Geology Reports⌕ Search

SEARCH · Geology Reports

Results for “Algorithms”

Search indexed USGS publications on groundwater, aquifers, geologic maps, mineral resources and earthquakes. Explore source records by subject and place.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 685 records · Page 38Linked to original sources
USGS2006

KGS-HighK: A Fortran 90 program for simulation of hydraulic tests in highly permeable aquifers

Slug and pumping tests (hydraulic tests) are frequently used by hydrogeologists to obtain in-situ estimates of the transmissive and storage properties of a formation (Streltsova, 1988; Kruseman and de Ridder, 1990; Butler, 1998). In aquifers of high hydraulic conductivity, hydraulic tests are affected by mechanisms that are not considered in the analysis of tests in less permeable media (Bredehoeft et al., 1966). Inertia-induced oscillations in hydraulic head are the most common manifestation of such mechanisms. Over the last three decades, a number of analytical solutions that incorporate these mechanisms have been developed for the analysis of hydraulic tests in highly permeable aquifers (see Butler and Zhan (2004) for a review of this previous work). These solutions, however, are restricted to a subset of the conditions commonly encountered in the field. Recently, a more general solution has been developed that builds on this previous work to remove many of the limitations imposed by these earlier approaches (Butler and Zhan, 2004). The purpose of this note is to present a Fortran 90 program, KGS-HighK, for the evaluation of this new solution. This note begins with a brief overview of the conceptual model that motivated the development of the solution of Butler and Zhan (2004) for pumping- and slug-induced flow to/from a central well. The major steps in the derivation of that solution are described, but no details are given. Instead, a Mathematica notebook is provided for those interested in the derivation details. The key algorithms used in KGS-HighK are then described and the program structure is briefly outlined. A field example is provided to demonstrate program performance. The note concludes with a short summary section. ?? 2005 Elsevier Ltd. All rights reserved.

Computers & Geosciences↗
USGS2013

A wetting and drying scheme for ROMS

The processes of wetting and drying have many important physical and biological impacts on shallow water systems. Inundation and dewatering effects on coastal mud flats and beaches occur on various time scales ranging from storm surge, periodic rise and fall of the tide, to infragravity wave motions. To correctly simulate these physical processes with a numerical model requires the capability of the computational cells to become inundated and dewatered. In this paper, we describe a method for wetting and drying based on an approach consistent with a cell-face blocking algorithm. The method allows water to always flow into any cell, but prevents outflow from a cell when the total depth in that cell is less than a user defined critical value. We describe the method, the implementation into the three-dimensional Regional Oceanographic Modeling System (ROMS), and exhibit the new capability under three scenarios: an analytical expression for shallow water flows, a dam break test case, and a realistic application to part of a wetland area along the Georgia Coast, USA.

Georgia↗
USGS2014

MTpy: A Python toolbox for magnetotellurics

We present the software package MTpy that allows handling, processing, and imaging of magnetotelluric (MT) data sets. Written in Python, the code is open source, containing sub-packages and modules for various tasks within the standard MT data processing and handling scheme. Besides the independent definition of classes and functions, MTpy provides wrappers and convenience scripts to call standard external data processing and modelling software. In its current state, modules and functions of MTpy work on raw and pre-processed MT data. However, opposite to providing a static compilation of software, we prefer to introduce MTpy as a flexible software toolbox, whose contents can be combined and utilised according to the respective needs of the user. Just as the overall functionality of a mechanical toolbox can be extended by adding new tools, MTpy is a flexible framework, which will be dynamically extended in the future. Furthermore, it can help to unify and extend existing codes and algorithms within the (academic) MT community. In this paper, we introduce the structure and concept of MTpy . Additionally, we show some examples from an everyday work-flow of MT data processing: the generation of standard EDI data files from raw electric ( E -) and magnetic flux density ( B -) field time series as input, the conversion into MiniSEED data format, as well as the generation of a graphical data representation in the form of a Phase Tensor pseudosection.

