Identifying tropical dry forests extent and succession via the use of machine learning techniques

Information on ecosystem services as a function of the successional stage for secondary tropical dry forests (TDFs) is scarce and limited. Secondary TDFs succession is defined as regrowth following a complete forest clearance for cattle growth or agriculture activities. In the context of large conservation initiatives, the identification of the extent, structure and composition of secondary TDFs can serve as key elements to estimate the effectiveness of such activities. As such, in this study we evaluate the use of a Hyperspectral MAPper (HyMap) dataset and a waveform LIDAR dataset for characterization of different levels of intra-secondary forests stages at the Santa Rosa National Park (SRNP) Environmental Monitoring Super Site located in Costa Rica. Specifically, a multi-task learning based machine learning classifier (MLC-MTL) is employed on the first shortwave infrared (SWIR1) of HyMap in order to identify the variability of aboveground biomass of secondary TDFs along a successional gradient. Our paper recognizes that the process of ecological succession is not deterministic but a combination of transitional forests types along a stochastic path that depends on ecological, edaphic, land use, and micro-meteorological conditions, and our results provide a new way to obtain the spatial distribution of three main types of TDFs successional stages.     

Above ground biomass and tree species richness estimation with airborne lidar in tropical Ghana forests

Estimates of forest aboveground biomass are fundamental for carbon monitoring and accounting; delivering information at very high spatial resolution is especially valuable for local management, conservation and selective logging purposes. In tropical areas, hosting large biomass and biodiversity resources which are often threatened by unsustainable anthropogenic pressures, frequent forest resources monitoring is needed. Lidar is a powerful tool to estimate aboveground biomass at fine resolution; however its application in tropical forests has been limited, with high variability in the accuracy of results. Lidar pulses scan the forest vertical profile, and can provide structure information which is also linked to biodiversity. In the last decade the remote sensing of biodiversity has received great attention, but few studies focused on the use of lidar for assessing tree species richness in tropical forests.This research aims at estimating aboveground biomass and tree species richness using discrete return airborne lidar in Ghana forests. We tested an advanced statistical technique, Multivariate Adaptive Regression Splines (MARS), which does not require assumptions on data distribution or on the relationships between variables, being suitable for studying ecological variables.We compared the MARS regression results with those obtained by multilinear regression and found that both algorithms were effective, but MARS provided higher accuracy either for biomass (R² = 0.72) and species richness (R² = 0.64). We also noted strong correlation between biodiversity and biomass field values. Even if the forest areas under analysis are limited in extent and represent peculiar ecosystems, the preliminary indications produced by our study suggest that instrument such as lidar, specifically useful for pinpointing forest structure, can also be exploited as a support for tree species richness assessment.     

Possibilities of discriminating tropical secondary succession in Amazônia using hyperspectral and multiangular CHRIS/PROBA data

CHRIS/PROBA data collected in the Brazilian Amazônia in 4 view angles (−36°, nadir, +36°, +55°) and 62 bands (410–1000 nm range) were evaluated for the discrimination between primary forest and 3 stages of secondary succession after deforestation: initial (SS1; <5 years), intermediate (SS2; 5–15 years) and advanced (SS3; >15 years). Single view angle and multiangular approaches (nadir plus anisotropic information derived from reflectance ratios between view angles) were tested for discrimination. Both approaches used principal components analysis (PCA) applied to pixel spectra representative of each class in order to reduce data dimensionality at each dataset, to enhance separability between the classes, and to provide input variables for multiple discriminant analysis (MDA). The results showed that the off-nadir viewing improved discrimination between the successional stages. Discrimination between SS2 and SS3 was enhanced with PCA at +36° view angle. Primary forest and SS3 presented a more anisotropic behavior than SS2 and SS1, especially in the backward scattering direction (positive view angles) in which great amounts of sunlit canopy components were viewed by the sensor. MDA classification results showed that the multiangular approach produced an overall improvement in the discrimination. From the single (nadir) to the multiangular approach, classification accuracy using a separate set of pixels increased from 83.3% to 98.3% for SS1, 53.3% to 70.0% for SS2, and 58.3% to 76.7% for SS3. The nadir and multiangular classifications were statistically different at a 0.05% level of significance. Kappa statistics increased from 0.63 to 0.82. The results showed that multiangular data can improve the differentiation between primary forest and old stages of natural vegetation regrowth, which have been reported in the literature as the most difficult classes to be mapped in the Amazonian environment.     

