Land Cover Fractions Research Articles

The complementary relationship between actual and potential evaporation for spatially heterogeneous surfaces

AbstractThe Complementary Relationship (CR) between actual and potential evaporation offers an attractive framework for estimating actual evaporation of drying land surfaces from simple meteorological measurements. Land surfaces are often heterogeneous with variable soil types, land cover, and local hydrologic conditions that give rise to spatially variable evaporation dynamics. The main aim is to incorporate effects of spatial heterogeneities on estimates of actual evaporation in the CR framework. The study extends the physically based approach of Aminzadeh et al. (2016) and proposes upscaling schemes for land‐atmosphere interactions affecting reference evaporation from heterogeneous surfaces comprised of vegetation and bare soil patches. For small‐scale surface heterogeneity relative to the extent of convective boundary layer (CBL), area‐averaged atmospheric boundary conditions were imposed over the domain of interest to integrate contributions from patches with different dynamics. For large‐scale heterogeneity (large patches relative to the scale of the mean CBL), fluxes from each patch were weighted by their respective areas. Preliminary results are in reasonable agreement with available field measurements and illustrate various effects of heterogeneous surface evaporative fluxes on the CR response. The results also highlight hidden dynamics not captured by standard CR, such as ability of vegetated patches to support steady evaporative fluxes until the onset of water stress while bare soil has already dried out. The study provides new insights into the roles of different vegetation types, land cover fraction, and atmospheric conditions on regional CR behavior hence advancing predictive capabilities of actual evapotranspiration from spatially heterogeneous land surfaces.

Spatial and temporal patterns of surface–atmosphere energy exchange in a dense urban environment using scintillometry

Spatially integrated measurements of the surface energy balance (SEB) are needed in urban areas to evaluate urban climate models and satellite observations. Scintillometers allow observations of sensible heat flux (QH) over much larger areas than techniques such as eddy covariance (EC), however methods are needed to partition between remaining unmeasured SEB terms. This is the first study to use observed spatial and temporal patterns of QH from a scintillometer network to constrain estimates of remaining SEB terms in a dense, heterogeneous urban environment. Results show that QH dominates the surface energy balance in central London throughout the year, with expected diurnal courses and seasonal trends in QH magnitude related to solar radiation input. Measurements also reveal a clear anthropogenic component of QH with winter (summer) weekday QH values 11.7% (5.1%) higher than weekends. Spatially, QH magnitude is correlated with vegetation and building landcover fraction in the measurement source areas. Spatial analysis provides additional evidence of anthropogenic influence with highest weekday/weekend ratios (1.55) from the City of London. Spatial differences are used to estimate horizontal advection and a novel method to estimate monthly latent heat flux is developed based on observed landcover and wet–dry surface variations in normalized QH. Annual anthropogenic heat emissions are estimated to be 46.3 W m−2 using an energy balance residual approach. The methods presented here have potential to significantly enhance understanding of urban areas, particularly in areas with tall buildings where there are few observational data.

Using Landsat, MODIS, and a Biophysical Model to Evaluate LST in Urban Centers

In this paper, we assessed and compared land surface temperature (LST) in urban centers using data from Landsat, MODIS, and the Simple Biosphere model (SiB2). We also evaluated the sensitivity of the model’s LST to different land cover types, fractions (percentages), and emissivities compared to reference points derived from Landsat thermal data. This was demonstrated in three climatologically- and morphologically-different cities of Atlanta, GA, New York, NY, and Washington, DC. Our results showed that in these cities SiB2 was sensitive to both the emissivity and the land cover type and fraction, but much more sensitive to the latter. The practical implications of these results are rather significant since they imply that the SiB2 model can be used to run different scenarios for evaluating urban heat island (UHI) mitigation strategies. This study also showed that using detailed emissivities per land cover type and fractions from Landsat-derived data caused a convergence of the model results towards the Landsat-derived LST for most of the studied cases. This study also showed that SiB2 LSTs are closer in magnitude to Landsat-derived LSTs than MODIS-derived LSTs. It is important, however, to emphasize that both Landsat and MODIS LSTs are not direct observations and, as such, do not represent a ground truth. More studies will be needed to compare these results to in situ LST data and provide further validation.

