Hunter Lutz
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Hunter Lutz has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Garrett Uthlaut
Department of Systems and Information Engineering, University of Virginia, Charlottesville, VA 22904, USA
2 shared publications - Venkataraman Lakshmi
Department of Environmental Health and Engineering Johns Hopkins University Baltimore MD USA
2 shared publications - Amir Farzin Nikkhah
Department of Systems and Information Engineering, University of Virginia
1 shared publication - Arsalan Heydarian
Department of Civil and Environmental Engineering, University of Virginia, Charlottesville, VA
1 shared publication - Bradford Campbell
University of Virginia, Charlottesville, Virginia, USA
1 shared publication - Kyung Kim
Department of Environmental Health and Engineering, Whiting School of Engineering, Johns Hopkins University, Baltimore, MD 21218, USA
1 shared publication
Latest Publications
- Open Access
- 7 November 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Sensor Networks
MASHI: Multi-Agent Vision-Language System for Home Inspection
Home inspection evaluates the visible components of a residential building and reports their deficiencies, safety concerns, and recommended actions. Vision–language models (VLMs) can describe what a photograph shows, but inspection findings also require regulatory knowledge, system-level assessment, and familiarity with inspection practice. We present MASHI, a multi-agent vision–language system that answers home inspection questions across the ten systems of the ASHI Standard of Practice. A planner agent routes each question to an ASHI system and assigns tasks to a perception agent and a system agent. The perception agent reports visual evidence from the question’s photograph. The system agent applies the agent skill of the selected ASHI system to the question, its task, and this evidence. When needed, it retrieves text and images through multimodal retrieval-augmented generation (RAG) from a knowledge base of 45 building code editions and 18 inspection reports. For evaluation, a semi-automated pipeline with human verification turns 84 other inspection reports into 32,554 question–answer pairs. Prometheus-2 and Phi-4 rate the answers on a five-level rubric. MASHI reaches a mean judge score of 4.083, higher than those of all other compared systems, including its single-agent variant. Every difference is statistically significant.
- 7 November 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Sensor Networks
Open AccessMASHI: Multi-Agent Vision-Language System for Home Inspection
Home inspection evaluates the visible components of a residential building and reports their deficiencies, safety concerns, and recommended actions. Vision–language models (VLMs) can describe what a photograph shows, but inspection findings also require regulatory knowledge, system-level assessment, and familiarity with inspection practice. We present MASHI, a multi-agent vision–language system that answers home inspection questions across the ten systems of the ASHI Standard of Practice. A planner agent routes each question to an ASHI system and assigns tasks to a perception agent and a system agent. The perception agent reports visual evidence from the question’s photograph. The system agent applies the agent skill of the selected ASHI system to the question, its task, and this evidence. When needed, it retrieves text and images through multimodal retrieval-augmented generation (RAG) from a knowledge base of 45 building code editions and 18 inspection reports. For evaluation, a semi-automated pipeline with human verification turns 84 other inspection reports into 32,554 question–answer pairs. Prometheus-2 and Phi-4 rate the answers on a five-level rubric. MASHI reaches a mean judge score of 4.083, higher than those of all other compared systems, including its single-agent variant. Every difference is statistically significant.
- Open Access
- 4 September 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (17) , 3009
Surface Thermal State, Antecedent Hydroclimate, and Post-Fire Vegetation–Water Response in the Zambezi River Basin: A Multi-Source Environmental Time-Series Analysis
Wildfire in African savannas reflects coupled surface thermal, hydroclimatic, vegetation, and disturbance processes, but basin-scale time-series analyses can overstate mechanisms when temporal dependence and spatial heterogeneity are ignored. We assembled a monthly 2003–2024 multi-source environmental dataset for the Zambezi River Basin () and nine Level-4 HydroBASINS units, using quality-controlled MODIS burned area, normalized difference vegetation index (NDVI), daytime land surface temperature (LST), and evapotranspiration (ET), together with CHIRPS precipitation and GLDAS-2.1 Noah 0–10 cm soil moisture. Primary inference used heteroskedasticity- and autocorrelation-consistent regressions, expanding-window Random Forest validation, false-discovery-rate-controlled distributed-lag tests, and sub-basin robustness; vector autoregression was retained as a secondary diagnostic. Current-month burned-area anomalies were positively associated with LST (, ) and negatively associated with near-surface soil moisture (, ). Over 1–9 months, precipitation and soil-moisture histories were jointly supported after multiplicity correction, whereas basin-wide NDVI, ET, and LST histories were not. Burned-area history was followed by cumulative 0–6 month declines in NDVI (, 95% CI ) and ET (, 95% CI ). Directions were broadly consistent across sub-basins, although magnitudes varied. The evidence supports time-scale-specific conditional associations without structural-causal claims.- 4 September 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (17) , 3009