Assessing the influence of return density on estimation of lidar-based aboveground biomass in tropical peat swamp forests of Kalimantan,Indonesia

The airborne lidar system (ALS) provides a means to efficiently monitor the status of remote tropical forests and continues to be the subject of intense evaluation. However, the cost of ALS acquisition can vary significantly depending on the acquisition parameters, particularly the return density (i.e., spatial resolution) of the lidar point cloud. This study assessed the effect of lidar return density on the accuracy of lidar metrics and regression models for estimating aboveground biomass (AGB) and basal area (BA) in tropical peat swamp forests (PSF) in Kalimantan, Indonesia. A large dataset of ALS covering an area of 123,000 ha was used in this study. This study found that cumulative return proportion (CRP) variables represent a better accumulation of AGB over tree heights than height-related variables. The CRP variables in power models explained 80.9% and 90.9% of the BA and AGB variations, respectively. Further, it was found that low-density (and low-cost) lidar should be considered as a feasible option for assessing AGB and BA in vast areas of flat, lowland PSF. The performance of the models generated using reduced return densities as low as 1/9 returns per m² also yielded strong agreement with the original high-density data. The use model-based statistical inferences enabled relatively precise estimates of the mean AGB at the landscape scale to be obtained with a fairly low-density of 1/4 returns per m², with less than 10% standard error (SE). Further, even when very low-density lidar data was used (i.e., 1/49 returns per m²) the bias of the mean AGB estimates were still less than 10% with a SE of approximately 15%. This study also investigated the influence of different DTM resolutions for normalizing the elevation during the generation of forest-related lidar metrics using various return densities point cloud. We found that the high-resolution digital terrain model (DTM) had little effect on the accuracy of lidar metrics calculation in PSF. The accuracy of low-density lidar metrics in PSF was more influenced by the density of aboveground returns, rather than the last return. This is due to the flat topography of the study area. The results of this study will be valuable for future economical and feasible assessments of forest metrics over large areas of tropical peat swamp ecosystems.     

A comprehensive but efficient framework of proposing and validating feature parameters from airborne LiDAR data for tree species classification

Tree species information is crucial for digital forestry, and efficient techniques for classifying tree species are extensively demanded. To this end, airborne light detection and ranging (LiDAR) has been introduced. However, the literature review suggests that most of the previous airborne LiDAR-based studies were only based on limited kinds of tree signatures. To address this gap, this study proposed developing a novel modular framework for LiDAR-based tree species classification, by deriving feature parameters in a systematic way. Specifically, feature parameters of point-distribution (PD), laser pulse intensity (IN), crown-internal (CI) and tree-external (TE) structures were proposed and derived. With a support-vector-machine (SVM) classifier used, the classifications were conducted in a leave-one-out-for-cross-validation (LOOCV) mode. Based on the samples of four typical boreal tree species, i.e., Picea abies, Pinus sylvestris, Populus tremula and Quercus robur, tests showed that the accuracies of the classifications based on the acquired PD-, IN-, CI- and TE-categorized feature parameters as well as the integration of their individual optimal parameters are 65.00%, 80.00%, 82.50%, 85.00% and 92.50%, respectively. These results indicate that the procedures proposed in this study can be used as a comprehensive but efficient framework of proposing and validating feature parameters from airborne LiDAR data for tree species classification.     

