Doris Klein

Temporal segmentation of MODIS time series for improving crop classification in Central Asian irrigation systems

Crop cover and crop rotation mapping is an important and still evolving field in remote sensing science for which robust and highly automated processing chains are required. This study presents an improved mapping procedure for crop rotations of irrigated areas in Central Asia by using classification and regression trees (CARTs) applied to transformations of 250 m Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) time series. The time series were divided into several temporal segments, from which metrics were derived as input features for classification. This temporal aggregation was applied to suppress within-class temporal variability. Various lengths of temporal segments were tested for their potential to increase classification accuracy. In addition, tests of enhancing the classification accuracy were done by combining different classification results using the majority rule for voting. These different processing strategies were applied to four annual time series (2004–2007) of the Khorezm region, where 270 000 ha of irrigated land is dominated by rotations of cotton, wheat and rice. Improved classification results were obtained for CARTs applied to metrics derived from a mixture of different segment lengths. The sole use of either long or short temporal segments was inferior. CART prioritized segments representing active phases of the phenological development. The best result, the optimized segment-based approach, achieved an overall accuracy between 83 and 85% for classifications between 2004 and 2007; in particular, the small range demonstrated the robustness regarding inter-annual variations. These accuracies exceeded those of the original time series without temporal segmentation by 6–7%. With some adjustments to other crops and field heterogeneity influencing the usefulness of a respective sensor, the approach can be applied to other irrigation systems in Central Asia.

Derivation of leaf area index for grassland within alpine upland using multi-temporal RapidEye data

Biophysical parameters such as leaf area index (LAI) are key variables for vegetation monitoring and particularly important for modelling energy and matter fluxes in the biosphere. Therefore LAI has been derived from remote sensing data operationally based on data with a somewhat coarse spatial resolution. This study aims at deriving high-spatial resolution (6.5 m) multi-temporal LAI for grasslands based on RapidEye data by statistical regressions between vegetation indices (VIs) and field samplings. However, the suitability of those data for grassland LAI derivation has not been tested to date. Thus, the potential of RapidEye data in general and its red edge band in particular are investigated, as well as the robustness of the established relationships for different points in time. LAI was measured repeatedly over summer 2011 at about 30 different meadows in the Bavarian alpine upland using the LAI-2000 and correlated with VI values. The best relationships resulted from using the ratio vegetation index and red edge indices (NDVIrededge, rededge ratio index 1, and relative length) in non-linear models. Thus the indices based on the red edge channel improved regression modelling. The associated transfer functions achieved R2 values ranging from 0.57 to 0.85. The temporal transferability of those transfer functions to other dates was shown to be limited, with the root mean square errors (RMSEs) of several scenes exceeding one. However, when the LAI ranges are similar, a reliable transfer is possible: for example, the transfer of the regression function based on early autumn measurements showed RMSEs of only 0.77–0.95 for the other scenes except for the high-density stage in July, when the LAI reaches unprecedented maximal values. Also, the combination of multi-temporal training data shows no saturation of the selected indices and enables a satisfactory LAI mapping of different dates (RMSE = 0.59 – 1.02).

Mapping of large irrigated areas in Central Asia using MODIS time series

Remote sensing offers the opportunity to produce land cover classifications for large and remote areas on a yearly basis and is an important tool in regions that lack these information. However often training and validation data to generate annual land cover maps are not available in necessary quantity - being from one year only or covering only a small extent of the region of interest. This study was focused on land use classifications at regional scale with a special emphasize on annual updates under the constraint of limited sampling data. Often, sampling is reduced to one year or to an unrepresentative area extend within the region of interest. The investigations for the period between 2004 and 2009 were conducted in the irrigation systems of the Amu Darya Delta in Central Asia, where reliable information on crop rotations is required for sustainable land and water management. Annual training and validation data were extracted from high resolution land use classifications. For classification, statistical features based on MODIS time series of vegetation indices, reflectance and land surface temperature (LST) were calculated and a random forest algorithm was applied. By a combination of training data from different years, the accuracy could be enhanced from an overall accuracy of 70% to more than 90% for a focused subregion and also good consistency with high resolution images for the other parts of the delta, which has to be confirmed using quantitative validation. A combination of a different number of years was tested. Already two years can be sufficient to generate a robust and transferable random forest to produce yearly land use maps. The study shows the possibility to combine training data from different years for the annual classification of irrigated croplands on a regional scale.

Modeling of impervious surface in Germany using Landsat images and topographic vector data

The constant expansion of urban agglomerations in most countries is closely associated with a significant increase of impervious surface (IS). In Europe, robust and cost-effective methods for detection and quantification of IS on a continental or national scale are still rare. Thus, our study focuses on determining the percentage of impervious surface (PIS) for whole Germany based on a combined analysis of Landsat images and vector data on roads and railway networks, using Support Vector Machines (SVM) and GIS functionalities. We developed a procedure which provides functionalities for 1) the modeling of IS for built-up areas (PISB) based on optical earth observation data, 2) the combination of PISB with vector data providing additional information on small-scale infrastructure (PIST) and 3) the spatial aggregation of the combined product (PISBT) to the administrative units of municipalities. Compared to reference data sets of four cities, the results showed a mean absolute error of 19.4 % and a mean standard deviation of 17.3 %. The mean PIS of the total of residential, industrial and transportation-related areas in Germany comes up to 43.0 %, with a minimum in the federal state of Brandenburg (39.3 %) and a maximum in Hessen (46.1 %).