Computers & Geosciences↗
USGS2018

Determining on-fault earthquake magnitude distributions from integer programming

Earthquake magnitude distributions among faults within a fault system are determined from regional seismicity and fault slip rates using binary integer programming. A synthetic earthquake catalog (i.e., list of randomly sampled magnitudes) that spans millennia is first formed, assuming that regional seismicity follows a Gutenberg-Richter relation. Each earthquake in the synthetic catalog can occur on any fault and at any location. The objective is to minimize misfits in the target slip rate for each fault, where slip for each earthquake is scaled from its magnitude. The decision vector consists of binary variables indicating which locations are optimal among all possibilities. Uncertainty estimates in fault slip rates provide explicit upper and lower bounding constraints to the problem. An implicit constraint is that an earthquake can only be located on a fault if it is long enough to contain that earthquake. A general mixed-integer programming solver, consisting of a number of different algorithms, is used to determine the optimal decision vector. A case study is presented for the State of California, where a 4 kyr synthetic earthquake catalog is created and faults with slip ≥3 mm/yr are considered, resulting in >10 6 variables. The optimal magnitude distributions for each of the faults in the system span a rich diversity of shapes, ranging from characteristic to power-law distributions.

Computers & Geosciences↗
USGS2026

3D semantic mapping of surface geological features

Semantic mapping in 3D is fundamental to a wide range of geoscientific studies and applications, including geomorphology, hazard assessment, and environmental monitoring. However, automatically segmenting geological features from large-scale photogrammetric datasets remains a significant challenge. We present a methodology to address this gap. Using overlapping images collected over environments of interest, Structure-from-Motion (SfM) produces georeferenced point clouds and estimates camera poses. Existing large vision models, such as Segment Anything Model, segment objects in the images, generating pixel-segmentation associations. To produce pixel-point associations, we project the points back onto the camera image planes. As objects are independently segmented across multiple images with different perspectives, we develop a segmentation mosaicking algorithm to build probabilistic point-segmentation associations that combines the pixel-segmentation associations and pixel-point associations. Our methodology is validated using both synthetic data generated by Kubric and real-world UAV-SfM data. The implementation is designed to be compatible with existing SfM software, including Agisoft and OpenDroneMap, for photogrammetry mapping in geoscience studies. As a case study, we apply our method to the semantic mapping of precariously balanced rocks (PBRs), which provide upper-bound constraints on historical ground motion shaking intensity. To support object-level identification of PBRs, we additionally integrated Grounding DINO, enabling text-prompted segmentation of features of interest within UAV imagery. This case study demonstrates the effectiveness of our method in generating a 3D semantic map of PBRs, enabling spatial distribution of PBR fragility for earthquake hazard analysis.

Computers & Geosciences↗
USGS2022

Fingerprinting historical tributary contributions to floodplain sediment using bulk geochemistry

Sediment deposition on floodplains is essential for the development and maintenance of riparian ecosystems. Upstream erosion is known to influence downstream floodplain construction, but linking these disparate processes is challenging, especially over large spatial and temporal scales. Sediment fingerprinting is thus a robust tool to establish process linkages between downstream floodplain development and sediment production in distal headwater basins. Here we use sediment geochemistry to connect historical erosion in several tributaries of the Yampa River in Colorado and Wyoming, USA, to the construction of downstream floodplains on which extensive cottonwood forests established. Using a combination of conventional techniques and the relatively novel machine-learning random forest algorithm, we build multiple fingerprints of diagnostic geochemical tracers that are then input into a Bayesian mixing model to apportion provenance of floodplain sediment. Sediment samples for provenance analysis were collected from an excavated floodplain in Deerlodge Park on the Yampa River at the rooting surface of the surrounding cottonwood forest and dominantly comprised of very fine (4Φ) sand. Fingerprinting analysis of the 4Φ fraction of collected floodplain sink (n = 38) and tributary source (n = 218) samples revealed floodplain sediment to be dominantly sourced from the tributaries of Muddy Creek (45 ± 4%) and Sand Wash (42 ± 6%). Dendrochronology results moreover indicate the Deerlodge floodplain sediment was deposited in ∼1912, which falls squarely within the time (1880–1940) these tributaries were actively eroding. Taken together, study results indicate a demonstrable link between historical tributary erosion and downstream floodplain construction and concomitant forest establishment. Our findings suggest processes operating in tributary watersheds play an important role in the dynamics of large rivers and emphasize both the need for holistic, collaborative management of sediment as an essential resource and the potential to utilize sediment fingerprinting to inform and direct river ecosystem management.

