Ahmed Derdouri
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Ahmed Derdouri has a total of 31 co-authors. Here are the top co-authors by number of shared publications:
- Yuji Murayama
Institute of Life and Environmental Sciences, University of Tsukuba, 1-1-1 Tennodai, Tsukuba 305-8572, Japan
8 shared publications - Ruci Wang
Center for Environmental Remote Sensing, Chiba University, Chiba 263-8522, Japan
5 shared publications - Takehiro Morimoto
Institute of Life and Environmental Sciences, University of Tsukuba, 1-1-1 Tennodai, Tsukuba 305-8572, Japan
4 shared publications - Toshihiro Osaragi
Institute of Science Tokyo
4 shared publications Toshihiro Osaragi
3 shared publicationsAnsar Khan
2 shared publications
Latest Publications
- Open Access
- 4 May 2026
- Published by Springer Nature in Research Square
LLM-assisted screening and extraction for systematic reviews: workload reduction and refusal-driven missingness in a land-use evidence synthesis
Generative AI may reduce the cost of systematic reviews, but deployed models can fail in ways that affect which evidence is included in a review. We evaluated a hosted large language model (LLM), Qwen Max, as an assistant for an environmental evidence synthesis on national-scale land-use/land-cover prediction. After retrieving 11,817 Web of Science records and screening 11,688 unique records, we compared LLM title–abstract screening, full-text screening, and structured extraction with an expert-curated baseline. The workflow processed the corpus in about 25 hours on a single workstation, incurring US$106 in API charges. Screening was useful as a high-throughput triage step, but disagreements concentrated in borderline records and require human adjudication. For 345 full-text extraction attempts, agreement was strongest for categorical, temporal, and set-based fields and weaker for narrative fields. Full-text calls also produced explicit refusals: 7/2,084 deep-screening papers and 2/345 extraction papers were blocked as sensitive geopolitical content. These refusals clustered in contested regions, creating non-random missingness rather than ordinary extraction error. LLMs can therefore support systematic review workflows when used as auditable decision aids, but evidence synthesis protocols should pre-specify prompts, log raw outputs and refusals, report workload and validation metrics, and route refusals and high-risk fields to human review.
- 4 May 2026
- Published by Springer Nature in Research Square
Open AccessLLM-assisted screening and extraction for systematic reviews: workload reduction and refusal-driven missingness in a land-use evidence synthesis
Generative AI may reduce the cost of systematic reviews, but deployed models can fail in ways that affect which evidence is included in a review. We evaluated a hosted large language model (LLM), Qwen Max, as an assistant for an environmental evidence synthesis on national-scale land-use/land-cover prediction. After retrieving 11,817 Web of Science records and screening 11,688 unique records, we compared LLM title–abstract screening, full-text screening, and structured extraction with an expert-curated baseline. The workflow processed the corpus in about 25 hours on a single workstation, incurring US$106 in API charges. Screening was useful as a high-throughput triage step, but disagreements concentrated in borderline records and require human adjudication. For 345 full-text extraction attempts, agreement was strongest for categorical, temporal, and set-based fields and weaker for narrative fields. Full-text calls also produced explicit refusals: 7/2,084 deep-screening papers and 2/345 extraction papers were blocked as sensitive geopolitical content. These refusals clustered in contested regions, creating non-random missingness rather than ordinary extraction error. LLMs can therefore support systematic review workflows when used as auditable decision aids, but evidence synthesis protocols should pre-specify prompts, log raw outputs and refusals, report workload and validation metrics, and route refusals and high-risk fields to human review.
- Open Access
- 14 March 2026
- Published by Copernicus GmbH
LLM Workflow for Land-Use Prediction Evidence Synthesis: Efficient Screening, Selective Refusals, Reportable Gaps
Large language models (LLMs) promise to make systematic reviews more scalable and less costly, but the validity of LLM-assisted evidence synthesis depends not only on accuracy, but also on which parts of the literature are effectively visible to a deployed model and how reliably they are interpreted. We report a large-scale, domain-specific evaluation of an end-to-end LLM-assisted workflow for a systematic review of national-scale land use/land cover (LULC) prediction research (11,817 records; 11,688 after de-duplication), using a single hosted LLM deployment (Qwen Max) as a concrete case study. At title–abstract screening, the model behaved as a recall-oriented filter, excluding 9,891/11,688 records (84.7%) and routing 1,797 records for human follow-up; compared with the human baseline, it excluded fewer studies (84.7% vs 91.8%) and shifted more records into OK and POSSIBLE (OK: 4.2% vs 1.5%; POSSIBLE: 10.6% vs 5.5%). For full-text extraction, structured fields showed high agreement with expert coding across 342 benchmark papers (mean scores: 0.84 categorical, 0.85 temporal, 0.87 set-based), whereas free-text summaries were more variable (mean 0.79 overall; cosine similarity 0.51–0.87 across narrative fields despite high BERT-F1). In our case study, the workflow was completed in approximately one day on a single workstation for ~US$106 in API costs. Critically, full-text processing also produced explicit refusals: 7/2,084 candidate papers in deep screening and 2/345 papers targeted for insight extraction were blocked as “sensitive” geopolitical content. Although rare, these refusals were non-random and concentrated in contested regions, illustrating how LLM-specific constraints can introduce structured missingness that systematically removes or misinterprets evidence in precisely those settings where land-use conflict and governance are most salient. LLM-assisted reviews can therefore make previously prohibitive syntheses tractable. However, they must be embedded in transparent, human-led workflows that monitor and log model failures including refusals, omissions, and misreadings, and apply targeted auditing to detect and correct systematic blind spots.
