Coupling high-resolution satellite imagery with ALS-based canopy height model and digital elevation model in object-based boreal forest habitat type classification

Räsänen, Aleksi; Kuitunen, Markku; Tomppo, Erkki; Lensu, Anssi

doi:10.1016/j.isprsjprs.2014.05.003

Cited by 34 publications

(18 citation statements)

References 168 publications

(204 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Regardless of the choice of classifier, accuracy assessments are used to determine the quality of the classification, and several factors can affect the results of an accuracy assessment, including training sample size [11,12], the number of classes in the classification [13], the ability of the training data to adequately characterize the classes being mapped [10], and dimensionality of the data [13]. Generally, when performing image classification and accuracy assessments, training and validation data should be statistically independent (e.g., not clustered) [14] and representative of the entire landscape [10,12], and there should be abundant training data in all classes [15]. Many different training and validation sampling schemes are used throughout the literature, but without careful scrutiny of each dataset used and the specific assessment method, it may be difficult to compare results of classifications [11,16].…”

Section: Introductionmentioning

confidence: 99%

On the Importance of Training Data Sample Selection in Random Forest Image Classification: A Case Study in Peatland Ecosystem Mapping

2015

View full text Add to dashboard Cite

Random Forest (RF) is a widely used algorithm for classification of remotely sensed data. Through a case study in peatland classification using LiDAR derivatives, we present an analysis of the effects of input data characteristics on RF classifications (including RF out-ofbag error, independent classification accuracy and class proportion error). Training data selection and specific input variables (i.e., image channels) have a large impact on the overall accuracy of the image classification. High-dimension datasets should be reduced so that only uncorrelated important variables are used in classifications. Despite the fact that RF is an ensemble approach, independent error assessments should be used to evaluate RF results, and iterative classifications are recommended to assess the stability of predicted classes. Results are also shown to be highly sensitive to the size of the training data set. In addition to being as large as possible, the training data sets used in RF classification should also be (a) randomly distributed or created in a manner that allows for the class proportions of the training data to be representative of actual class proportions in the landscape; and (b) should have minimal spatial autocorrelation to improve classification results and to mitigate inflated estimates of RF out-of-bag classification accuracy.

show abstract

Section: Introductionmentioning

confidence: 99%

On the Importance of Training Data Sample Selection in Random Forest Image Classification: A Case Study in Peatland Ecosystem Mapping

2015

View full text Add to dashboard Cite

show abstract

“…Overall, we had 19-50 training segments for each class. We then reduced the number of features from 262 to 109 using the RF-based wrapper feature selection algorithm Boruta (Kursa and Rudnicki, 2010;Räsänen et al, 2014;Li et al, 2016) in R 3.2.2 (R Core Team, 2015). After 1000 RF runs in Boruta, 10 variables were deemed confirmed, rejected or tentative, and if tentative, a tentative rough fix (Kursa and Rudnicki, 2010) was carried out.…”

Section: Land Cover Classification and Landscape Estimates Of Plant Amentioning

confidence: 99%

Spatial variation and linkages of soil and vegetation in the Siberian Arctic tundra – coupling field observations with remote sensing data