Persistent Organic Pollutants in Streamwater: Influence of Hydrological Conditions and Landscape Type

Concentrations of polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/Fs), polychlorinated biphenyls (PCBs), and hexachlorobenzene (HCB) in streamwater were measured in a remote catchment in northern Sweden and downstream to the Baltic Sea. Sampling took place at seven sites during two years and under different hydrological conditions: during the snow-free, snow-covered, and spring-flood seasons. Concentrations varied substantially between seasons and were up to 20 times higher during the spring flood compared to the preceding snow-covered period. The increase in concentrations with runoff was due to higher levels of particle-associated contaminants, while the dissolved concentrations remained stable. Particulate-contaminant concentrations were positively correlated primarily to suspended particulate matter (SPM) at sites in areas with a high land-cover fraction of sorted sediment. When upstream sampling locations were compared, a mire-dominated stream had higher concentrations and a lower retention of atmospherically deposited contaminants than a forest stream of the same catchment size. Contaminant concentrations (normalized to volume) did not increase consistently downstream despite the presence of several point sources. However, when normalized to the amount of SPM, concentrations were on average >20 times higher at the outlet in the Baltic Sea compared to the outlet from the remote catchment without point sources.

A Superresolution Land-Cover Change Detection Method Using Remotely Sensed Images With Different Spatial Resolutions

The development of remote sensing has enabled the acquisition of information on land-cover change at different spatial scales. However, a tradeoff between spatial and temporal resolutions normally exists. Fine-spatial-resolution images have low temporal resolutions, whereas coarse-spatial-resolution images have high temporal repetition rates. A novel superresolution change detection method (SRCD) is proposed to detect land-cover changes at both fine spatial and temporal resolutions with the use of a coarse-resolution image and a fine-resolution land-cover map acquired at different times. SRCD is an iterative method that involves endmember estimation, spectral unmixing, land-cover fraction change detection, and superresolution land-cover mapping. Both the land-cover change/no-change map and from-to change map at fine spatial resolution can be generated by SRCD. In this paper, SRCD was applied to a synthetic multispectral image, a Moderate-Resolution Imaging Spectroradiometer multispectral image, and a Landsat-8 Operational Land Imager multispectral image. The land-cover from-to change maps are found to have the highest overall accuracy (higher than 85%) in all of the three experiments. Most of the changed land-cover patches, which were larger than the coarse-resolution pixel, were correctly detected.

Error decomposition of nine passive and active microwave satellite soil moisture data sets over Australia

Abstract Soil moisture is one of the essential climate variables for the Global Climate Observing System (GCOS) that has been prioritized by the ESA's Climate Change Initiative to construct its homogeneous long-term climate record. This requires a consistent characterization of the error structures in the individual data sets, which vary due to changes in instrument configuration and calibration, and retrieval algorithm design. In this paper, the random error and systematic differences in nine passive and active microwave satellite soil moisture products over Australia (time coverage: 1978–present) are estimated in a same manner for SM components at subseasonal and seasonal-to-interannual timescales separately. The multi-scale error structures are found to be non-trivial and vary between the products, giving cause for conducting multi-scale merging with awareness of these differences. Noticeable similarities between the error structures of the satellite products derived from same retrieval algorithm and same measuring frequency however suggest transferability of error parameters between them. Using partial rank correlation analysis, the error maps are linked to statistics on vegetation index, digital elevation, soil moisture and soil temperature, and land cover fractions and mixing in order to explain the observed variability and the similarities between the products.