Open AccessSurface Thermal State, Antecedent Hydroclimate, and Post-Fire Vegetation–Water Response in the Zambezi River Basin: A Multi-Source Environmental Time-Series Analysis
Wildfire in African savannas reflects coupled surface thermal, hydroclimatic, vegetation, and disturbance processes, but basin-scale time-series analyses can overstate mechanisms when temporal dependence and spatial heterogeneity are ignored. We assembled a monthly 2003–2024 multi-source environmental dataset for the Zambezi River Basin () and nine Level-4 HydroBASINS units, using quality-controlled MODIS burned area, normalized difference vegetation index (NDVI), daytime land surface temperature (LST), and evapotranspiration (ET), together with CHIRPS precipitation and GLDAS-2.1 Noah 0–10 cm soil moisture. Primary inference used heteroskedasticity- and autocorrelation-consistent regressions, expanding-window Random Forest validation, false-discovery-rate-controlled distributed-lag tests, and sub-basin robustness; vector autoregression was retained as a secondary diagnostic. Current-month burned-area anomalies were positively associated with LST (, ) and negatively associated with near-surface soil moisture (, ). Over 1–9 months, precipitation and soil-moisture histories were jointly supported after multiplicity correction, whereas basin-wide NDVI, ET, and LST histories were not. Burned-area history was followed by cumulative 0–6 month declines in NDVI (, 95% CI ) and ET (, 95% CI ). Directions were broadly consistent across sub-basins, although magnitudes varied. The evidence supports time-scale-specific conditional associations without structural-causal claims. - Open Access
- 29 July 2026
- Published by MDPI AG in Preprints.org
Surface Thermal State, Hydrologic Mediation, and Post-Fire Memory: A Multimethod Satellite Analysis of Wildfire–Vegetation–Water Coupling in the Zambezi River Basin
Wildfire regimes in African savannas reflect interactions among surface heating, fuel availability, moisture constraints, and post-fire vegetation recovery. These processes are often analyzed jointly, making it difficult to distinguish near-term fire susceptibility from antecedent fuel–moisture controls and post-fire land-surface feedbacks. We examine these coupled processes in the Zambezi River Basin using monthly basin-aggregated satellite time series from January 2003 through November 2024 for burned area, land surface temperature (LST), precipitation, soil moisture, evapotranspiration (ET), and normalized difference vegetation index (NDVI). Random Forest regression, vector autoregression, impulse-response analysis, Granger predictive tests, and distributed lag regression are assigned non-overlapping inferential roles. Among the included predictors, surface soil moisture is the dominant contemporaneous land-surface signal in the Random Forest, while positive LST anomalies are associated with elevated burned-area anomalies at a one-month lag. Precipitation, soil moisture, NDVI, and ET predict burned area over seasonal lags through moisture-suppression and fuel-accumulation pathways. Burned-area history also improves prediction of subsequent NDVI and ET variability, indicating post-fire vegetation–water memory. The results support a three-component framework separating short-lag surface thermal state, antecedent hydrologic mediation, and post-fire vegetation–water response.
- 29 July 2026
- Published by MDPI AG in Preprints.org
Open AccessSurface Thermal State, Hydrologic Mediation, and Post-Fire Memory: A Multimethod Satellite Analysis of Wildfire–Vegetation–Water Coupling in the Zambezi River Basin
Wildfire regimes in African savannas reflect interactions among surface heating, fuel availability, moisture constraints, and post-fire vegetation recovery. These processes are often analyzed jointly, making it difficult to distinguish near-term fire susceptibility from antecedent fuel–moisture controls and post-fire land-surface feedbacks. We examine these coupled processes in the Zambezi River Basin using monthly basin-aggregated satellite time series from January 2003 through November 2024 for burned area, land surface temperature (LST), precipitation, soil moisture, evapotranspiration (ET), and normalized difference vegetation index (NDVI). Random Forest regression, vector autoregression, impulse-response analysis, Granger predictive tests, and distributed lag regression are assigned non-overlapping inferential roles. Among the included predictors, surface soil moisture is the dominant contemporaneous land-surface signal in the Random Forest, while positive LST anomalies are associated with elevated burned-area anomalies at a one-month lag. Precipitation, soil moisture, NDVI, and ET predict burned area over seasonal lags through moisture-suppression and fuel-accumulation pathways. Burned-area history also improves prediction of subsequent NDVI and ET variability, indicating post-fire vegetation–water memory. The results support a three-component framework separating short-lag surface thermal state, antecedent hydrologic mediation, and post-fire vegetation–water response.