Colorado, Wyoming↗
USGS2022

Machine learning and data augmentation approach for identification of rare earth element potential in Indiana Coals, USA

Rare earth elements and yttrium (REYs) are critical elements and valuable commodities due to their limited availability and high demand in a wide range of applications and especially in high-technology products. The increased demand and geopolitical pressures motivate the search for alternative sources of REYs, and coal, coal waste, and coal ash are considered as new sources for these critical elements. This research evaluates the REY potential of coals from Indiana (USA). However, although coal data revealed REY potential, it suffered from sparse samples with complete REY measurements. Therefore, we explore the applicability of machine learning (ML) models and data augmentation techniques to demonstrate their applicability to evaluate REY potential in Indiana, and other areas in coal basins, using selected coal parameters (Al2O3, Fe2O3, C, Ash, S, P, Mo, Zn, and As contents) as covariates (indicators). Due to the relatively small sample size with complete REY data in the Indiana Coal Database, two data augmentation techniques (Random Over-Sampling Examples and Synthetic Minority Over-Sampling Technique) were used. Four machine learning algorithms (linear discriminate analysis, support vector machine, random forest, and artificial neural networks) were applied for modeling REY potential as a classification problem. The results show that application of Synthetic Minority Over-Sampling Technique prior to development of the support vector machine (SVM) models generated the best REY classification with an accuracy of 95%. The encouraging results based on Indiana coal data may suggest that a similar approach can be used for other coal basins for screening the locations with REY potential. Those locations then can be targeted for more detailed geochemical surveys to identify most promising areas and evaluate overall REY resources.

Indiana↗
USGS2023

Merging machine learning and geostatistical approaches for spatial modeling of geoenergy resources

Geostatistics is the most commonly used probabilistic approach for modeling earth systems, including quality parameters of various geoenergy resources. In geostatistics, estimates, either on a point or block support, are generated as a spatially-weighted average of surrounding samples. The optimal weights are determined through the stationary variogram model which accounts for the spatial structure of the samples. Recently, efficient modeling workflows using various machine learning algorithms (MLAs) have been expanded to the spatial context for modeling geological heterogeneity. The flexible use of MLAs as a spatial estimation tool stems mainly from the fact that unlike kriging, they do not require any variogram, nor do they depend strongly on a prior stationarity assumption (i.e., second order stationarity). This study evaluates the performance of two MLAs (ensemble super learner and elliptical radial basis neural network), ordinary kriging, and hybrid spatial modeling approaches using ordinary intrinsic collocated cokriging. The aforementioned modeling techniques are compared for estimating resources for four coal variables (wash yield, ash yield, calorific value and thickness) as an example. The results suggest that MLAs, when implemented alone, do not outperform ordinary kriging, but the estimation accuracy of the final model, measured by the root mean squared error tends to subtly improve (

Virginia↗
USGS2019

Debris-flow monitoring and warning: Review and examples

Debris flows represent one of the most dangerous types of mass movements, because of their high velocities, large impact forces and long runout distances. This review describes the available debris-flow monitoring techniques and proposes recommendations to inform the design of future monitoring and warning/alarm systems. The selection and application of these techniques is highly dependent on site and hazard characterization, which is illustrated through detailed descriptions of nine monitoring sites: five in Europe, three in Asia and one in the USA. Most of these monitored catchments cover less than ∼10 km 2 and are topographically rugged with Melton Indices greater than 0.5. Hourly rainfall intensities between 5 and 15 mm/h are sufficient to trigger debris flows at many of the sites, and observed debris-flow volumes range from a few hundred up to almost one million cubic meters. The sensors found in these monitoring systems can be separated into two classes: a class measuring the initiation mechanisms, and another class measuring the flow dynamics. The first class principally includes rain gauges, but also contains of soil moisture and pore-water pressure sensors. The second class involves a large variety of sensors focusing on flow stage or ground vibrations and commonly includes video cameras to validate and aid in the data interpretation. Given the sporadic nature of debris flows, an essential characteristic of the monitoring systems is the differentiation between a continuous mode that samples at low frequency (“non-event mode”) and another mode that records the measurements at high frequency (“event mode”). The event detection algorithm, used to switch into the “event mode” depends on a threshold that is typically based on rainfall or ground vibration. Identifying the correct definition of these thresholds is a fundamental task not only for monitoring purposes, but also for the implementation of warning and alarm systems.