Mikola¹,

Virtanen²,

Linkosalmi³

et al. 2018

Preprint

Self Cite

View full text Add to dashboard Cite

Abstract. Arctic tundra ecosystems will have a key role in future climate change due to intensifying permafrost thawing, plant growth and ecosystem carbon exchange, but monitoring these changes may be challenging due to the heterogeneity 20 of Arctic landscapes. We examined spatial variation and linkages of soil and plant attributes in a site of Siberian Arctic tundra in Tiksi, northeast Russia, and evaluated possibilities to capture this variation by remote sensing for the benefit of carbon exchange measurements and landscape extrapolation. We distinguished nine land cover types (LCTs) -bare soil, lichen tundra, shrub tundra, flood meadow, graminoid tundra, bog, dry fen, wet den and water -to classify the variation in our site. To characterize the LCTs, we sampled 92 study plots for plant (biomass and leaf area index, LAI) and soil 25 (organic matter OM%, bulk density, moisture, pH, litter layer depth, litter mass loss, temperature and active layer depth) attributes in 2014. Moreover, to test if variation in plant and soil attributes can be detected using remote sensing, we produced a normalized difference vegetation index (NDVI) and topographical parameters for each study plot using three very high spatial resolution multispectral satellite images (QuickBird and WorldView-2, portraying vegetation at 180, 220 and 750 growing degree days, DD with 0 °C threshold) and a digital elevation model (derived from a WV-2 stereo-30 pair image). We found that soils in our site ranged from mineral soils in bare soil and lichen tundra (on average 3.9 % OM) to soils of high OM% in graminoid tundra, bog, dry fen and wet fen (38 %), with shrub tundra and flood meadow Vascular shoot mass and LAI were also positively associated with soil OM content, and LAI with active layer depth, but the amount of variation explained was significantly lower (6-15 %). NDVI captured variation in peak season vascular LAI better than variation in moss biomass, but the difference depended on the phase of the growing season in the image:180-DD, 220-DD and 750-DD NDVI captured 23, 17 and 7 % of moss mass variation and 25, 34 and 50% of vascular 5 LAI variation, respectively. For this reason, soil attributes associated with moss mass were better captured by early season NDVI and those associated with LAI by late season NDVI. Topographic attributes were related to LAI and many soil attributes, but not to moss biomass and they could not increase the amount of spatial variation explained in plant and soil attributes above that achieved by NDVI. Our results illustrate a typical tundra ecosystem with great fine-scale spatial variation in both plant and soil attributes. Mosses dominate plant biomass and control many soil attributes, including 10 OM% and temperature, but variation in moss biomass is difficult to capture by remote sensing reflectance or topography.This suggests that using simple reflectance indices and DEM for spatial extrapolation of those vegetation and soil attributes that are relevant for regional ecosystem and global climate models warran...

show abstract

“…The overall accuracy (89%) achieved for all validation grids was also within the range of other studies, but did not reach the accuracies achieved by Hellesen and Matikainen [28] and Sasaki, et al [13] who produced classification accuracies over 95%. It is worth noting that almost all comparable classification studies were applied on a single study area with limited spatial coverage, ranging from 100 ha to 1500 ha [13,23,28,52]. This study area contains 69 sample grids that are geographically separated from each other, with a total area of 59,616 ha.…”

Section: Land Cover Classification Of All Validation Gridsmentioning

confidence: 99%

Mapping Net Stocked Plantation Area for Small-Scale Forests in New Zealand Using Integrated RapidEye and LiDAR Sensors

Xu¹,

Morgenroth²,

Manley³

2017

Forests

View full text Add to dashboard Cite

Abstract:In New Zealand, approximately 70% of plantation forests are large-scale (over 1000 ha) with accurate resource description. In contrast, the remaining 30% of plantation forests are small-scale (less than 1000 ha). It is forecasted that these small-scale forests will supply nearly 40% of the national wood production in the next decade. However, in-depth description of these forests, especially those under 100 ha, is very limited. This research evaluates the use of remote sensing datasets to map and estimate the net stocked plantation area for small-scale forests. We compared a factorial combination of two classification approaches (Nearest Neighbour (NN), Classification and Regression Tree (CART)) and two remote sensing datasets (RapidEye, RapidEye plus LiDAR) for their ability to accurately classify planted forest area. CART with a combination of RapidEye and LiDAR metrics outperformed the other three combinations producing the highest accuracy for mapping forest plantations (user's accuracy = 90% and producer's accuracy = 88%). This method was further examined by comparing the mapped plantations with manually digitised plantations based on aerial photography. The mapping approach overestimated the plantation area by 3%. It was also found that forest patches exceeding 10 ha achieved higher conformance with the digitised areas. Overall, the mapping approach in this research provided a proof of concept for deriving forest area and mapping boundaries using remote sensing data, and is especially relevant for small-scale forests where limited information is currently available.

show abstract

Coupling high-resolution satellite imagery with ALS-based canopy height model and digital elevation model in object-based boreal forest habitat type classification

Cited by 34 publications

References 168 publications

On the Importance of Training Data Sample Selection in Random Forest Image Classification: A Case Study in Peatland Ecosystem Mapping

On the Importance of Training Data Sample Selection in Random Forest Image Classification: A Case Study in Peatland Ecosystem Mapping

Spatial variation and linkages of soil and vegetation in the Siberian Arctic tundra – coupling field observations with remote sensing data

Mapping Net Stocked Plantation Area for Small-Scale Forests in New Zealand Using Integrated RapidEye and LiDAR Sensors

Contact Info

Product

Resources

About