Forest Canopy Cover Research Articles

Forest cover mediates genetic connectivity of northwestern cougars

Population structure, connectivity, and dispersal success of individuals can be challenging to demonstrate for solitary carnivores with low population densities. Though the cougar (Puma concolor) is widely distributed throughout North America and is capable of dispersing long distances, populations can be geographically structured and genetic isolation has been documented in some small populations. We described genetic structure and explored the relationship between landscape resistance and genetic variation in cougars in Washington and southern British Columbia using allele frequencies of 17 microsatellite loci for felids. We evaluated population structure of cougars using the Geneland clustering algorithm and spatial principal components analysis. We then used Circuitscape to estimate the landscape resistance between pairs of individuals based on rescaled GIS layers for forest canopy cover, elevation, human population density and highways. We quantified the effect of landscape resistance on genetic distance using multiple regression on distance matrices and boosted regression tree analysis. Cluster analysis identified four populations in the study area. Multiple regression on distance matrices and boosted regression tree models indicated that only forest canopy cover and geographic distance between individuals had an effect on genetic distance. The boundaries between genetic clusters largely corresponded with breaks in forest cover, showing agreement between population structure and genetic gradient analyses. Our data indicate that forest cover promotes gene flow for cougars in the Pacific Northwest, which provides insight managers can use to preserve or enhance genetic connectivity.

The role of deterministic factors and stochasticity on the trophic interactions between birds and fish in temporary floodplain ponds

Trophic interaction between contiguous habitats is an understudied area despite their ecological importance. We tested four hypotheses related to the visit of foraging birds in temporary ponds inhabited by fish: the local characteristics of ponds influence the (1) abundance and (2) composition of foraging birds; (3) the number of foraging bird visits increases in higher fish abundance; and (4) pond characteristics lead to a non-random spatial structure of fish assemblage. We studied 18 temporary ponds in the floodplain area of Pantanal (center-west Brazil), where we measured environmental variables (pond size, depth, macrophyte coverage, and forest canopy coverage), recorded the number of foraging birds and sampled fish. Foraging birds’ abundance and composition were mainly influenced by forest canopy coverage and pond size, corroborating our first and second hypotheses. The hypotheses 3 and 4 were rejected. Fish abundance was not correlated with higher number of visits of foraging birds and fish distribution was random. Local environmental variables did not affect significantly fish richness and composition. Our results suggest that the abundance of foraging birds and fish presence are determined by different assembly processes (deterministic vs. random), which may limit ponds selection by birds due to the uncertainty in fish distribution and ephemeral nature of temporary ponds.

Modern surface pollen assemblages from the Middle and High Atlas, Morocco: insights into pollen representation and transport

Thirty-three modern surface samples were collected in the environmentally and climatologically contrasting regions of the Middle and High Atlas Mountains, Morocco. Samples representing forest and steppe montane environments (1935–2760 m above sea level) are clustered around study sites at Lake Tislit (High Atlas, semi-arid oro-Mediterranean bioclime) and Lake Sidi Ali and Michliffen (Middle Atlas, sub-humid montane Mediterranean bioclime). Good discrimination between regional pollen spectra is evident, with Middle Atlas samples reflecting higher arboreal cover (Cedrus and evergreen Quercus) and High Atlas samples with high abundances of non-arboreal taxa, including Artemisia and Fabaceae. These four taxa (Cedrus, evergreen Quercus, Artemisia and Fabaceae) are furthermore shown to be reliable indicators of local source vegetation within a 100 m2 quadrat, taking into account threshold abundances of 7%, 20%, 4% and 10%, respectively. Deciduous Quercus, Olea and Phillyrea show long-distance pollen dispersal across both regions, contributing to non-trivial arboreal pollen (AP) values of up to 35% (typically 20–30%) in the High Atlas spectra. In the Middle Atlas, AP values of 40 to 50% occur in open sampling locations and > 60% under forest canopy cover. These insights should be taken into account when interpreting ancient pollen spectra from regional lakes and bogs for palaeoenvironmental reconstruction.