Earth-Science Reviews↗
USGS2023

Evaluating a tandem human-machine approach to labelling of wildlife in remote camera monitoring

Remote cameras (“trail cameras”) are a popular tool for non-invasive, continuous wildlife monitoring, and as they become more prevalent in wildlife research, machine learning (ML) is increasingly used to automate or accelerate the labor-intensive process of labelling (i.e., tagging) photos. Human-machine hybrid tagging approaches have been shown to greatly increase tagging efficiency (i.e., time to tag a single image). However, those potential increases hinge on the extent to which an ML model makes correct vs. incorrect predictions. We performed an experiment using a ML model that produces bounding boxes around animals, people, and vehicles in remote camera imagery (MegaDetector) to consider the impact of a ML model’s performance on its ability to accelerate human labeling. Six participants tagged trail camera images collected from 12 sites in Vermont and Maine, USA (January–September 2022) using three tagging methods (one with ML bounding box assistance and two without assistance). We used a generalized linear mixed model to examine the influence of ML model performance and tagging method on tagging efficiency. We found that ML bounding boxes offer significant improvement in tagging efficiency when labelling data compared to unassisted tagging. Additionally, the time taken to label with bounding boxes was not statistically different from an unassisted tagging approach. However, we found that gains in efficiency are contingent on the ML algorithm’s performance and that incorrect ML predictions, particularly the 4.2% false positive and 3.6% false negative predictions, can slow the tagging process compared to a non-hybrid approach. These findings indicate that although practitioners usually forgo the production of bounding boxes when selecting a data labelling process due to the increased effort, ML bounding box-assisted tagging can offer an efficient method for labeling. More broadly, ML-assisted data labelling offers an opportunity to accelerate the analysis of trail camera imagery, but an assessment of the ML model’s performance can illuminate whether the hybrid-tagging approach is ultimately a help or hinderance.

Maine, Vermont↗
USGS2013

Evapotranspiration and water balance of an anthropogenic coastal desert wetland: responses to fire, inflows and salinities

Evapotranspiration (ET) and other water balance components were estimated for Cienega de Santa Clara, an anthropogenic brackish wetland in the delta of the Colorado River in Mexico. The marsh is in the Biosphere Reserve of the Upper Gulf of California and Delta of the Colorado River, and supports a high abundance and diversity of wildlife. Over 95% of its water supply originates as agricultural drain water from the USA, sent for disposal in Mexico. This study was conducted from 2009 to 2011, before, during and after a trial run of the Yuma Desalting Plant in the USA, which will divert water from the wetland and replace it with brine from the desalting operation. The goal was to estimate the main components in the water budget to be used in creating management scenarios for this marsh. We used a remote sensing algorithm to estimate ET from meteorological data and Enhanced Vegetation Index values from the Moderate Resolution Imaging Spectrometer (MODIS) sensors on the Terra satellite. ET estimates from the MODIS method were then compared to results from a mass balance of water and salt inflows and outflows over the study period. By both methods, mean annual ET estimates ranged from 2.6 to 3.0 mm d −1 , or 50 to 60% of reference ET (ET o ). Water entered at a mean salinity of 2.6 g L −1 TDS and mean salinity in the wetland was 3.73 g L −1 TDS over the 33 month study period. Over an annual cycle, 54% of inflows supported ET while the rest exited the marsh as outflows; however, in winter when ET was low, up to 90% of the inflows exited the marsh. An analysis of ET estimates over the years 2000–2011 showed that annual ET was proportional to the volume of inflows, but was also markedly stimulated by fires. Spring fires in 2006 and 2011 burned off accumulated thatch, resulting in vigorous growth of new leaves and a 30% increase in peak summer ET compared to non-fire years. Following fires, peak summer ET estimates were equal to ET o , while in non-fire years peak ET was equal to only one-half to two-thirds of ET o . Over annual cycles, estimated ET was always lower than ET o , because T. domingensis is dormant in winter and shades the water surface, reducing direct evaporation. Thus, ET of a Typha marsh is likely to be less than an open water surface under most conditions.