Development of a Class-Based Multiple Endmember Spectral Mixture Analysis (C-MESMA) Approach for Analyzing Urban Environments

Multiple endmember spectral mixture analysis (MESMA) has been widely applied for estimating fractional land covers from remote sensing imagery. MESMA has proven effective in addressing inter-class and intra-class endmember variability by allowing pixel-specific endmember combinations. This method, however, assumes that each land cover type has an equal probability of being included in the model, and the one with the least estimation error (e.g., root mean square error) was chosen as the “best-fit” model. Such an approach may mistakenly include a land cover class in the model and overestimate its abundance, or it might omit a class from the model and subsequently lead to underestimation. To address this problem, this paper developed a land cover class-based multiple endmember spectral mixture analysis (C-MESMA) method. In particular, a support vector machine (SVM) method with reflectance spectra and spectral indices, including the normalized difference vegetation index (NDVI), the biophysical composition index (BCI), and the ratio normalized difference soil index (RNDSI), were employed to classify the image into six land cover classes: pure impervious surface area (ISA), pure vegetation, pure soil, ISA-vegetation, vegetation-soil, and vegetation-ISA-soil. With the information of land cover classes, an individual MESMA method was applied to each mixed class. Finally, the fractional maps were derived through integrating land cover fractions of each land cover class. Quantitative analysis of the resulting percent ISA (%ISA) and comparative analyses with traditional MESMA indicate that C-MESMA improved the estimation accuracy of %ISA.

A geostatistical temporal mixture analysis approach to address endmember variability for estimating regional impervious surface distributions

Spectral mixture analysis (SMA) is a major approach for estimating fractional land covers through modeling the relationship between the spectral signatures of a mixed remote sensing pixel and those of the comprised pure land covers (also termed as endmembers). When SMA is implemented, endmember variability has proven to have significant impact on the accuracy of land cover fraction estimates. To address the endmember variability problem, this article developed a geostatistical temporal mixture analysis (GTMA) technique, with which spatially varying per-pixel endmember sets were estimated using an ordinary kriging interpolation technique. The method was applied to time-series moderate-resolution imaging spectroradiometer normalized difference vegetation index imagery in Wisconsin and North Carolina, United States to estimate regional impervious surface distributions. Analysis of results suggests that GTMA has achieved a promising accuracy. Detailed analysis indicates that a better performance has been achieved in less-developed areas than developed areas, and slight underestimation and slight overestimation have been detected in developed areas and less-developed areas, respectively. Moreover, while the performance of GTMA is comparable to those of phenology-based TMA and phenology-based multiple endmember TMA over the entire study area and in less-developed areas, a much better performance has been achieved in developed areas. Finally, this article argues that endmember variability may be more essential in developed areas when compared to less-developed areas.

Minimizing Reflectance Anisotropy Effects in Airborne Spectroscopy Data Using Ross–Li Model Inversion With Continuous Field Land Cover Stratification

The spectral and radiometric quality of airborne imaging spectrometer data is affected by the anisotropic reflectance behavior of the imaged surface. Illumination and observation angle-dependent patterns of surface reflected radiation propagate into products, hinder quantitative assessment of biophysical/biochemical parameters, and decrease the comparability of data from multiple flight lines. The Ross–Li model, originally developed for multiangular observations, can be inverted to estimate and correct for surface anisotropy effects. This requires land cover be stratified into distinct types of scattering behavior. When the observations subsumed in these classes cover a range of view angles, a pseudo multiangular view on the surface can be employed to invert the Ross–Li model. A discrete land cover classification, however, bears the risk of inappropriate scattering correction resulting in spatial artifacts in the corrected data, predominantly in transition regions of two land cover types (e.g., soil and sparse vegetation with varying fractions). We invert the Ross–Li model on continuous land cover fraction layers. We decompose land cover in dominating structural types using linear spectral unmixing. Ross–Li kernel weights and formulations are estimated for each type independently; the correction is then applied pixel-wise according to the fractional distribution. The corrected Airborne Prism EXperiment imaging spectrometer data show significant reduction of anisotropic reflectance effects of up to 90% (average 60% to 75%, $p=0.05$ ), measured in the overlapping regions of adjacent flight lines. No spatial artifacts or spectral irregularities are observed after correction.