Rapid forest carbon assessments of oceanic islands: a case study of the Hawaiian archipelago

BackgroundSpatially explicit forest carbon (C) monitoring aids conservation and climate change mitigation efforts, yet few approaches have been developed specifically for the highly heterogeneous landscapes of oceanic island chains that continue to undergo rapid and extensive forest C change. We developed an approach for rapid mapping of aboveground C density (ACD; units = Mg or metric tons C ha−1) on islands at a spatial resolution of 30 m (0.09 ha) using a combination of cost-effective airborne LiDAR data and full-coverage satellite data. We used the approach to map forest ACD across the main Hawaiian Islands, comparing C stocks within and among islands, in protected and unprotected areas, and among forests dominated by native and invasive species.ResultsTotal forest aboveground C stock of the Hawaiian Islands was 36 Tg, and ACD distributions were extremely heterogeneous both within and across islands. Remotely sensed ACD was validated against U.S. Forest Service FIA plot inventory data (R2 = 0.67; RMSE = 30.4 Mg C ha−1). Geospatial analyses indicated the critical importance of forest type and canopy cover as predictors of mapped ACD patterns. Protection status was a strong determinant of forest C stock and density, but we found complex environmentally mediated responses of forest ACD to alien plant invasion.ConclusionsA combination of one-time airborne LiDAR data acquisition and satellite monitoring provides effective forest C mapping in the highly heterogeneous landscapes of the Hawaiian Islands. Our statistical approach yielded key insights into the drivers of ACD variation, and also makes possible future assessments of C storage change, derived on a repeat basis from free satellite data, without the need for additional LiDAR data. Changes in C stocks and densities of oceanic islands can thus be continually assessed in the face of rapid environmental changes such as biological invasions, drought, fire and land use. Such forest monitoring information can be used to promote sustainable forest use and conservation on islands in the future.Electronic supplementary materialThe online version of this article (doi:10.1186/s13021-015-0043-4) contains supplementary material, which is available to authorized users.

Thirty Year Analysis of Forest and Scrub Canopy Cover on the Big Sur Coast of California using Landsat Imagery

The potential for negative effects such as habitat loss and high severity wildfires on endemic and threatened biological resources on the United States Pacific coast from recent climate warming has received increasing attention. This study was designed to quantify, for the first time, variations in forest and coastal bluff scrub canopy density and related growth rates using 30 consecutive years of April-to-June Landsat satellite image data for the central California coast around Big Sur. A trend model was fit to multi-year Landsat NDVI time series using least squares non-linear regression on time. Analysis showed that annual growth rates for coast redwood stands in the Big Sur area, the southern-most range for Sequoia sempervirens on the Pacific coast, have been depressed by periods of drought that exceeded -1.2 standard deviations of long-term annual precipitation. There was no evidence from the Landsat record that recent large wildfires have affected region-wide coast redwood stand structure or growth rates. Nonetheless, extensive wildfires of 2008 did have a marked (albeit temporary) impact on canopy density of the region’s mixed hardwood and Coulter pine forest zones, equivalent to the extreme drought effects of 1990-1991 and 2013-2014 on these higher elevation forest zones.

A comparison of two sampling approaches for assessing the urban forest canopy cover from aerial photography

Two different sampling approaches for estimating urban tree canopy cover were applied to two medium-sized cities in the United States, in conjunction with two freely available remotely sensed imagery products. A random point-based sampling approach, which involved 1000 sample points, was compared against a plot/grid sampling (cluster sampling) approach that involved a 1.83m square grid of points embedded within 0.04ha circular plots. The imagery products included aerial photography from the U.S. Department of Agriculture National Agricultural Imagery Program (viewed within ArcGIS), and Google Earth imagery. For Tallahassee, Florida, the estimate of tree canopy cover was 48.6–49.1% using Google Earth imagery and 44.5–45.1% using NAIP imagery within ArcGIS. Statistical tests suggested that the two sampling approaches produced significantly different estimates using the two different imagery sources. For Tacoma, Washington, the estimated tree canopy cover was about 19.2–20.0% using Google Earth imagery and 17.3–18.1% when using NAIP imagery in ArcGIS. Here, there seemed to be no significant difference between the random point-based sampling efforts when used with the two different image sources, while the opposite was true when using the plot/grid sampling approach. However, our findings showed some similarities between the two sampling approaches; hence, the random point-based sampling approach might be preferred due to the time and effort required, and because fewer opportunities for classification problems might arise. Continuous review of urban canopy cover estimation procedures suggested by organizations such as the Climate Action Reserve and others can provide society with information on the accuracy and effectiveness resource assessment methods employed for making wise decisions about climate change and carbon management.

