John Odindi

Land-use/cover classification in a heterogeneous coastal landscape using RapidEye imagery: evaluating the performance of random forest and support vector machines classifiers

Mapping of patterns and spatial distribution of land-use/cover (LULC) has long been based on remotely sensed data. In the recent past, efforts to improve the reliability of LULC maps have seen a proliferation of image classification techniques. Despite these efforts, derived LULC maps are still often judged to be of insufficient quality for operational applications, due to disagreement between generated maps and reference data. In this study we sought to pursue two objectives: first, to test the new-generation multispectral RapidEye imagery classification output using machine-learning random forest (RF) and support vector machines (SVM) classifiers in a heterogeneous coastal landscape; and second, to determine the importance of different RapidEye bands on classification output. Accuracy of the derived thematic maps was assessed by computing confusion matrices of the classifiers’ cover maps with respective independent validation data sets. An overall classification accuracy of 93.07% with a kappa value of 0.92, and 91.80 with a kappa value of 0.92 was achieved using RF and SVM, respectively. In this study, RF and SVM classifiers performed comparatively similarly as demonstrated by the results of McNemer’s test (Z = 1.15). An evaluation of different RapidEye bands using the two classifiers showed that incorporation of the red-edge band has a significant effect on the overall classification accuracy in vegetation cover types. Consequently, pursuit of high classification accuracy using high-spatial resolution imagery on complex landscapes remains paramount.

Exploring the potential of in situ hyperspectral data and multivariate techniques in discriminating different fertilizer treatments in grasslands

Abstract. Optimizing the productivity of native rangelands has received considerable attention in range management. Rangeland fertilizer application has emerged as a popular intervention for improving rangeland quality. To achieve optimal range quality from such intervention, there is a need for quick and accurate methods of assessing the effects of different fertilizer combinations. The utility of in situ hyperspectral data and multivariate techniques in distinguishing 12 complex ammonium nitrate, ammonium sulfate, lime, and phosphorus fertilizer combinations on a grassland is assessed. Partial least squares regression discriminant analysis (PLS-DA) and discriminant analysis (DA) classification results derived using hyperspectral grass reflectance that were (1) fertilized using 11 combinations of ammonium sulfate, ammonium nitrate, phosphorus, and lime and (2) unfertilized experimental plots were compared. Results illustrate the strength of in situ hyperspectral data and multivariate techniques in detecting and discriminating grasses with different fertilizer treatments. Specifically, four bands within the red edge (731 and 737 nm) and the shortwave infrared (1310 and 1777 nm) regions of the electromagnetic spectrum demonstrated a high potential for discriminating the effects of fertilizer treatments on grasslands. DA outperformed PLS-DA in discriminating complex combinations of ammonium nitrate, ammonium sulfate combined with lime and phosphorus, as well as unfertilized grasses. Overall, spectroscopy and DA offer great potential for discriminating complex fertilizer combinations.

Assessing the potential of integrated Landsat 8 thermal bands, with the traditional reflective bands and derived vegetation indices in classifying urban landscapes

Abstract Reliable and up-to-date urban land cover information is valuable in urban planning and policy development. Due to the increasing demand for reliable land cover information there has been a growing need for robust methods and datasets to improve the classification accuracy from remotely sensed imagery. This study sought to assess the potential of the newly launched Landsat 8 sensor’s thermal bands and derived vegetation indices in improving land cover classification in a complex urban landscape using the support vector machine classifier. This study compared the individual and combined performance of Landsat 8’s reflective, thermal bands and vegetation indices in classifying urban land use-land cover. The integration of Landsat 8 reflective bands, derived vegetation indices and thermal bands overall produced significantly higher accuracy classification results than using traditional bands as standalone (i.e. overall, user and producer accuracies). An overall accuracy above 89.33% and a kappa index of 0.86, significantly higher than the one obtained with the use of the traditional reflective bands as a standalone data-set and other analysis stages. On average, the results also indicate high producer and user accuracies (i.e. above 80%) for most of the classes with a McNemar’s Z score of 9.00 at 95% confidence interval showing significant improvement compared with classification using reflective bands as standalone. Overall, the results of this study indicate that the integration of the Landsat 8’s OLI and TIR data presents an invaluable potential for accurate and robust land cover classification in a complex urban landscape, especially in areas where the availability of high resolution datasets remains a challenge.

