation and, where appropriate, the dismantling or demolition of the project (MITERD, 2024).

Therefore, this document becomes the main road map for the establishment of preventive, corrective and compensatory measures that will be established during the different phases of the project, such as:

• Phase prior to the start of the works
• Construction or construction phase
• Operating or operating phase
• Dismantling phase

The analysis of information from Environmental Impact Statements it is a complex process, due to the heterogeneity of structure, content and case studies of projects. However, it is considered a crucial step for the proper establishment, maintenance and monitoring of preventive, corrective and compensatory measures as directed by the administration.

The main target The purpose of this study is the location, analysis and interpretation of the information collected. For this reason, for this study, the documents published for the autonomous communities with the highest development of photovoltaic projects and those issued by the Ministry for the Ecological Transition and the Demographic Challenge, in the period between 2020 and 2023, were selected.

• Official Gazette of Castilla-La Mancha (DOCM), issued by the Government of Castilla-La Mancha.
• Official Gazette of Castilla y León (BOCYL), issued by the Junta de Castilla y León.
• Official Gazette of the Junta de Andalucía (BOJA), issued by the Junta de Andalucía.
• Official Gazette of the Community of Madrid (BOCM), issued by the Community of Madrid.
• Official Gazette of Aragon (BOA), issued by the Government of Aragon.
• Extremadura Official Document (DOE), issued by the Extremadura Government.
• Official State Gazette (BOE), issued by the Ministry for the Ecological Transition and the Demographic Challenge (MITERD).

Materials and methods

This study differs from other analyses carried out for this documentation by two fundamental factors. On the one hand, the intensive manual labeling work carried out by expert technical personnel to transfer specialized knowledge to AI models, and on the other, the refinement of AI models (Prompting) of advanced work, interaction and adjustment for results. This synergy between processes is essential to adapt models to the heterogeneity, complexity and specific detail of this type of document.

The analyzed parameters have been extracted using Generative Artificial Intelligence (IAG) guided and supervised by experienced technical personnel, establishing the following steps:

• Programmatic document access and collection

The first phase involves programmatic access to the official web portals of the main Autonomous Communities with the development of this type of project, as well as to the Official State Gazette (BOE). In this way, the documents relevant to Environmental Impact Statements are collected.

• Data filtering

Once the documents have been collected, an initial filtering is carried out. Given the heterogeneity in the presentation of information on the different web portals, a specific filtering and selection process was implemented for each source, ensuring precision in the identification and collection of relevant documents, especially those that correspond to photovoltaic plant project resolutions and other specific criteria selected.

• Extraction and analysis, applying advanced AI processes, including the use of Large Language Models (LLMs).

The third step is aimed at extracting the necessary data for the present study, as well as others that are currently being worked on. The values have been obtained for both favorable, unfavorable or conditional Environmental Impact Statements.

• Qualitative validation and measurement

Considering the probabilistic nature of AI models and the potential performance risks, an exhaustive evaluation of the results obtained is carried out. This stage includes the review of performance and accuracy metrics, verified with data previously processed, labeled and analyzed manually by technical personnel experienced in reviewing this type of documentation. This step is essential to confirm the effectiveness and reliability of the artificial intelligence models used.

• Statistical analysis for the parameters obtained

The system developed after this analytical process has proven to be high-performance, reliable and effective in the automatic evaluation of Environmental Impact Statements. The results achieved and validated by rigorous comparisons with reference data, they emphasize the capacity of this AI-driven technology to generate accurate and valuable analyses, offering new perspectives for environmental research, optimization of efforts and serving as decision-making support for personnel who are experts in Environmental Impact Assessment.

Results

The results obtained show significant differences between documents issued by different administrations. This study focuses on the analysis of the following points, however, a larger work is under development where the variations for all the parameters obtained will be detailed.

• Percentage of authorized projects and unfavorable.
• Request for a monitoring plan and environmental monitoring.
• Request for mortality monitoring for project infrastructures.
• Principals Target species of conservation.
• Measures compensatory steppe habitat.

The results have been obtained after the analysis of approximately 1,500 Environmental Impact Statements issued between 2020 and 2023 for the selected newsletters. The average values show a percentage of favorable resolutions that reach 92.2%, demonstrating the improvement in the Environmental Impact Assessment processes carried out previously, as well as the modifications that occurred in the implementations before reaching the administrative authorization process after the detection of critical or limiting environmental parameters.

Among all the bulletins analyzed, the values obtained for the Official State Gazette (BOE) for procedures carried out through the Ministry for the Ecological Transition and the Demographic Challenge (MITERD) where the average percentages between the years 2020 and 2023 correspond to 82.4% of positive resolutions compared to 16.4% of negative resolutions. In addition, analyzing the data by year of publication, we can see that the number of resolutions issued shows an increasing trend, going from values close to 10 during the years 2020-2021 to 100 documents issued in 2022 and more than 220 in the year 2023.

The results are similar for the presence of environmental monitoring with values generally greater than 80% of the total number of authorized projects. However, there are large differences for monitoring mortality, where several communities show negative trends and percentages of less than 20-40% of presence in authorizations. These data contrast with the growing trend observed in the results for the Official State Gazette, with requests exceeding 82% by 2023.

The most commonly mentioned species have a protection category, with large raptors and steppes standing out. The 20 most cited species account for 85% of total citations annually. In addition, the 37% is obtained from 7 steppe birds, the kestrel being primilla (Falco Naumanni), the ashy harrier (Circus Pygargus), the royal kite (Milvus Milvus) and the common tern (Tetrax Tetrax) the species with the highest number of mentions. This result is directly related to the need to establish steppe habitat compensation. However, this compensation is a measure with a high disparity of criteria between projects and administrations, both in terms of compensation for occupancy of the plant and in total hectares.

Conclusions

The detailed analysis of the Environmental Impact Statements provides an overview of the evolution for the established preventive, corrective and compensatory measures, allowing us to analyze the variability of criteria and critical points for each Administration. Knowledge of these parameters improves the establishment, maintenance and monitoring of measures, as well as allows the advancement and equalization of criteria between the different regions of our territory.

Authors

Arenas Martinez, Gaspar

Martinez Iniesta, Cristobal

Roldan Arroyo, Juan Manuel

Gonzalez Medina, Daniel

Sanchez Lopez, Adrian