Assessing the potential for leaf-off LiDAR data to model canopy closure in temperate deciduous forests

Estimates of canopy closure have many important uses in forest management and ecological research. Field measurements, however, are typically not practical to acquire over expansive areas or for large numbers of locations. This problem has been addressed, in recent years, through the use of airborne light detection and ranging (LiDAR) technology which has proven effective in modeling canopy closure remotely. The techniques developed to use LiDAR for this purpose have been designed and evaluated for datasets acquired during leaf-on conditions. However, a large number of LiDAR datasets are acquired during leaf-off conditions since their primary purpose is to generate bare-earth Digital Elevation Models. In this paper, we develop and evaluate techniques for leveraging small-footprint leaf-off LiDAR data to model leaf-on canopy closure in temperate deciduous forests.We evaluate three techniques for modeling canopy closure: (1) the canopy-to-total-return-ratio (CTRR), (2) the canopy-to-total-pixel-ratio (CTPR), and (3) the hemispherical-viewshed (HV). The first technique has been used widely, in various forms, and has been shown to be effective with leaf-on LiDAR datasets. The CTRR technique that we tested uses the first-return LiDAR data only. The latter two techniques are new contributions that we develop and present in this paper. These techniques use Canopy Height Models (CHM) to detect significant gaps in the forest canopy which are of primary importance in estimating closure.The techniques we tested each showed good promise for predicting canopy closure using leaf-off LiDAR data with the CTPR and HV models having particularly high correlations with closure estimates from hemispherical photographs. The CTRR model had performance on par with results from previous studies that used leaf-on LiDAR, although, with leaf-off data the model tended to be negatively biased with respect to species having simple and compound leaf types and positively biased for coniferous species. The CTPR and HV models also showed some slight negative biases for compound-leaf species. The biases for the CTPR and HV models were mitigated when the CHM data were smoothed to fill in small gaps. The CHM-based models were robust to changes in the CHM model resolution which suggests that these methods may be applicable to a variety of small-footprint LiDAR datasets. In this research, the new CTPR and HV methods showed a strong ability to predict canopy closure using leaf-off data, however, future work will be needed to test the applicability of the models to variations in LiDAR datasets, forest types, and topography.     

A physically based approach to model LAI from MODIS 250 m data in a tropical region

A time series of leaf area index (LAI) has been developed based on 16-day normalized difference vegetation index (NDVI) data from the Moderate Resolution Imaging Spectroradiometer (MODIS) at 250 m resolution (MOD250_LAI). The MOD250_LAI product uses a physical radiative transfer model which establishes a relationship between LAI, fraction of vegetation cover (FVC) and given patterns of surface reflectance, view-illumination conditions and optical properties of vegetation. In situ measurements of LAI and FVC made at 166 plots using hemispherical photography served for calibration of model parameters and validation of modelling results. Optical properties of vegetation cover, summarized by the light extinction coefficient, were computed at the local (pixel) level based on empirical models between ground-measured tree crown architecture at 85 sampling plots and spectral values in Landsat ETM+ bands. Influence of view-illumination conditions on optical properties of canopy was simulated by a view angle geometry model incorporating the solar zenith angle and the sensor viewing angle. The results revealed high compatibility of the produced MOD250_LAI data set with ground truth information and the 30 m resolution Landsat ETM+ LAI estimated using the similar algorithm. The produced MOD250_LAI was also compared with the global MODIS 1000-m LAI product (MOD15A2 LAI). Results show good consistency of the spatial distribution and temporal dynamics between the two LAI products. However, the results also showed that the annual LAI amplitude by the MOD15A2 product is significantly higher than by the MOD250_LAI. This higher amplitude is caused by a considerable underestimation of the tropical rainforest LAI by the MOD15A2 during the seasonal phases of low leaf production.     

Estimation of aboveground biomass and carbon in a tropical rain forest in Gabon using remote sensing and GPS data

The knowledge of biomass stocks in tropical forests is critical for climate change and ecosystem services studies. This research was conducted in a tropical rain forest located near the city of Libreville (the capital of Gabon), in the Akanda Peninsula. The forest cover was stratified in terms of mature, secondary and mangrove forests using Landsat-ETM data. A field inventory was conducted to measure the required basic forest parameters and estimate the aboveground biomass (AGB) and carbon over the different forest classes. The Shuttle Radar Topography Mission (SRTM) data were used in combination with ground-based GPS measurements to derive forest heights. Finally, the relationships between the estimated heights and AGB were established and validated. Highest biomass stocks were found in the mature stands (223 ± 37 MgC/ha), followed by the secondary forests (116 ± 17 MgC/ha) and finally the mangrove forests (36 ± 19 MgC/ha). Strong relationships were found between AGB and forest heights (R² > 0.85).     