Decisive environmental characteristics for woody regrowth in forest edges – Patterns along complex environmental gradients in Southern Sweden

Early and late successional stages of forest edge development were studied across long, complex environmental gradients in order to disentangle and assess the woody species composition and functional traits’ relations to abiotic and biotic environmental characteristics at site and landscape scale. Data were sampled from 78 randomly stratified locations along a 610-km railway corridor between the cities of Stockholm and Malmö, Sweden, and subsequently analysed in multiple steps of unconstrained, constrained and partial constrained ordination and variation partitioning. Different analytical methods were applied at each step and only the environmental characteristics that repeatedly showed significant impacts on woody species composition were selected for analysis of average weighted community traits to provide verification and a more detailed understanding. Following this approach, 10 environmental characteristics were identified as decisive for the woody species composition of the forest edges. The variation partitioning revealed a marked shift in the relative explanatory power of the environmental variables in relation to the succession of the forest edges. In the early successional stage, the abiotic variables related to site productivity (as reflected by field layer type and soil moisture) and climate (humidity and altitude) dominate, while in the late successional stages of forest edges more complex patterns evolved due to hierarchical and dynamic filter effects, where biotic variables reflecting structural aspects at site (forest edge profile, canopy stratification and canopy cover) and landscape (forest continuity and the shape and contrast of edges in the surrounding landscape) level were equally important as the abiotic variables. We concluded that management efforts in early succession stages of forest edges should depart from basic abiotic gradients of site fertility and moisture; and that it is crucial to incorporate vegetation structure variables at site and landscape level into long-term management planning.

Light Detection and Ranging for Implementing Water‐Oriented Forest Management in a Semiarid Sub‐Catchment (Valencia, Spain)

Diminishing rainfall interception in semiarid forests may offset reductions of available water either for ecosystem or human needs under the climate change predictions. Forest managers should be aware of this issue and develop specific hydrology‐oriented silvicultural treatments. In this sense, relating forest–water processes to detailed forest structure information, as provided by light detection and ranging (LiDAR), is essential. In this study, low‐density LiDAR data are used to characterize forest structure and throughfall water. First, LiDAR was validated as an estimator of the forest structure variables (canopy cover, canopy height, and leaf area index) in a semiarid forest, with acceptable root mean square errors (6.2% cover, 0.4 m2 m−2 LAI and 0.7 m height). Secondly, the LiDAR‐derived variables were related to the measured throughfall by means of simple and multiple regression models. The best model included canopy cover and stand height, and improved upon previous results based on field measures from r2 = 0.76–0.95. Finally, the model was used to estimate the throughfall in the whole sub‐catchment, where very low or low throughfall (<74%) was identified in 47.5% of the area. Low‐density LiDAR was proved to be useful in guiding hydrology‐oriented silviculture to increase the amount of throughfall water.

Higher Soil Water Availability after Removal of a Dominant, Nonnative Tree (Casuarina equisetifolia Forst.) from a Subtropical Forest

Abstract: The eradication of dominant, nonnative trees can alter soil conditions in forest ecosystems by changing the forest's water balance. To test this idea, we measured water in surface soil in forests dominated by an invasive nonnative tree (Casuarina equisetifolia Forst.) on Nishijima Island, a subtropical island in the Ogasawara Islands group, northwestern Pacific Ocean. The volumetric water content of surface soils at sites where all trees of C. equisetifolia were killed by herbicide was compared with adjacent invaded control sites, and effects of time since tree removal (up to 3 yr) were assessed. Dry weights of accumulated litter, total canopy openness of the forest canopy, maximum heights, and total plant cover of herbaceous vegetation were also compared between treatment and control plots. Volumetric soil water content in quadrats where C. equisetifolia trees were removed (removal area) was significantly higher than at control quadrats. We found that the effect of C. equisetifolia removal on s...