Examining the impact of urban biophysical composition and neighboring environment on surface urban heat island effect

Due to atmospheric and surface modifications associated with urbanization, surface urban heat island (SUHI) effects have been considered essential in examining urban ecological environments. With remote sensing technologies, numerous land cover type related variables, including spectral indices and land cover fractions, have been applied to estimate land surface temperature (LST), thereby further examining SUHI. This study begins with the reexamination of the commonly used indicators of LST using Landsat Enhanced Thematic Mapper Plus (ETM+) and Landsat Thematic Mapper (TM) images which cover four counties of Wisconsin, United States. Origin of the large variation of LST found in urban areas is then investigated by discriminating soil and impervious surfaces. Except land cover types, neighboring environment is another key factor which may affect LST in urban areas. Thus, a neighboring effect considered method is proposed at the end of the study to better understand the relationship between impervious surfaces fraction (%ISA) and LST by taking the influence of neighboring environment into account. Results indicate that spectral indices have better performance in predicting LST than land cover fractions do within the study area. However, the result remains arguable due to the complexity and uncertainty of spectral mixture analysis. Impervious surfaces are found responsible for the large variation of LST in urban areas, which indicates that impervious surfaces should not be simply considered as a single land cover type has stable negative correlation with LST. Moreover, a better relationship is found between %ISA and LST when neighboring effect is considered, when compared to the traditional method which ignores the neighboring effect.

The effect of imposing ‘fractional abundance constraints’ onto the multilayer perceptron for sub-pixel land cover classification

To be physically interpretable, sub-pixel land cover fractions or abundances should fulfill two constraints, the Abundance Non-negativity Constraint (ANC) and the Abundance Sum-to-one Constraint (ASC). This paper focuses on the effect of imposing these constraints onto the MultiLayer Perceptron (MLP) for a multi-class sub-pixel land cover classification of a time series of low resolution MODIS-images covering the northern part of Belgium. Two constraining modes were compared, (i) an in-training approach that uses ‘softmax’ as the transfer function in the MLP’s output layer and (ii) a post-training approach that linearly rescales the outputs of the unconstrained MLP. Our results demonstrate that the pixel-level prediction accuracy is markedly increased by the explicit enforcement, both in-training and post-training, of the ANC and the ASC. For aggregations of pixels (municipalities), the constrained perceptrons perform at least as well as their unconstrained counterparts. Although the difference in performance between the in-training and post-training approach is small, we recommend the former for integrating the fractional abundance constraints into MLPs meant for sub-pixel land cover estimation, regardless of the targeted level of spatial aggregation.

Incorporating Endmember Variability into Linear Unmixing of Coarse Resolution Imagery: Mapping Large-Scale Impervious Surface Abundance Using a Hierarchically Object-Based Spectral Mixture Analysis

As an important indicator of anthropogenic impacts on the Earth’s surface, it is of great necessity to accurately map large-scale urbanized areas for various science and policy applications. Although spectral mixture analysis (SMA) can provide spatial distribution and quantitative fractions for better representations of urban areas, this technique is rarely explored with 1-km resolution imagery. This is due mainly to the absence of image endmembers associated with the mixed pixel problem. Consequently, as the most profound source of error in SMA, endmember variability has rarely been considered with coarse resolution imagery. These issues can be acute for fractional land cover mapping due to the significant spectral variations of numerous land covers across a large study area. To solve these two problems, a hierarchically object-based SMA (HOBSMA) was developed (1) to extrapolate local endmembers for regional spectral library construction; and (2) to incorporate endmember variability into linear spectral unmixing of MODIS 1-km imagery for large-scale impervious surface abundance mapping. Results show that by integrating spatial constraints from object-based image segments and endmember extrapolation techniques into multiple endmember SMA (MESMA) of coarse resolution imagery, HOBSMA improves the discriminations between urban impervious surfaces and other land covers with well-known spectral confusions (e.g., bare soil and water), and particularly provides satisfactory representations of urban fringe areas and small settlements. HOBSMA yields promising abundance results at the km-level scale with relatively high precision and small bias, which considerably outperforms the traditional simple mixing model and the aggregated MODIS land cover classification product.