Carbon sinks and tropical forest biomass estimation: a review on role of remote sensing in aboveground-biomass modelling

Tropical forest embraces a large stock of carbon and contributes to the enormous amount of above- and below-ground biomass and the global carbon cycle. The carbon kept in the above-ground living biomass of trees is typically the largest pool and the most directly impacted by deforestation and degradation. Hence, quantifying carbon stock and fluxes from tropical forests by estimating the above-ground forest biomass is the critical step that will be investigated further in this paper. Remote sensing technology can provide many advantages in quantifying and mapping forest structure and monitoring and mapping above-ground biomass, and is both temporally and spatially accurate. Therefore, a good data-set of biomass which comprises canopy height and canopy structure can provide carbon sequestration potential for forest reserves. This paper reviews a thorough research of biomass estimation using remote sensing and geospatial technologies.     

A hybrid framework for single tree detection from airborne laser scanning data: A case study in temperate mature coniferous forests in Ontario,Canada

This study presents a hybrid framework for single tree detection from airborne laser scanning (ALS) data by integrating low-level image processing techniques into a high-level probabilistic framework. The proposed approach modeled tree crowns in a forest plot as a configuration of circular objects. We took advantage of low-level image processing techniques to generate candidate configurations from the canopy height model (CHM): the treetop positions were sampled within the over-extracted local maxima via local maxima filtering, and the crown sizes were derived from marker-controlled watershed segmentation using corresponding treetops as markers. The configuration containing the best possible set of detected tree objects was estimated by a global optimization solver. To achieve this, we introduced a Gibbs energy, which contains a data term that judges the fitness of the objects with respect to the data, and a prior term that prevents severe overlapping between tree crowns on the configuration space. The energy was then embedded into a Markov Chain Monte Carlo (MCMC) dynamics coupled with a simulated annealing to find its global minimum. In this research, we also proposed a Monte Carlo-based sampling method for parameter estimation. We tested the method on a temperate mature coniferous forest in Ontario, Canada and also on simulated coniferous forest plots with different degrees of crown overlap. The experimental results showed the effectiveness of our proposed method, which was capable of reducing the commission errors produced by local maxima filtering, thus increasing the overall detection accuracy by approximately 10% on all of the datasets.     

Development and implementation of a multiscale biomass model using hyperspectral vegetation indices for winter wheat in the North China Plain

Crop monitoring during the growing season is important for regional management decisions and biomass prediction. The objectives of this study were to develop, improve and validate a scale independent biomass model. Field studies were conducted in Huimin County, Shandong Province of China, during the 2006–2007 growing season of winter wheat (Triticum aestivum L.). The field design had a multiscale set-up with four levels which differed in their management, such as nitrogen fertilizer inputs and cultivars, to create different biomass conditions: small experimental fields (L1), large experimental fields (L2), small farm fields (L3), and large farm fields (L4). L4, planted with different winter wheat varieties, was managed according to farmers’ practice while L1 through L3 represented controlled field experiments. Multitemporal spectral measurements were taken in the fields, and biomass was sampled for each spectral campaign. In addition, multitemporal Hyperion data were obtained in 2006 and 2007. L1 field data were used to develop biomass models based on the relation between the winter wheat spectra and biomass: several published vegetation indices, including NRI, REP, OSAVI, TCI, and NDVI, were investigated. A new hyperspectral vegetation index, which uses a four-band combination in the NIR and SWIR domains, named GnyLi, was developed. Following the multiscale concept, the data of higher levels (L2 through L4) were used stepwise to validate and improve the models of the lower levels, and to transfer the improved models to the next level. Lastly, the models were transferred and validated at the regional scale using Hyperion images of 2006 and 2007. The results showed that the GnyLi and NRI models, which were based on the NIR and SWIR domains, performed best with R² > 0.74. All the other indices explained less than 60% model variability. Using the Hyperion data for regionalization, GnyLi and NRI explained 81–89% of the biomass variability. These results highlighted that GnyLi and NRI can be used together with hyperspectral images for both plot and regional level biomass estimation. Nevertheless, additional studies and analyses are needed to test its replicability in other environmental conditions.     