Mixed severity fire effects within the Rim fire: Relative importance of local climate, fire weather, topography, and forest structure

Recent and projected increases in the frequency and severity of large wildfires in the western U.S. makes understanding the factors that strongly affect landscape fire patterns a management priority for optimizing treatment location. We compared the influence of variations in the local environment on burn severity patterns on the large 2013 Rim fire that burned under extreme drought with those of previous smaller fires for a study area in the Sierra Nevada, California, USA. Although much of the Rim fire burned during plume-dominated conditions resulting in large high-severity patches, our study area burned under milder fire weather resulting in a mix of fire severities. In our study area the Rim fire produced a higher proportion of moderate- and high-severity effects than occurred in previous fires. Random forest modeling explained up to 63% of the Rim fire burn variance using seven predictors: time since previous fire, actual evapotranspiration (AET), climatic water deficit (Deficit), previous maximum burn severity, burning index, slope position, and solar radiation. Models using only a subset of biophysical predictors (AET, Deficit, slope position and steepness, and solar radiation) explained 55% of the Rim fire and 58% of the maximum fire burn severity of previous fires. The relationship of burn severity to patterns of AET, however, reversed for the Rim fire (positive) compared to earlier fires (negative). Measurements of pre-Rim fire forest structure from LiDAR did not improve our ability to explain burn severity patterns. We found that accounting for spatial autocorrelation in burn severity and biophysical environment was important to model quality and stability. Our results suggest water balance and topography can help predict likely burn severity patterns under moderate climate and fire weather conditions, providing managers with general guidance for prioritizing fuel treatments and identifying where fire is less likely to burn with higher severities even for locations with higher forest density and canopy cover.

Multi-scale analysis of snow dynamics at the southern margin of the North American continental snow distribution

Abstract Snow provides a key water source for stream flow and agricultural production across western North America and drinking water for large populations in the Southwest. Accurate estimates of snow cover spatial distribution and temporal dynamics are important at regional and local scales as snow cover is projected to decrease due to global climate change. We examined regional-scale temporal trends in snow distribution across central and northern Arizona using two tiles of 2928 daily images of MOD10 snow product. The analysis included the entire MODIS archive time period, October 1, 2003–June 1, 2014, and a 245,041 km2 area of 51 HUC8 watersheds. We also examined the effects of a regional forest restoration effort, known as the Four Forest Restoration Initiative (4FRI), aimed at enhancing snow accumulation and retention for increased groundwater recharge through forest thinning and burning treatments. We analyzed 66 Landsat TM/ETM + images spanning 26 years between 1988 and 2014 at five sites and one hyperspectral image from 2014 at two sites. The MOD10 snow product performs well in estimating Arizona's thin and discontinuous snow distribution. Mann-Kendall time-series analysis indicate significantly increasing trends in the annual number of snow cover days (SCD) over the 12-year period in 1.6% of the region at elevation transitions such as the Mogollon Rim in central Arizona, while significantly decreasing trends are observed at a few locations of lower elevations leading to the desert margins in eastern Arizona. The observed temporal trends are mostly consistent with ground-based SNOTEL snow measurements. An Arizona specific, Landsat sensor-derived binary classification model, similar to the MOD10 product, was developed at a local scale. It performs better than commonly-used simple threshold-based approaches, but demonstrates the continued challenges associated with Landsat sensor-derived snow classification in Arizona likely due to its coarse temporal resolution. Landsat-derived multi-temporal Normalized Difference Snow Index (NDSI) analysis indicate that treated (thinned and thinned-and-burned) forest sites had significantly greater NDSI values than untreated control sites. Snowpack at treated sites also appeared to persist longer into the spring season with potentially greater contributions to groundwater recharge in this semi-arid region. The high-resolution hyperspectral data analysis indicate that sites treated to approximately 24% forest canopy cover appear to have an optimum threshold for accumulating and maintaining snowpack. It balances canopy cover versus canopy gap, which reduces snow interception and sublimation by canopy, while providing enough shade. These results are encouraging for the 4FRI, the first and largest forest restoration effort in the US history, aimed at improving watershed health and function in the face of changing climate.

Detecção da Infraestrutura para Exploração Florestal em Rondônia Utilizando Dados de Sensoriamento Remoto

RESUMOA infraestrutura para exploração florestal é fundamental nas atividades de exploração madeireira, mas responde por grande parte dos impactos ambientais dessa atividade. O presente estudo aplicou e testou cinco técnicas de processamento digital de imagens orbitais com três resoluções espaciais distintas para a detecção de estradas e pátios florestais construídos em áreas sob manejo florestal sustentável na Floresta Nacional do Jamari, estado de Rondônia. Os resultados mostraram que o NDVI e a Análise de componentes principais apresentaram a melhor acurácia global nas resoluções espaciais de 5 e 10 metros e na de 30 metros, respectivamente. De maneira geral, a acurácia do produtor para a classe de interesse não foi alta, alcançando no máximo 39,2%, e índice Kappa de 0,38. A baixa performance das técnicas de geoprocessamento utilizadas na detecção de florestas exploradas seletivamente está relacionada às alterações ocorridas no dossel da floresta manejada, pouco perceptíveis em imagens de satélite.