Ecological Engineering↗
USGS2018

Estimating vegetation biomass and cover across large plots in shrub and grass dominated drylands using terrestrial lidar and machine learning

Terrestrial laser scanning (TLS) has been shown to enable an efficient, precise, and non-destructive inventory of vegetation structure at ranges up to hundreds of meters. We developed a method that leverages TLS collections with machine learning techniques to model and map canopy cover and biomass of several classes of short-stature vegetation across large plots. We collected high-definition TLS scans of 26 1-ha plots in desert grasslands and big sagebrush shrublands in southwest Idaho, USA. We used the Random Forests machine learning algorithm to develop decision tree models predicting the biomass and canopy cover of several vegetation classes from statistical descriptors of the aboveground heights of TLS points. Manual measurements of vegetation characteristics collected within each plot served as training and validation data. Models based on five or fewer TLS descriptors of vegetation heights were developed to predict the canopy cover fraction of shrubs (R 2 = 0.77, RMSE = 7%), annual grasses (R 2 = 0.70, RMSE = 21%), perennial grasses (R 2 = 0.36, RMSE = 12%), forbs (R 2 = 0.52, RMSE = 6%), bare earth or litter (R 2 = 0.49, RMSE = 19%), and the biomass of shrubs (R 2 = 0.71, RMSE = 175 g) and herbaceous vegetation (R 2 = 0.61, RMSE = 99 g) (all values reported are out-of-bag). Our models explained much of the variability between predictions and manual measurements, and yet we expect that future applications could produce even better results by reducing some of the methodological sources of error that we encountered. Our work demonstrates how TLS can be used efficiently to extend manual measurement of vegetation characteristics from small to large plots in grasslands and shrublands, with potential application to other similarly structured ecosystems. Our method shows that vegetation structural characteristics can be modeled without classifying and delineating individual plants, a challenging and time-consuming step common in previous methods applying TLS to vegetation inventory. Improving application of TLS to studies of shrub-steppe ecosystems will serve immediate management needs by enhancing vegetation inventories, environmental modeling studies, and the ability to train broader datasets collected from air and space.

Idaho↗
USGS2021

Are drought indices and climate data good indicators of ecologically relevant soil moisture dynamics in drylands?

Droughts are disproportionately impacting global dryland regions where ecosystem health and function are tightly coupled to moisture availability. Drought severity is commonly estimated using algorithms such as the standardized precipitation-evapotranspiration index (SPEI), which can estimate climatic water balance impacts at various hydrologic scales by varying computational length. However, the performance of these metrics as indicators of soil moisture dynamics at ecologically relevant scales, across soil depths, and in consideration of broader scale ecohydrological processes, requires more attention. In this study, we tested components of climatic water balance, including SPEI and SPEI computation lengths, to recreate multi-decadal and periodic soil-moisture patterns across soil profiles at 866 sites in the western United States. Modeling results show that SPEI calculated over the prior 12-months was the most predictive computation length and could recreate changes in moisture availability within the soil profile over longer periods of time and for annual recharge of deeper soil moisture stores. SPEI was slightly less successful with recreating spring surface-soil moisture availability, which is key to dryland ecosystems dominated by winter precipitation. Meteorological drought indices like SPEI are intended to be convenient and generalized indicators of meteorological water deficit. However, the inconsistent ability of SPEI to recreate ecologically relevant patterns of soil moisture at regional scales suggests that process-based models, and the larger data requirements they involve, remain an important tool for dryland ecohydrology

California, Idaho, Nevada, Oregon, Utah↗
USGS2022

Satellites quantify the spatial extent of cyanobacterial blooms across the United States at multiple scales