- 14 March 2026
- Published by Copernicus GmbH
Open AccessLLM Workflow for Land-Use Prediction Evidence Synthesis: Efficient Screening, Selective Refusals, Reportable Gaps
Large language models (LLMs) promise to make systematic reviews more scalable and less costly, but the validity of LLM-assisted evidence synthesis depends not only on accuracy, but also on which parts of the literature are effectively visible to a deployed model and how reliably they are interpreted. We report a large-scale, domain-specific evaluation of an end-to-end LLM-assisted workflow for a systematic review of national-scale land use/land cover (LULC) prediction research (11,817 records; 11,688 after de-duplication), using a single hosted LLM deployment (Qwen Max) as a concrete case study. At title–abstract screening, the model behaved as a recall-oriented filter, excluding 9,891/11,688 records (84.7%) and routing 1,797 records for human follow-up; compared with the human baseline, it excluded fewer studies (84.7% vs 91.8%) and shifted more records into OK and POSSIBLE (OK: 4.2% vs 1.5%; POSSIBLE: 10.6% vs 5.5%). For full-text extraction, structured fields showed high agreement with expert coding across 342 benchmark papers (mean scores: 0.84 categorical, 0.85 temporal, 0.87 set-based), whereas free-text summaries were more variable (mean 0.79 overall; cosine similarity 0.51–0.87 across narrative fields despite high BERT-F1). In our case study, the workflow was completed in approximately one day on a single workstation for ~US$106 in API costs. Critically, full-text processing also produced explicit refusals: 7/2,084 candidate papers in deep screening and 2/345 papers targeted for insight extraction were blocked as “sensitive” geopolitical content. Although rare, these refusals were non-random and concentrated in contested regions, illustrating how LLM-specific constraints can introduce structured missingness that systematically removes or misinterprets evidence in precisely those settings where land-use conflict and governance are most salient. LLM-assisted reviews can therefore make previously prohibitive syntheses tractable. However, they must be embedded in transparent, human-led workflows that monitor and log model failures including refusals, omissions, and misreadings, and apply targeted auditing to detect and correct systematic blind spots.
- Open Access
- 16 July 2025
- Published by Elsevier in City and Environment Interactions
- Vol. 28
Impact of urban morphology on land surface temperature: A case study of the central Tokyo, Japan
- 16 July 2025
- Published by Elsevier in City and Environment Interactions
- Vol. 28
Open AccessImpact of urban morphology on land surface temperature: A case study of the central Tokyo, Japan
- 24 March 2025
- Published by Elsevier in Travel Behaviour and Society
- Vol. 40
A complex network analysis of urban human mobility in Tokyo
- 24 March 2025
- Published by Elsevier in Travel Behaviour and Society
- Vol. 40
A complex network analysis of urban human mobility in Tokyo
- Top Cited PapersOpen Access
- 28 September 2024
- Published by Elsevier in Landscape and Urban Planning
- Vol. 253
Urban green space in transition: A cross-continental perspective from eight Global North and South cities
- 28 September 2024
- Published by Elsevier in Landscape and Urban Planning
- Vol. 253
Top Cited PapersOpen AccessUrban green space in transition: A cross-continental perspective from eight Global North and South cities
- Open Access
Spatiotemporal Thermal Variations in Moroccan Cities: A Comparative Analysis
This study examines the Land Surface Temperature (LST) trends in eight key Moroccan cities from 1990 to 2020, emphasizing the influential factors and disparities between coastal and inland areas. Geographically weighted regression (GWR), machine learning (ML) algorithms, namely XGBoost and LightGBM, and SHapley Additive exPlanations (SHAP) methods are utilized. The study observes that urban areas are often cooler due to the presence of urban heat sinks (UHSs), more noticeably in coastal cities. However, LST is seen to increase across all cities due to urbanization and the degradation of vegetation cover. The increase in LST is more pronounced in inland cities surrounded by barren landscapes. Interestingly, XGBoost frequently outperforms LightGBM in the analyses. ML models and SHAP demonstrate efficacy in deciphering urban heat dynamics despite data quality and model tuning challenges. The study’s results highlight the crucial role of ongoing urbanization, topography, and the existence of water bodies and vegetation in driving LST dynamics. These findings underscore the importance of sustainable urban planning and vegetation cover in mitigating urban heat, thus having significant policy implications. Despite its contributions, this study acknowledges certain limitations, primarily the use of data from only four discrete years, thereby overlooking inter-annual, seasonal, and diurnal variations in LST dynamics.
Open AccessSpatiotemporal Thermal Variations in Moroccan Cities: A Comparative Analysis
This study examines the Land Surface Temperature (LST) trends in eight key Moroccan cities from 1990 to 2020, emphasizing the influential factors and disparities between coastal and inland areas. Geographically weighted regression (GWR), machine learning (ML) algorithms, namely XGBoost and LightGBM, and SHapley Additive exPlanations (SHAP) methods are utilized. The study observes that urban areas are often cooler due to the presence of urban heat sinks (UHSs), more noticeably in coastal cities. However, LST is seen to increase across all cities due to urbanization and the degradation of vegetation cover. The increase in LST is more pronounced in inland cities surrounded by barren landscapes. Interestingly, XGBoost frequently outperforms LightGBM in the analyses. ML models and SHAP demonstrate efficacy in deciphering urban heat dynamics despite data quality and model tuning challenges. The study’s results highlight the crucial role of ongoing urbanization, topography, and the existence of water bodies and vegetation in driving LST dynamics. These findings underscore the importance of sustainable urban planning and vegetation cover in mitigating urban heat, thus having significant policy implications. Despite its contributions, this study acknowledges certain limitations, primarily the use of data from only four discrete years, thereby overlooking inter-annual, seasonal, and diurnal variations in LST dynamics.