Mapping vegetation in a late Quaternary landform of the Amazonian wetlands using object-based image analysis and decision tree classification

Fan-shaped morphologies related to late Quaternary residual megafan depositional systems are common features over wide areas in northern Amazonia. These features were formed by ancient distributary drainage systems that are in great contrast to tributary drainage networks that typify the modern Amazon basin. The surfaces of the Amazonian megafans constitute vegetacional mosaic wetlands with different campinarana types. A fine-scale-resolution investigation is required to provide detailed classification maps for the various campinarana and surrounding forest types associated with the Amazonian megafans. This approach remains to be presented, despite its relevance for analysing the relationship between stages of plant succession and sedimentary dynamics associated with the evolution of megafans. In this work, we develop a methodology for classifying vegetation over a fan-shaped megafan palaeoform from a northern Amazonian wetland. The approach included object-based image analysis (OBIA) and data-mining (DM) techniques combining Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) images, land-cover fractions derived by the linear spectral mixing model, synthetic aperture radar (SAR) images, and the digital elevation model (DEM) acquired during the Shuttle Radar Topography Mission (SRTM). The DEM, vegetation fraction, and ASTER band 3 were the most useful parameters for defining the forest classes. The normalized difference vegetation index (NDVI), ASTER band 1, vegetation fraction, and the Advanced Land Observing Satellite (ALOS)/Phased Array type L-band Synthetic Aperture Radar (PALSAR) transmitting and receiving horizontal polarization (HH) and transmitting horizontal and receiving vertical polarization (HV) were all effective in distinguishing the wetland classes campinarana and Mauritia. Tests of statistical significance indicated the overall accuracies and kappa coefficients (κ) of 88% and 0.86 for the final map, respectively. The allocation disagreement coefficient of 5% and a quantity disagreement value of 7% further attested the statistical significance of the classification results. Hence, in addition to water, exposed soil, and deforestation areas, OBIA and DM were successful for differentiating a large number of open (forest, wood, shrub, and grass campinaranas), forest (terra firme, várzea, igapó, and alluvial), as well as Mauritia wetland classes in the inner and outer areas of the studied megafan.

Comparison of data gap-filling methods for Landsat ETM+ SLC-off imagery for monitoring forest degradation in a semi-deciduous tropical forest in Mexico

A number of methods to overcome the 2003 failure of the Landsat 7 Enhanced Thematic Mapper (ETM+) scan-line corrector (SLC) are compared in this article in a forest-monitoring application in the Yucatan Peninsula, Mexico. The objective of this comparison is to determine the best approach to accomplish SLC-off image gap-filling for the particular landscape in this region, and thereby provide continuity in the Landsat data sensor archive for forest-monitoring purposes. Four methods were tested: (1) local linear histogram matching (LLHM); (2) neighbourhood similar pixel interpolator (NSPI); (3) geostatistical neighbourhood similar pixel interpolator (GNSPI); and (4) weighted linear regression (WLR). All methods generated reasonable SLC-off gap-filling data that were visually consistent and could be employed in subsequent digital image analysis. Overall accuracy, kappa coefficients (κ), and quantity and allocation disagreement indices were used to evaluate unsupervised Iterative Self-Organizing Data Analysis (ISODATA) land-cover classification maps. In addition, Pearson correlation coefficients (r) and root mean squares of the error (RMSEs) were employed for estimates agreement with fractional land cover. The best results visually (overall accuracy > 85%, κ < 9%, quantity disagreement index < 5.5%, and allocation disagreement index < 12.5%) and statistically (r > 0.84 and RMSE < 7%) were obtained from the GNSPI method. These results suggest that the GNSPI method is suitable for routine use in reconstructing the imagery stack of Landsat ETM+ SLC-off gap-filled data for use in forest-monitoring applications in this type of heterogeneous landscape.