Previous studies indicate that cyanobacterial harmful algal bloom (cyanoHAB) frequency, extent, and magnitude have increased globally over the past few decades. However, little quantitative capability is available to assess these metrics of cyanoHABs across broad geographic scales and at regular intervals. Here, the spatial extent was quantified from a cyanobacteria algorithm applied to two European Space Agency satellite platforms—the MEdium Resolution Imaging Spectrometer (MERIS) onboard Envisat and the Ocean and Land Colour Instrument (OLCI) onboard Sentinel-3. CyanoHAB spatial extent was defined for each geographic area as the percentage of valid satellite pixels that exhibited cyanobacteria above the detection limit of the satellite sensor. This study quantified cyanoHAB spatial extent for over 2,000 large lakes and reservoirs across the contiguous United States (CONUS) during two time periods: 2008–2011 via MERIS and 2017–2020 via OLCI when cloud-, ice-, and snow-free imagery was available. Approximately 56% of resolvable lakes were glaciated, 13% were headwater, isolated, or terminal lakes, and the rest were primarily drainage lakes. Results were summarized at national-, regional-, state-, and lake-scales, where regions were defined as nine climate regions which represent climatically consistent states. As measured by satellite, changes in national cyanoHAB extent did have a strong increase of 6.9% from 2017 to 2020 (|Kendall’s tau (τ)| = 0.56; gamma (γ) = 2.87 years), but had negligible change (|τ| = 0.03) from 2008 to 2011. Two of the nine regions had moderate (0.3 ≤ |τ| < 0.5) increases in spatial extent from 2017 to 2020, and eight of nine regions had negligible (|τ| < 0.2) change from 2008 to 2011. Twelve states had a strong or moderate increase from 2017 to 2020 (|τ| ≥ 0.3), while only one state had a moderate increase and two states had a moderate decrease from 2008 to 2011. A decrease, or no change, in cyanoHAB spatial extent did not indicate a lack of issues related to cyanoHABs. Sensitivity results of randomly omitted daily CONUS scenes confirm that even with reduced data availability during a short four-year temporal assessment, the direction and strength of the changes in spatial extent remained consistent. We present the first set of national maps of lake cyanoHAB spatial extent across CONUS and demonstrate an approach for quantifying past and future changes at multiple spatial scales. Results presented here provide water quality managers information regarding current cyanoHAB spatial extent and quantify rates of change.

Ecological Indicators↗
USGS2024

Passive acoustic monitoring and convolutional neural networks facilitate high-resolution and broadscale monitoring of a threatened species

Population monitoring is an essential component of biodiversity conservation and management, but low detection probabilities for rare and/or cryptic species makes estimating abundance and occupancy challenging. Passive acoustic monitoring combined with machine learning algorithms represents a potential path forward to effectively and efficiently monitor the occurrence of rare vocalizing species across entire forest landscapes. Our objectives were to develop and implement a convolutional neural network (PNW-Cnet) to identify vocalizations of a rare and threatened forest nesting bird species – the marbled murrelet ( Brachyramphus marmoratus ) – in the Pacific Northwest, U.S.A., 2018–2021. We used PNW-Cnet predictions from broadscale passive acoustic monitoring data to examine spatiotemporal patterns in the distribution of murrelets. PNW-Cnet showed sufficiently high prediction accuracy (overall precision > 0.9) to enable broadscale population monitoring. Spatiotemporal analysis showed that annual peak murrelet call abundance occurs in ordinal weeks 28–32 (late July–Mid August) but this varied by study area. The greatest number of detections typically occurred in the Olympic Peninsula and Oregon Coast Range where late-successional forest dominates and nearer to ocean habitats. We demonstrate that passive acoustic monitoring can be used to understand intensity of use across broad scales for a rare and cryptic species in addition to the typical detection/non-detection data that are often collected. Passive acoustic monitoring combined with PNW-Cnet offers considerable promise for species distribution modeling and long-term population monitoring for rare species.