Biomass allometric equation and expansion factor for a mountain moist evergreen forest in Mozambique

Background

Worldwide, forests are an important carbon sink and thus are key to mitigate the effects of climate change. Mountain moist evergreen forests in Mozambique are threatened by agricultural expansion, uncontrolled logging, and firewood collection, thus compromising their role in carbon sequestration. There is lack of local tools for above-ground biomass (AGB) estimation of mountain moist evergreen forest, hence carbon emissions from deforestation and forest degradation are not adequately known. This study aimed to develop biomass allometric equations (BAE) and biomass expansion factor (BEF) for the estimation of total above-ground carbon stock in mountain moist evergreen forest.

Methods

The destructive method was used, whereby 39 trees were felled and measured for diameter at breast height (DBH), total height and the commercial height. We determined the wood basic density, the total dry weight and merchantable timber volume by Smalian’s formula. Six biomass allometric models were fitted using non-linear least square regression. The BEF was determined based on the relationship between bole stem dry weight and total dry weight of the tree. To estimate the mean AGB of the forest, a forest inventory was conducted using 27 temporary square plots. The applicability of Marzoli’s volume equation was compared with Smalian’s volume equation in order to check whether Marzoli’s volume from national forest inventory can be used to predict AGB using BEF.

Results

The best model was the power model with only DBH as predictor variable, which provided an estimated mean AGB of 291?±?141 Mg ha^?1 (mean?±?95% confidence level). The mean wood basic density of sampled trees was 0.715?±?0.182 g cm^?3. The average BEF was of 2.05?±?0.15 and the estimated mean AGB of 387?±?126 Mg ha^?1. The BAE from miombo woodland within the vicinity of the study area underestimates the AGB for all sampled trees. Chave et al.’s pantropical equation of moist forest did not fit to the Moribane Forest Reserve, while Brown’s equation of moist forest had a good fit to the Moribane Forest Reserve, having generated 1.2% of bias, very close to that generated by the selected model of this study. BEF showed to be reliable when combined with stand mean volume from Marzoli’s National Forestry Inventory equation.

Conclusion

The BAE and the BEF function developed in this study can be used to estimate the AGB of the mountain moist evergreen forests at Moribane Forest Reserve in Mozambique. However, the use of the biomass allometric model should be preferable when DBH information is available.

     

A sampling approach to estimate the log determinant used in spatial likelihood problems

Likelihood-based methods for modeling multivariate Gaussian spatial data have desirable statistical characteristics, but the practicality of these methods for massive georeferenced data sets is often questioned. A sampling algorithm is proposed that exploits a relationship involving log-pivots arising from matrix decompositions used to compute the log determinant term that appears in the model likelihood. We demonstrate that the method can be used to successfully estimate log-determinants for large numbers of observations. Specifically, we produce an log-determinant estimate for a 3,954,400 by 3,954,400 matrix in less than two minutes on a desktop computer. The proposed method involves computations that are independent, making it amenable to out-of-core computation as well as to coarse-grained parallel or distributed processing. The proposed technique yields an estimated log-determinant and associated confidence interval.