Modelling forest canopy trends with on-demand spatial simulation

Understanding trends in forest canopy cover at local, national, and global scales is important for many applications, including policymaking related to forest carbon sequestration. Globally consistent land-cover data sets derived from MODerate-resolution Imaging Spectroradiometer (MODIS) are now available for a period of more than 10 years, long enough to detect trends both in deforestation and in afforestation. However, methods of modelling land-cover change normally require specialized software and expertise, limiting the availability of this information. This barrier to access can be eliminated through the use of web services that construct models on demand based on user-specified regions of interest, so that parameters are inferred from, and relevant to, local conditions. In this paper we present a proof-of-concept system for building and running spatial Markov chain models of forest-cover change on demand, and demonstrate how the on-demand approach may be implemented for similar applications.

Geostatistical modeling using LiDAR-derived prior knowledge with SPOT-6 data to estimate temperate forest canopy cover and above-ground biomass via stratified random sampling

Forest canopy cover (CC) and above-ground biomass (AGB) are important ecological indicators for forest monitoring and geoscience applications. This study aimed to estimate temperate forest CC and AGB by integrating airborne LiDAR data with wall-to-wall space-borne SPOT-6 data through geostatistical modeling. Our study involved the following approach: (1) reference maps of CC and AGB were derived from wall-to-wall LiDAR data and calibrated by field measurements; (2) twelve discrete LiDAR flights were simulated by assuming that LiDAR data were only available beneath these flights; (3) training/testing samples of CC and AGB were extracted from the reference maps inside and outside the simulated flights using stratified random sampling; (4) The simple linear regression, ordinary kriging and regression kriging model were used to extend the sparsely sampled CC/AGB data to the entire study area by incorporating a selection of SPOT-6 variables, including vegetation indices and texture variables. The regression kriging model was superior at estimating and mapping the spatial distribution of CC and AGB, as it featured the lowest mean absolute error (MAE; 11.295% and 18.929 t/ha for CC and AGB, respectively) and root mean squared error (RMSE; 17.361% and 21.351 t/ha for CC and AGB, respectively). The predicted and reference values of both CC and AGB were highly correlated for the entire study area based on the estimation histograms and error maps. Finally, we concluded that the regression kriging model was superior and more effective at estimating LiDAR-derived CC and AGB values using the spatially-reduced samples and the SPOT-6 variables. The presented modeling workflow will greatly facilitate future forest growth monitoring and carbon stock assessments for large areas of temperate forest in northeast China. It also provides guidance on how to take full advantage of future sparsely collected LiDAR data in cases where wall-to-wall LiDAR coverage is not available from the perspective of geostatistics.

Canopy cover and drought influence the landscape epidemiology of an amphibian chytrid fungus

Batrachochytrium dendrobatidis (Bd) is a globally distributed fungal pathogen of amphibians that has caused amphibian population declines. Prevalence and infection intensity of Bd should vary spatially across heterogeneous landscapes because survival of the pathogen depends on environmental conditions including humidity and temperature. Wetland temperatures should vary with forest canopy cover, but few studies have examined the relationship between Bd prevalence and canopy cover. We sampled anurans for Bd at 33 wetlands in Illinois across a canopy cover gradient during the severe drought of 2012 and during more typical weather in 2013. We hypothesized that anurans from closed canopy wetlands would have higher Bd prevalence than anurans from open canopy wetlands because high canopy cover would prevent temperatures from exceeding Bd's critical maximum temperature as often. We used occupancy modeling to examine how Bd prevalence varied spatially in relation to environmental covariates, and to determine if weather conditions altered such relationships. We analyzed all common species combined and Spring Peepers (Pseudacris crucifer) separately. Bd was widespread among the study wetlands. During the drought year of 2012, however, Bd prevalence and infection intensity were substantially reduced. Although wetland temperatures were related negatively to canopy cover, the influence of canopy cover on Bd was complex. Bd prevalence for Spring Peepers was related positively to canopy cover during the drought, but not in the non‐drought year. Bd prevalence for all common species was not related positively to canopy cover in either year. Effects of other environmental covariates, such as water depth, also differed between years. Weather conditions had a large impact on Bd infection patterns, which makes predicting pathogen dynamics difficult. In the Midwest region of the United States, climate change models predict more frequent droughts, which may generally inhibit Bd while also altering the effect of environmental factors such as canopy cover on Bd prevalence for amphibians.

