Using large language models for safety-related table summarization in clinical study reports.

Objectives: The generation of structured documents for clinical trials is a promising application of large language models (LLMs). We share opportunities, insights, and challenges from a competitive challenge that used LLMs for automating clinical trial documentation. Materials and Methods: As part of a challenge initiated by Pfizer (organizer), several teams (participant) created a pilot for generating summaries of safety tables for clinical study reports (CSRs). Our evaluation framework used automated metrics and expert reviews to assess the quality of AI-generated documents. Results: The comparative analysis revealed differences in performance across solutions, particularly in factual accuracy and lean writing. Most participants employed prompt engineering with generative pre-trained transformer (GPT) models. Discussion: We discuss areas for improvement, including better ingestion of tables, addition of context and fine-tuning. Conclusion: The challenge results demonstrate the potential of LLMs in automating table summarization in CSRs while also revealing the importance of human involvement and continued research to optimize this technology.

Three-phase hierarchical model-based and hybrid inference.

Global commitments to mitigating climate change and halting biodiversity loss require reliable information about Earth's ecosystems. Increasingly, such information is obtained from multiple sources of remotely sensed data combined with data acquired in the field. This new wealth of data poses challenges regarding the combination of different data sources to derive the required information and assess uncertainties. In this article, we show how predictors and their variances can be derived when hierarchically nested models are applied. Previous studies have developed methods for cases involving two modeling steps, such as biomass prediction relying on tree-level allometric models and models linking plot-level field data with remotely sensed data. This study extends the analysis to cases involving three modeling steps to cover new important applications. The additional step might involve an intermediate model, linking field and remotely sensed data available from a small sample, for making predictions that are subsequently used for training a final prediction model based on remotely sensed data:•In cases where the data in the final step are available wall-to-wall, we denote the approach three-phase hierarchical model-based inference (3pHMB),•In cases where the data in the final step are available as a probability sample, we denote the approach three-phase hierarchical hybrid inference (3pHHY).

Assessing the Effects of Fire Disturbances and Timber Management on Carbon Storage in the Greater Yellowstone Ecosystem.

Accurate characterization of Carbon (C) consequences of forest disturbances and management is critical for informed climate mitigation and adaptation strategies. While research into generalized properties of the forest C cycle informs policy and provides abstract guidance to managers, most management occurs at local scales and relies upon monitoring systems that can consistently provide C cycle assessments that explicitly apply to a defined time and place. We used an inventory-based forest monitoring and simulation tool to quantify C storage effects of actual fires, timber harvests, and forest regeneration conditions in the Greater Yellowstone Ecosystem (GYE). Results show that (1) the 1988 fires had a larger impact on GYE's C storage than harvesting during 1985-2011; (2) continuation of relatively high harvest rates of the region's National Forest land, which declined after 1990, would have shifted the disturbance agent primary importance on those lands from fire to harvest; and (3) accounting for local heterogeneity of post-disturbance regeneration patterns translates into large regional effects on total C storage. Large fires in 1988 released about 8.3 ± 0.3 Mg/ha of C across Yellowstone National Park (YNP, including both disturbed and undisturbed area), compared with total C storage reductions due to harvest of about 2.3 ± 0.3 Mg/ha and 2.6 ± 0.2 Mg/ha in adjacent Caribou-Targhee and Gallatin National Forests, respectively, from 1985-2011. If the high harvest rates observed in 1985-1989 had been maintained through 2011 in GYE National Forests, the C storage effect of harvesting would have quintupled to 10.5 ± 1.0 Mg/ha, exceeding the immediate losses associated with YNP's historic fire but not the longer-term net loss of carbon (16.9 ± 0.8 Mg/ha). Following stand-replacing disturbance such as the 1988 fires, the actual regeneration rate was slower than the default regional average rate assumed by empirically calibrated forest growth models. If regeneration following the 1988 fire had reached regionally average rates, either through different natural circumstances or through more active management, YNP would have had approximately 4.1 Mg/ha more forest carbon by year 2020. This study highlights the relative effects of fire disturbances and management activities on regional C storage, and demonstrates a forest carbon monitoring system that can be both applied consistently across the US and tailored to questions of specific local management interest.

A sample design for globally consistent biomass estimation using lidar data from the Geoscience Laser Altimeter System (GLAS).