A general framework for error analysis in measurement-based GIS Part 1: The basic measurement-error model and related concepts

This is the first of a four-part series of papers which proposes a general framework for error analysis in measurement-based geographical information systems (MBGIS). The purpose of the series is to investigate the fundamental issues involved in measurement error (ME) analysis in MBGIS, and to provide a unified and effective treatment of errors and their propagation in various interrelated GIS and spatial operations. Part 1 deals with the formulation of the basic ME model together with the law of error propagation. Part 2 investigates the classic point-in-polygon problem under ME. Continuing to Part 3 is the analysis of ME in intersections and polygon overlays. In Part 4, error analyses in length and area measurements are made. In this present part, a simple but general model for ME in MBGIS is introduced. An approximate law of error propagation is then formulated. A simple, unified, and effective treatment of error bands for a line segment is made under the name of covariance-based error band. A new concept, called maximal allowable limit, which guarantees invariance in topology or geometric-property of a polygon under ME is also advanced. To handle errors in indirect measurements, a geodetic model for MBGIS is proposed and its error propagation problem is studied on the basis of the basic ME model as well as the approximate law of error propagation. Simulation experiments all substantiate the effectiveness of the proposed theoretical construct.This project was supported by the earmarked grant CUHK 4362/00H of the Hong Kong Research grants Council.     

A general framework for error analysis in measurement-based GIS Part 2: The algebra-based probability model for point-in-polygon analysis

This is the second paper of a four-part series of papers on the development of a general framework for error analysis in measurement-based geographic information systems (MBGIS). In this paper, we discuss the problem of point-in-polygon analysis under randomness, i.e., with random measurement error (ME). It is well known that overlay is one of the most important operations in GIS, and point-in-polygon analysis is a basic class of overlay and query problems. Though it is a classic problem, it has, however, not been addressed appropriately. With ME in the location of the vertices of a polygon, the resulting random polygons may undergo complex changes, so that the point-in-polygon problem may become theoretically and practically ill-defined. That is, there is a possibility that we cannot answer whether a random point is inside a random polygon if the polygon is not simple and cannot form a region. For the point-in-triangle problem, however, such a case need not be considered since any triangle always forms an interior or region. To formulate the general point-in-polygon problem in a suitable way, a conditional probability mechanism is first introduced in order to accurately characterize the nature of the problem and establish the basis for further analysis. For the point-in-triangle problem, four quadratic forms in the joint coordinate vectors of a point and the vertices of the triangle are constructed. The probability model for the point-in-triangle problem is then established by the identification of signs of these quadratic form variables. Our basic idea for solving a general point-in-polygon (concave or convex) problem is to convert it into several point-in-triangle problems under a certain condition. By solving each point-in-triangle problem and summing the solutions, the probability model for a general point-in-polygon analysis is constructed. The simplicity of the algebra-based approach is that from using these quadratic forms, we can circumvent the complex geometrical relations between a random point and a random polygon (convex or concave) that one has to deal with in any geometric method when probability is computed. The theoretical arguments are substantiated by simulation experiments.This project was supported by the earmarked grant CUHK 4362/00H of the Hong Kong Research grants Council.     

生物量估测模型中遥感信息与植被光合参数的关系研究

本文通过对建立估测植被生物量遥感模型中所涉及的遥感信息参数与植被光合参数的关系分析,从理论和实验中阐明了反映植物长势的量发化植被指数和反映植物光合面积的叶面积指数,光合有效辐射及吸收光合有效辐射的相互关系,对在实际中建立更为机理的生物量遥感模型提供可供进一步参考的依据。     

A computational framework for generalized moving windows and its application to landscape pattern analysis

Land cover products based on remotely sensed data are commonly investigated in terms of landscape composition and configuration; i.e. landscape pattern. Traditional landscape pattern indicators summarize an aspect of landscape pattern over the full study area. Increasingly, the advantages of representing the scale-specific spatial variation of landscape patterns as continuous surfaces are being recognized. However, technical and computational barriers hinder the uptake of this approach. This article reduces such barriers by introducing a computational framework for moving window analysis that separates the tasks of tallying pixels, patches and edges as a window moves over the map from the internal logic of landscape indicators. The framework is applied on data covering the UK and Ireland at 250 m resolution, evaluating a variety of indicators including mean patch size, edge density and Shannon diversity at window sizes ranging from 2.5 km to 80 km. The required computation time is in the order of seconds to minutes on a regular personal computer. The framework supports rapid development of indicators requiring little coding. The computational efficiency means that methods can be integrated in iterative computational tasks such as multi-scale analysis, optimization, sensitivity analysis and simulation modelling.     