Historical and current landscape‐scale ponderosa pine and mixed conifer forest structure in the Southern Sierra Nevada

Many managers today are tasked with restoring forests to mitigate the potential for uncharacteristically severe fire. One challenge to this mandate is the lack of large‐scale reference information on forest structure prior to impacts from Euro‐American settlement. We used a robust 1911 historical dataset that covers a large geographic extent (>10,000 ha) and has unbiased sampling locations to compare past and current forest conditions for ponderosa pine and mixed conifer forests in the southern Sierra Nevada. The 1911 dataset contained records from 18,052 trees in 378 sampled transects, totaling just over 300 ha in transect area. Forest structure was highly variable in 1911 and shrubs were found in 54% of transects. Total tree basal area ranged from 1 to 60 m2 ha−1 and tree density from 2 to 170 ha−1 (based on trees >30 cm dbh). K‐means cluster analysis divided transects into four groups: mixed conifer‐high basal area (MC High BA), mixed conifer‐average basal area (MC Ave BA), mixed conifer‐average basal area‐high shrubs (MC Ave BA Shrubs), and ponderosa pine (Pond Pine). The percentage of this 1911 landscape that experienced high severity fire was low and varied from 1–3% in mixed conifer forests and 4–6% in ponderosa pine forests. Comparing forest inventory data from 1911 to the present indicates that current forests have changed drastically, particularly in tree density, canopy cover, the density of large trees, dominance of white fir in mixed conifer forests, and the similarity of tree basal area in contemporary ponderosa pine and mixed conifer forests. Average forest canopy cover increased from 25–49% in mixed conifer forests, and from 12–49% in ponderosa pine forests from 1911 to the present; canopy cover in current forest types is similar but in 1911 mixed conifer forests had twice the canopy cover as ponderosa pine forests. Current forest restoration goals in the southern Sierra Nevada are often skewed toward the higher range of these historical values, which will limit the effectiveness of these treatments if the objective is to produce resilient forest ecosystems into the future.

Potential improvement for forest cover and forest degradation mapping with the forthcoming Sentinel-2 program

Abstract. The forthcoming European Space Agency’s Sentinel-2 mission promises to provide high (10 m) resolution optical data at higher temporal frequencies (5 day revisit with two operational satellites) than previously available. CNES, the French national space agency, launched a program in 2013, ‘SPOT4 take 5’, to simulate such a dataflow using the SPOT HRV sensor, which has similar spectral characteristics to the Sentinel sensor, but lower (20m) spatial resolution. Such data flow enables the analysis of the satellite images using temporal analysis, an approach previously restricted to lower spatial resolution sensors. We acquired 23 such images over Tanzania for the period from February to June 2013. The data were analysed with aim of discriminating between different forest cover percentages for landscape units of 0.5 ha over a site characterised by deciduous intact and degraded forests. The SPOT data were processed by one extracting temporal vegetation indices. We assessed the impact of the high acquisition rate with respect to the current rate of one image every 16 days. Validation data, giving the percentage of forest canopy cover in each land unit were provided by very high resolution satellite data. Results show that using the full temporal series it is possible to discriminate between forest units with differences of more than 40% tree cover or more. Classification errors fell exclusively into the adjacent forest canopy cover class of 20% or less. The analyses show that forestation mapping and degradation monitoring will be substantially improved with the Sentinel-2 program.

Exploring issues of training data imbalance and mislabelling on random forest performance for large area land cover classification using the ensemble margin