The impact of impervious water-storage parametrization on urban climate modelling

In order to improve the representation of the water balance in urban land-surface models, we present a new impervious water-storage parametrization that assumes a distribution of water reservoirs. It has been implemented in TERRA-URB, a new urban parametrization for COSMO-CLM’s standard land-surface module TERRA-ML. The water-storage capacity and the maximal wet surface fraction of the urban impervious land cover consisting of streets and buildings are estimated for Toulouse centre by matching the modelled and observed evapotranspiration (ET) rates. They amount to 1.31±0.20kgm−2 and 12±4%, respectively. The model successfully reproduces the timespan and magnitude of increased ET for both urban observations campaigns CAPITOUL and BUBBLE. Our sensitivity study reveals that water-storage parametrization largely affects the performance of modelled ET rates. Hereby, the simulation employing the new water-storage parametrization is improved compared to arbitrary or existing water-storage parametrizations. The ET, surface sensible heat exchange and upwelling infra-red radiation are all affected until 12 day-time hours after rainfall on average. The modelled annual-mean ET during the CAPITOUL campaign from the urban land in Toulouse is an order of magnitude lower than that observed for the natural surroundings.

Extending the vegetation–impervious–soil model using simulated EnMAP data and machine learning

The upcoming hyperspectral satellite mission Environmental Mapping and Analysis Program (EnMAP) will provide timely and globally sampled imaging spectrometer data on a frequent basis. This will create unprecedented opportunities for a variety of environmental research fields and lead to manifold novel applications. These opportunities specifically apply to challenging environments, including heterogeneous urban landscapes. In this paper, we explored the potential of EnMAP data for quantifying land cover along the urban–rural gradient of Berlin, Germany. Land cover fraction maps from a simulated EnMAP scene at 30m spatial resolution were derived based on support vector regression (SVR) combined with synthetically mixed training data. Results demonstrate that EnMAP imagery will be well suited for mapping impervious, vegetation and soil surface types according to the VIS framework. Moreover, EnMAP data will allow extending the VIS framework by more detailed sub-categories such as roof and pavement, or low vegetation and tree. However, we advise caution that spaceborne imaging spectrometer data of improved quality will not completely help to overcome well known phenomena of spectral similarity between materials and spectral confusion caused by the presence of shaded areas. To identify possible benefits and limitations of EnMAP data, comparisons to fraction maps derived from a higher resolution Hyperspectral Mapper (HyMap) image at 9m spatial resolution and a multispectral Landsat ETM+-like image at 30m spatial resolution were drawn. First, we demonstrate that both VIS and extended VIS mapping reveal similar accuracies compared to maps from spatially higher resolution data. Second, we illustrate the superiority of the higher spectral information content for improved and extended urban land cover mapping compared to multispectral data. Overall, this study provides important insights into the potential of spaceborne imaging spectrometer and specifically future EnMAP data for urban remote sensing.

Management alters C allocation in turfgrass lawns

Turfgrass is a large fraction of land cover in urbanized areas around the world, but little information exists about basic ecological processes in lawns or the controls on these processes. Here, we characterize biomass carbon stocks and fluxes in tall fescue (Festuca arundinacea) appropriate for use on home lawns, managed with differing levels of irrigation, mowing height, and fertilization to test for effects of common management practices. Total standing biomass C (to 10cm depth) averaged 6.04Mgha−1 over the course of the study, with slightly more C in roots than in stubble; management had little effect on the average standing biomass C pool, but there was substantial intraannual variation in biomass C stocks (>2MgCha−1), and the pattern of variation was affected by all three management practices. Net primary production (NPP) averaged 4.50MgCha−1yr−1; management did not affect NPP, but had significant effects on how C was allocated to each component of production (roots, stubble, or clippings). Despite low standing biomass, turfgrass lawns generate a large annual flux of C, and the allocation of that flux is strongly modified by management.