BACKGROUND: Lidar height data collected by the Geosciences Laser Altimeter System (GLAS) from 2002 to 2008 has the potential to form the basis of a globally consistent sample-based inventory of forest biomass. GLAS lidar return data were collected globally in spatially discrete full waveform "shots," which have been shown to be strongly correlated with aboveground forest biomass. Relationships observed at spatially coincident field plots may be used to model biomass at all GLAS shots, and well-established methods of model-based inference may then be used to estimate biomass and variance for specific spatial domains. However, the spatial pattern of GLAS acquisition is neither random across the surface of the earth nor is it identifiable with any particular systematic design. Undefined sample properties therefore hinder the use of GLAS in global forest sampling. RESULTS: We propose a method of identifying a subset of the GLAS data which can justifiably be treated as a simple random sample in model-based biomass estimation. The relatively uniform spatial distribution and locally arbitrary positioning of the resulting sample is similar to the design used by the US national forest inventory (NFI). We demonstrated model-based estimation using a sample of GLAS data in the US state of California, where our estimate of biomass (211 Mg/hectare) was within the 1.4% standard error of the design-based estimate supplied by the US NFI. The standard error of the GLAS-based estimate was significantly higher than the NFI estimate, although the cost of the GLAS estimate (excluding costs for the satellite itself) was almost nothing, compared to at least US$ 10.5 million for the NFI estimate. CONCLUSIONS: Global application of model-based estimation using GLAS, while demanding significant consolidation of training data, would improve inter-comparability of international biomass estimates by imposing consistent methods and a globally coherent sample frame. The methods presented here constitute a globally extensible approach for generating a simple random sample from the global GLAS dataset, enabling its use in forest inventory activities.

Estimates of carbon stored in harvested wood products from the United States forest service northern region, 1906-2010.

BACKGROUND: Global forests capture and store significant amounts of CO2 through photosynthesis. When carbon is removed from forests through harvest, a portion of the harvested carbon is stored in wood products, often for many decades. The United States Forest Service (USFS) and other agencies are interested in accurately accounting for carbon flux associated with harvested wood products (HWP) to meet greenhouse gas monitoring commitments and climate change adaptation and mitigation objectives. This paper uses the Intergovernmental Panel on Climate Change (IPCC) production accounting approach and the California Forest Project Protocol (CFPP) to estimate HWP carbon storage from 1906 to 2010 for the USFS Northern Region, which includes forests in northern Idaho, Montana, South Dakota, and eastern Washington. RESULTS: Based on the IPCC approach, carbon stocks in the HWP pool were increasing at one million megagrams of carbon (MgC) per year in the mid 1960s, with peak cumulative storage of 28 million MgC occurring in 1995. Net positive flux into the HWP pool over this period is primarily attributable to high harvest levels in the mid twentieth century. Harvest levels declined after 1970, resulting in less carbon entering the HWP pool. Since 1995, emissions from HWP at solid waste disposal sites have exceeded additions from harvesting, resulting in a decline in the total amount of carbon stored in the HWP pool. The CFPP approach shows a similar trend, with 100-year average carbon storage for each annual Northern Region harvest peaking in 1969 at 937,900 MgC, and fluctuating between 84,000 and 150,000 MgC over the last decade. CONCLUSIONS: The Northern Region HWP pool is now in a period of negative net annual stock change because the decay of products harvested between 1906 and 2010 exceeds additions of carbon to the HWP pool through harvest. However, total forest carbon includes both HWP and ecosystem carbon, which may have increased over the study period. Though our emphasis is on the Northern Region, we provide a framework by which the IPCC and CFPP methods can be applied broadly at sub-national scales to other regions, land management units, or firms.

Changes in timber haul emissions in the context of shifting forest management and infrastructure.

BACKGROUND: Although significant amounts of carbon may be stored in harvested wood products, the extraction of that carbon from the forest generally entails combustion of fossil fuels. The transport of timber from the forest to primary milling facilities may in particular create emissions that reduce the net sequestration value of product carbon storage. However, attempts to quantify the effects of transport on the net effects of forest management typically use relatively sparse survey data to determine transportation emission factors. We developed an approach for systematically determining transport emissions using: 1) -remotely sensed maps to estimate the spatial distribution of harvests, and 2) - industry data to determine landscape-level harvest volumes as well as the location and processing totals of individual mills. These data support spatial network analysis that can produce estimates of fossil carbon released in timber transport. RESULTS: Transport-related emissions, evaluated as a fraction of transported wood carbon at 4 points in time on a landscape in western Montana (USA), rose from 0.5% in 1988 to 1.7% in 2004 as local mills closed and spatial patterns of harvest shifted due to decreased logging on federal lands. CONCLUSION: The apparent sensitivity of transport emissions to harvest and infrastructure patterns suggests that timber haul is a dynamic component of forest carbon management that bears further study both across regions and over time. The monitoring approach used here, which draws only from widely available monitoring data, could readily be adapted to provide current and historical estimates of transport emissions in a consistent way across large areas.