Oregon, Washington↗
USGS2006

Predicting tree species presence and basal area in Utah: A comparison of stochastic gradient boosting, generalized additive models, and tree-based methods

Many efforts are underway to produce broad-scale forest attribute maps by modelling forest class and structure variables collected in forest inventories as functions of satellite-based and biophysical information. Typically, variants of classification and regression trees implemented in Rulequest's?? See5 and Cubist (for binary and continuous responses, respectively) are the tools of choice in many of these applications. These tools are widely used in large remote sensing applications, but are not easily interpretable, do not have ties with survey estimation methods, and use proprietary unpublished algorithms. Consequently, three alternative modelling techniques were compared for mapping presence and basal area of 13 species located in the mountain ranges of Utah, USA. The modelling techniques compared included the widely used See5/Cubist, generalized additive models (GAMs), and stochastic gradient boosting (SGB). Model performance was evaluated using independent test data sets. Evaluation criteria for mapping species presence included specificity, sensitivity, Kappa, and area under the curve (AUC). Evaluation criteria for the continuous basal area variables included correlation and relative mean squared error. For predicting species presence (setting thresholds to maximize Kappa), SGB had higher values for the majority of the species for specificity and Kappa, while GAMs had higher values for the majority of the species for sensitivity. In evaluating resultant AUC values, GAM and/or SGB models had significantly better results than the See5 models where significant differences could be detected between models. For nine out of 13 species, basal area prediction results for all modelling techniques were poor (correlations less than 0.5 and relative mean squared errors greater than 0.8), but SGB provided the most stable predictions in these instances. SGB and Cubist performed equally well for modelling basal area for three species with moderate prediction success, while all three modelling tools produced comparably good predictions (correlation of 0.68 and relative mean squared error of 0.56) for one species. ?? 2006 Elsevier B.V. All rights reserved.

Ecological Modelling↗
USGS2008

Resolving model parameter values from carbon and nitrogen stock measurements in a wide range of tropical mature forests using nonlinear inversion and regression trees

Objectively assessing the performance of a model and deriving model parameter values from observations are critical and challenging in landscape to regional modeling. In this paper, we applied a nonlinear inversion technique to calibrate the ecosystem model CENTURY against carbon (C) and nitrogen (N) stock measurements collected from 39 mature tropical forest sites in seven life zones in Costa Rica. Net primary productivity from the Moderate-Resolution Imaging Spectroradiometer (MODIS), C and N stocks in aboveground live biomass, litter, coarse woody debris (CWD), and in soils were used to calibrate the model. To investigate the resolution of available observations on the number of adjustable parameters, inversion was performed using nine setups of adjustable parameters. Statistics including observation sensitivity, parameter correlation coefficient, parameter sensitivity, and parameter confidence limits were used to evaluate the information content of observations, resolution of model parameters, and overall model performance. Results indicated that soil organic carbon content, soil nitrogen content, and total aboveground biomass carbon had the highest information contents, while measurements of carbon in litter and nitrogen in CWD contributed little to the parameter estimation processes. The available information could resolve the values of 2-4 parameters. Adjusting just one parameter resulted in under-fitting and unacceptable model performance, while adjusting five parameters simultaneously led to over-fitting. Results further indicated that the MODIS NPP values were compressed as compared with the spatial variability of net primary production (NPP) values inferred from inverse modeling. Using inverse modeling to infer NPP and other sensitive model parameters from C and N stock observations provides an opportunity to utilize data collected by national to regional forest inventory systems to reduce the uncertainties in the carbon cycle and generate valuable databases to validate and improve MODIS NPP algorithms.

Ecological Modelling↗
USGS2011

Resolving structural uncertainty in natural resources management using POMDP approaches

In recent years there has been a growing focus on the uncertainties of natural resources management, and the importance of accounting for uncertainty in assessing management effectiveness. This paper focuses on uncertainty in resource management in terms of discrete-state Markov decision processes (MDP) under structural uncertainty and partial observability. It describes the treatment of structural uncertainty with approaches developed for partially observable resource systems. In particular, I show how value iteration for partially observable MDPs (POMDP) can be extended to structurally uncertain MDPs. A key difference between these process classes is that structurally uncertain MDPs require the tracking of system state as well as a probability structure for the structure uncertainty, whereas with POMDPs require only a probability structure for the observation uncertainty. The added complexity of the optimization problem under structural uncertainty is compensated by reduced dimensionality in the search for optimal strategy. A solution algorithm for structurally uncertain processes is outlined for a simple example in conservation biology. By building on the conceptual framework developed for POMDPs, natural resource analysts and decision makers who confront structural uncertainties in natural resources can take advantage of the rapid growth in POMDP methods and approaches, and thereby produce better conservation strategies over a larger class of resource problems.

Ecological Modelling↗