Determining extreme heat vulnerability of Harare Metropolitan City using multispectral remote sensing and socio-economic data

Abstract Urbanisation alters surface landscape characteristics through conversion of natural landscapes to impervious surfaces. Such changes alter the thermal properties of urban landscape mosaics, increasing the urban heat island intensity and the population’s vulnerability to heat-related stress. This study aimed at deriving detailed area-specific spatial information on the distribution of heat vulnerability in Harare city, Zimbabwe, valuable for informed urban thermal mitigation, planning and decision-making. Using Landsat-8-derived bio-physical surface properties and socio-demographic factors, findings show that vulnerability to heat-related distress was high in over 40 percent of the city, mainly in densely built-up areas with low-income groups. Comparatively, low to moderate heat vulnerability was observed in the high-income northern suburbs with low physical exposure and population density. Results also showed a strong spatial correlation (α = 0.61) between heat vulnerability and observed surface temperatures in the hot season, signifying that land surface temperature is a good indicator of heat vulnerability in the area. Furthermore, the study showed that indices derived from moderate-resolution Landsat 8 data improve thermal risk assessment in areas of close proximity. These findings demonstrate the value of readily available multispectral data-sets in determining areas vulnerable to temperature extremes within a heterogeneous urban landscape. The findings are particularly valuable for designing heat-mitigation strategies as well as identifying highly vulnerable areas during heat waves.

Comparison between WorldView-2 and SPOT-5 images in mapping the bracken fern using the random forest algorithm

Abstract Plant species invasion is known to be a major threat to socioeconomic and ecological systems. Due to high cost and limited extents of urban green spaces, high mapping accuracy is necessary to optimize the management of such spaces. We compare the performance of the new-generation WorldView-2 (WV-2) and SPOT-5 images in mapping the bracken fern [Pteridium aquilinum (L) kuhn] in a conserved urban landscape. Using the random forest algorithm, grid-search approaches based on out-of-bag estimate error were used to determine the optimal ntree and mtry combinations. The variable importance and backward feature elimination techniques were further used to determine the influence of the image bands on mapping accuracy. Additionally, the value of the commonly used vegetation indices in enhancing the classification accuracy was tested on the better performing image data. Results show that the performance of the new WV-2 bands was better than that of the traditional bands. Overall classification accuracies of 84.72 and 72.22% were achieved for the WV-2 and SPOT images, respectively. Use of selected indices from the WV-2 bands increased the overall classification accuracy to 91.67%. The findings in this study show the suitability of the new generation in mapping the bracken fern within the often vulnerable urban natural vegetation cover types.

Detecting the Early Stage of Phaeosphaeria Leaf Spot Infestations in Maize Crop Using In Situ Hyperspectral Data and Guided Regularized Random Forest Algorithm

Phaeosphaeria leaf spot (PLS) is considered one of the major diseases that threaten the stability of maize production in tropical and subtropical African regions. The objective of the present study was to investigate the use of hyperspectral data in detecting the early stage of PLS in tropical maize. Field data were collected from healthy and the early stage of PLS over two years (2013 and 2014) using a handheld spectroradiometer. An integration of a newly developed guided regularized random forest (GRRF) and a traditional random forest (RF) was used for feature selection and classification, respectively. The 2013 dataset was used to train the model, while the 2014 dataset was used as independent test dataset. Results showed that there were statistically significant differences in biochemical concentration between the healthy leaves and leaves that were at an early stage of PLS infestation. The newly developed GRRF was able to reduce the high dimensionality of hyperspectral data by selecting key wavelengths with less autocorrelation. These wavelengths are located at 420 nm, 795 nm, 779 nm, 1543 nm, 1747 nm, and 1010 nm. Using these variables (), a random forest classifier was able to discriminate between the healthy maize and maize at an early stage of PLS infestation with an overall accuracy of 88% and a kappa value of 0.75. Overall, our study showed potential application of hyperspectral data, GRRF feature selection, and RF classifiers in detecting the early stage of PLS infestation in tropical maize.