Studies have demonstrated the robust performance of the ensemble machine learning classifier, random forests, for remote sensing land cover classification, particularly across complex landscapes. This study introduces new ensemble margin criteria to evaluate the performance of Random Forests (RF) in the context of large area land cover classification and examines the effect of different training data characteristics (imbalance and mislabelling) on classification accuracy and uncertainty. The study presents a new margin weighted confusion matrix, which used in combination with the traditional confusion matrix, provides confidence estimates associated with correctly and misclassified instances in the RF classification model. Landsat TM satellite imagery, topographic and climate ancillary data are used to build binary (forest/non-forest) and multiclass (forest canopy cover classes) classification models, trained using sample aerial photograph maps, across Victoria, Australia. Experiments were undertaken to reveal insights into the behaviour of RF over large and complex data, in which training data are not evenly distributed among classes (imbalance) and contain systematically mislabelled instances. Results of experiments reveal that while the error rate of the RF classifier is relatively insensitive to mislabelled training data (in the multiclass experiment, overall 78.3% Kappa with no mislabelled instances to 70.1% with 25% mislabelling in each class), the level of associated confidence falls at a faster rate than overall accuracy with increasing amounts of mislabelled training data. In general, balanced training data resulted in the lowest overall error rates for classification experiments (82.3% and 78.3% for the binary and multiclass experiments respectively). However, results of the study demonstrate that imbalance can be introduced to improve error rates of more difficult classes, without adversely affecting overall classification accuracy.

Runoff and soil erosion characteristics in different periods of an intensive forest management system in a tropical indonesian rainforest

Tropical Indonesian rainforest is managed using an Intensive Forest Management System (IFMS). The main activity is selective logging for timber harvesting and intensive line planting to enrich the standing stock. These activities signifi cantly alter the forest canopy cover and the hydrologic response of catchments, including peak discharge, direct runoff and soil erosion. Understanding the hydrologic effects in an IFMS is helpful to develop a forest management strategy. This study investigated the impact of vegetation changes on runoff and soil erosion characteristics in different periods of an IFMS resulting from their respective forestry treatments. This study was conducted in a natural tropical rainforest at Bukit Baka Experimental Catchment, Central Kalimantan, Indonesia. Runoff and soil erosion characteristics were investigated in three small catchments, a virgin forest catchment (C1), a 10-year-old line plantation catchment (C2) and a 1-year-old line plantation catchment (C3). The IFMS has increased in large scale of open forests and exposed the bare soil, especially in the early years after forest operation. This condition has changed drastically in the catchment hydrologic responses. The increase in discharge, direct runoff, and suspended sediment discharge in the C3 was higher than those in the C2 and C1, particularly for large rainfall events. The proportion of annual rainfall as water yield in the C1, C2, and C3 was 27, 30, and 41%, respectively. The annual suspended sediment yield in the C1, C2, and C3 was 0.15, 3.6, and 14.9 t ha −1 y −1 , respectively. The results showed that the magnitude of runoff and soil erosion depends on the interaction among the rainfall, forest cover changes, forest treatment applied and catchment characteristics. Ten years after forest operation, forest cover has recovery close to natural condition, but still there are differences in hydrological response. Controlling soil erosion during logging activities by reducing the impact of logging and combining ecologically based vegetation structure design is an effective way to control runoff and soil erosion.

Characterizing stand-level forest canopy cover and height using Landsat time series, samples of airborne LiDAR, and the Random Forest algorithm

Many forest management activities, including the development of forest inventories, require spatially detailed forest canopy cover and height data. Among the various remote sensing technologies, LiDAR (Light Detection and Ranging) offers the most accurate and consistent means for obtaining reliable canopy structure measurements. A potential solution to reduce the cost of LiDAR data, is to integrate transects (samples) of LiDAR data with frequently acquired and spatially comprehensive optical remotely sensed data. Although multiple regression is commonly used for such modeling, often it does not fully capture the complex relationships between forest structure variables. This study investigates the potential of Random Forest (RF), a machine learning technique, to estimate LiDAR measured canopy structure using a time series of Landsat imagery. The study is implemented over a 2600ha area of industrially managed coastal temperate forests on Vancouver Island, British Columbia, Canada. We implemented a trajectory-based approach to time series analysis that generates time since disturbance (TSD) and disturbance intensity information for each pixel and we used this information to stratify the forest land base into two strata: mature forests and young forests. Canopy cover and height for three forest classes (i.e. mature, young and mature and young (combined)) were modeled separately using multiple regression and Random Forest (RF) techniques. For all forest classes, the RF models provided improved estimates relative to the multiple regression models. The lowest validation error was obtained for the mature forest strata in a RF model (R2=0.88, RMSE=2.39m and bias=−0.16 for canopy height; R2=0.72, RMSE=0.068% and bias=−0.0049 for canopy cover). This study demonstrates the value of using disturbance and successional history to inform estimates of canopy structure and obtain improved estimates of forest canopy cover and height using the RF algorithm.