Model selection in Bayesian framework to identify the best WorldView-2 based vegetation index in predicting green biomass of salt marshes in the northern Gulf of Mexico

Coastal wetlands are among the most productive ecosystems globally but have experienced dramatic degradation and loss within the past several decades. Vegetation biomass of coastal wetlands is not only the key component of blue carbon storage but also plays an important role in vertical accretion, important for maintaining these habitats under relative sea-level rise. Remote sensing offers a cost-effective approach to study vegetation biomass at a broad spatial scale. We developed statistical models to predict peak aboveground green biomass of Spartina alterniflora and Juncus roemerianus, two dominant species of salt marshes using WorldView-2 satellite imagery at the Grand Bay National Estuarine Research Reserve (NERR) on the Mississippi coast in the northern Gulf of Mexico. The model accounted for nested data structures in the sampled biomass, assimilated uncertainties from data, parameters and model structures, and helped determine the best vegetation index among a variety of commonly-used indices to predict aboveground green biomass. We developed a series of mixed-effects models, which included different combinations of fixed effect(s), random intercept, and random slope(s). The fixed effects were species and one of the 60 vegetation indices derived from a WorldView-2 image obtained on 6 October 2012. The random effect used was site. We implemented the models in a Bayesian framework and selected the best model structure and vegetation index based on minimum posterior predictive loss and deviance information criterion. The results showed that the best vegetation index to predict peak green biomass was the green chlorophyll index derived from the reflectance values of band 8 (near-infrared) and band 3 (green), and its effect on biomass prediction varied among sites. The inclusion of species as a fixed effect improved the model prediction. The study demonstrated the need to account for spatial dependence of data in developing a robust model, and the importance of the second WorldView-2 near-infrared band (860–1040 nm) in predicting aboveground green biomass for the Grand Bay NERR. The analysis using mixed-effects modeling in Bayesian inference which coherently combined field and WorldView-2 data with uncertainties accounted for provides a robust and nondestructive tool for resource managers to monitor the status of coastal wetlands at a high spatial resolution in a timely manner. Through this study, we hope to emphasize the importance of appropriately accounting for nested data structures using mixed-effects models and promote wider application of Bayesian inference to facilitate assimilation of uncertainties in remote sensing applications.     

A comparison of geographic datasets and field measurements to model soil carbon using random forests and stepwise regressions (British Columbia,Canada)

We used geographic datasets and field measurements to examine the mechanisms that affect soil carbon (SC) storage for 65 grazed and non-grazed pastures in southern interior grasslands of British Columbia, Canada. Stepwise linear regression (SR) modeling was compared with random forest (RF) modeling. Models produced with SR performed better than those produced using RF models (r² = 0.56–0.77 AIC = 0.16–0.30 for SR models; r² = 0.38–0.53 and AIC = 0.18–0.30 for RF models). The factors most significant when predicting SC were elevation, precipitation, and the normalized difference vegetation index (NDVI). NDVI was evaluated at two scales using: (1) the MOD 13Q1 (250 m/16-day resolution) NDVI data product from the moderate resolution imaging spectro-radiometer (MODIS) (NDVI_MODIS), and (2) a handheld multispectral radiometer (MSR, 1 m resolution) (NDVI_MSR) in order to understand the potential for increasing model accuracy by increasing the spatial resolution of the gridded geographic datasets. When NDVI_MSR data were used to predict SC, the percentage of the variance explained by the model was greater than for models that relied on NDVI_MODIS data (r² = 0.68 for SC for non-grazed systems, modeled with SR based on NDVI_MODIS data; r² = 0.77 for SC for non-grazed systems, modeled with SR based on NDVI_MSR data). The outcomes of this study provide the groundwork for effective monitoring of SC using geographic datasets to enable a carbon offset program for the ranching industry.