Application of topo-edaphic factors and remotely sensed vegetation indices to enhance biomass estimation in a heterogeneous landscape in the Eastern Arc Mountains of Tanzania

Estimating tropical biomass is critical for establishment of conservation inventories and landscape monitoring. However, monitoring biomass in a complex and dynamic environment using traditional methods is challenging. Recently, biomass estimates based on remotely sensed data and ecological variables have shown great potential. The present study explored the utility of remotely sensed data and topo-edaphic factors to improve biomass estimation in the Eastern Arc Mountains of Tanzania. Twenty-nine vegetation indices were calculated from RapidEye data, while topo-edaphic factors were taken from field measurements. Results showed that using topo-edaphic variables or vegetation indices, biomass could be predicted with an R2 of 0.4. A combination of topo-edaphic variables and vegetation indices improved the prediction accuracy to an R2 of 0.6. Results further showed a decrease in biomass estimates from 1162 ton ha−1 in 1980 to 285.38 ton ha−1 in 2012. This study demonstrates the value of combining remotely sensed data with topo-edaphic variables in biomass estimation.

Comparison of partial least squares and support vector regressions for predicting leaf area index on a tropical grassland using hyperspectral data

Abstract. Leaf area index (LAI) is a key biophysical parameter commonly used to determine vegetation status, productivity, and health in tropical grasslands. Accurate LAI estimates are useful in supporting sustainable rangeland management by providing information related to grassland condition and associated goods and services. The performance of support vector regression (SVR) was compared to partial least square regression (PLSR) on selected optimal hyperspectral bands to detect LAI in heterogeneous grassland. Results show that PLSR performed better than SVR at the beginning and end of summer. At the peak of the growing season (mid-summer), during reflectance saturation, SVR models yielded higher accuracies (R2=0.902 and RMSE=0.371  m2 m−2) than PLSR models (R2=0.886 and RMSE=0.379  m2 m−2). For the combined dataset (all of summer), SVR models were slightly more accurate (R2=0.74 and RMSE=0.578  m2 m−2) than PLSR models (R2=0.732 and RMSE=0.58  m2 m−2). Variable importance on the projection scores show that most of the bands were located in the near-infrared and shortwave regions of the electromagnetic spectrum, thus providing a basis to investigate the potential of sensors on aerial and satellite platforms for large-scale grassland LAI prediction.

Varietal discrimination of common dry bean (Phaseolus vulgaris L.) grown under different watering regimes using multitemporal hyperspectral data

Abstract. Globally, the common dry bean varieties (Phaseolus vulgaris L.) are regarded as valuable food crops. Due to diversion-farm and postharvest management requirements, quick, reliable, and cost-effective varietal discrimination is critical for optimal management during growth and after harvesting. The large number of valuable wavelengths that characterize hyperspectral remotely sensed datasets in concert with emerging robust discriminant analysis techniques offers great potential for on-farm dry bean varietal discrimination. In this study, an integrated approach of partial least-squares discriminant analysis (PLS-DA) on hyperspectral data was used to determine the bean’s optimal timing for on-farm varietal discrimination. Based on experimental plots underirrigated and rain-fed watering regimes, hyperspectral data were collected at three major phenological stages. Data at each stage were first used to generate PLS-DA models to determine variable (wavebands) importance in the projection (VIP) and the VIP bands used to generate VIP conditioned PLS-DA models. The study identified 6 weeks (branching and rapid vegetative growth) and 10 weeks (flowering and pod development) after seed sowing as optimal stages for varietal discrimination. The study offers insight into the optimal period to discriminate dry bean varieties using spectroscopy, valuable for on-farm and after-farm management and crop monitoring sensor development.

Land surface temperature and emissivity estimation for Urban Heat Island assessment using medium- and low-resolution space-borne sensors: A review

Abstract The Urban Heat Island (UHI) phenomenon, a typical characteristic on urban landscapes, has been recognised as a key driver to the transformation of local climate. Reliable retrieval of urban and intra-urban thermal characteristics using satellite thermal data depends on accurate removal of the effects of atmospheric attenuations, angular and land surface emissivity. Several techniques have been proposed to retrieve land surface temperature (LST) from coarse resolution sensors. Medium spatial resolution sensors like the Advanced Space-borne Thermal Emission and Reflection Radiometer and the Landsat series offer a viable option for assessing LST within urban landscapes. This paper reviews the theoretical background of LST estimates from the thermal infrared part of the electromagnetic spectrum, LST retrieval algorithms applicable to each of the commonly used medium-resolution sensors and required variables for each algorithm. The paper also highlights LST validation techniques and concludes by stipulating the requirements for LST temporal and spatial configuration.