On quantifying post-classification subpixel landcover changes

The post-classifications change matrix is well defined for hard classifications. However, for soft classifications, where partial membership of pixels to landcover classes is allowed, there is no single definition for such a matrix. In this paper, we argue that a natural definition of the post-classification change matrix for subpixel classifications can be done in terms of a constrained optimization problem, according to which the change matrix should allow an optimal prediction of the subpixel landcover fractions at the latest date from those of the earliest date. We first show that the traditional change matrix for crisp classification corresponds to the optimal solution of the unconstrained problem. Then, the formulation is generalized for subpixel classifications by incorporating certain constraints pertaining to desirable properties of a change matrix, thus resulting in a constrained least square (CLS) change matrix. In addition, based on intuitive criteria, a generalized product (GPROD) was parameterized in terms of an exponent parameter and used to derive another change matrix. It was shown that when the exponent parameter of the GPROD operator tends to infinity, one of the most commonly used methods for map comparison from subpixel fractions, namely the MINPROD composite operator, results. The three matrices (CLS, GPROD and MINPROD) were tested on both simulated and real subpixel changes derived from QuickBird and Landsat TM images. Results indicated that, for small exponent values (0–0.5), the GPROD matrix yielded the lowest errors of estimated landcover changes, whereas the MINPROD generally yielded the highest errors for the same estimations.

A new data aggregation technique to improve landscape metric downscaling

Scale is a fundamental concept in landscape ecology and considerable attention has been given to the scale-dependent relationships of landscape metrics. Many metrics have been found to exhibit very consistent scaling relationships as map resolution (i.e., pixel or grain size) is increased. However, these scaling relationships tend to break down when attempting to ‘downscale’ them, and the scaling function is often unable to accurately predict metric values for finer resolutions than the original data. The reasons for this breakdown are not well understood. This research examines the downscaling behavior of metrics using various data aggregation techniques in an attempt to better understand the characteristics of metric scaling behavior. First, downscaling performance is examined using the traditional method of aggregation known as ‘majority rules’. Second, a new data aggregation technique is introduced that utilizes fractional land cover abundances obtained from sub-pixel remote sensing classifications in order to capture a greater amount of the spatial heterogeneity present in the landscape. The goal of this new aggregation technique is to produce a more accurate scaling relationship that can be downscaled to predict metric values at fine resolutions. Results indicate that sub-pixel classifications have the potential to transform data aggregation to allow more accurate downscaling for certain landscapes, but accuracy is linked to the spatial heterogeneity of the landscape.

A Comparison of Advanced Regression Algorithms for Quantifying Urban Land Cover

Quantitative methods for mapping sub-pixel land cover fractions are gaining increasing attention, particularly with regard to upcoming hyperspectral satellite missions. We evaluated five advanced regression algorithms combined with synthetically mixed training data for quantifying urban land cover from HyMap data at 3.6 and 9 m spatial resolution. Methods included support vector regression (SVR), kernel ridge regression (KRR), artificial neural networks (NN), random forest regression (RFR) and partial least squares regression (PLSR). Our experiments demonstrate that both kernel methods SVR and KRR yield high accuracies for mapping complex urban surface types, i.e., rooftops, pavements, grass- and tree-covered areas. SVR and KRR models proved to be stable with regard to the spatial and spectral differences between both images and effectively utilized the higher complexity of the synthetic training mixtures for improving estimates for coarser resolution data. Observed deficiencies mainly relate to known problems arising from spectral similarities or shadowing. The remaining regressors either revealed erratic (NN) or limited (RFR and PLSR) performances when comprehensively mapping urban land cover. Our findings suggest that the combination of kernel-based regression methods, such as SVR and KRR, with synthetically mixed training data is well suited for quantifying urban land cover from imaging spectrometer data at multiple scales.