Monitoring of the 2015 Villarrica Volcano Eruption by Means of DLR’s Experimental TET-1 Satellite

Villarrica Volcano is one of the most active volcanoes in the South Andes Volcanic Zone. This article presents the results of a monitoring of the time before and after the 3 March 2015 eruption by analyzing nine satellite images acquired by the Technology Experiment Carrier-1 (TET-1), a small experimental German Aerospace Center (DLR) satellite. An atmospheric correction of the TET-1 data is presented, based on the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Emissivity Database (GDEM) and Moderate Resolution Imaging Spectroradiometer (MODIS) water vapor data with the shortest temporal baseline to the TET-1 acquisitions. Next, the temperature, area coverage, and radiant power of the detected thermal hotspots were derived at subpixel level and compared with observations derived from MODIS and Visible Infrared Imaging Radiometer Suite (VIIRS) data. Thermal anomalies were detected nine days before the eruption. After the decrease of the radiant power following the 3 March 2015 eruption, a stronger increase of the radiant power was observed on 25 April 2015. In addition, we show that the eruption-related ash coverage of the glacier at Villarrica Volcano could clearly be detected in TET-1 imagery. Landsat-8 imagery was analyzed for comparison. The information extracted from the TET-1 thermal data is thought be used in future to support and complement ground-based observations of active volcanoes.

High Temperature Fire Experiment for TET-1 and Landsat 8 in Test Site DEMMIN (Germany)

In 2012, the German Aerospace Center (DLR) launched the small satellite TET-1 (Experimental Technology Carrier) as a test platform for new satellite technologies and as a carrier for the Multi-Spectral Camera System (MSC) with five spectral bands (Green, Red, Near Infrared, Middle Infrared, and Thermal Infrared). The MSC has been designed to provide quantitative parameters (e.g. fire radiative power, burned area) observing high-temperature events. The detection of such events provides information for operational support to fire brigades, to change detection of hotspots, to assess CO2 emissions of burning vegetation, and, finally, contributes to the monitoring programs that support climate models. In order to investigate the sensitivity and accuracy of the MSC system, a calibration and validation fire campaign was developed and executed, to derive characteristic signal changes of corresponding pixels in the MWIR and LWIR bands. The planning and execution of the validation campaign and the results are presented.

Comparison of leaf area indices for grasslands within the Alpine upland based on multi-scale satellite data time series and radiation transfer modeling

In this study, the leaf area index (LAI) of grasslands in the Bavarian Alpine uplands has been derived using inverted radiation transfer modeling (RTM) on original as well as simulated remote sensing data time series. The spatial resolutions of the data sets range from 6.5 to 250 m. While the high resolution data are available for four points in the vegetation period, the medium resolution time series consist of weekly scenes. The aim is to investigate the performance of the inverse RTM when applied to satellite data of different spatial and temporal resolutions. Further, we determine the adequate resolutions of remote sensing data for LAI retrieval in a heterogeneous landscape. All results were validated using in situ measurements. While the algorithm proves to be generally applicable in this challenging landscape on different scales, retrieval accuracy increases with higher spatial resolution. Satellite images with a spatial resolution up to 20 m are identified as a good compromise between accurate results and spatial detail. The 250 m resolution LAI time series on the other hand provides valuable information on the phenology and sudden LAI reductions caused by harvest, which are not captured by the high spatial resolution time series with few scenes.

Land Surface Phenology from MODIS data in Germany

Changes in the timing of phenological events are a critical parameter for quantifying the impact of climate alterations on vegetation development that in reverse alters patterns of biosphere-atmosphere interactions. By the means of remote sensing, numerous approaches have been developed for deriving Land Surface Phenology. This study aims at identifying the best combination of data quality integration, filtering and threshold choice for the Start of Season (SOS) in Germany using MODIS data. The resulting remote sensing based SOS is related to phenological ground observations. For the year 2010, SOS data was produced by 5 different methods and the best method was identified by the comparison of the SOS dates to ground observations of the leaf unfolding of Broad-Leaved forest and beginning of turning green for Pastures, respectively. The results show that the impact of time series data quality screening, filter choice and threshold setting differs among the selected land cover classes.

Comparison of leaf area index derived by statistical relationships and inverse radiation transport modeling using RapidEye data in the European alpine upland

Leaf Area Index (LAI) is a relevant input parameter for flux modeling of energy and matter in the biosphere. However, in a landscape such as the European alpine upland with small-scale land use patterns and high vegetation heterogeneity, existing global products are less suited and a high spatial resolution is required. Within this study two methods are compared to derive the LAI for grassland in the prealpine River Ammer catchment from high spatial resolution RapidEye data: the empirical approach based on regression functions, and the physical approach of inverted radiation transfer modeling (RTM). Established vegetation indices (VIs) as well as new ones incorporating RapidEye’s red edge band are calculated for four dates of the vegetation period 2011 and correlated with in situ LAI data. The statistical regressions between VIs and LAI of the different time steps show high correlations (R2 of 0.57 up to 0.85). However, the established regressions are scene specific and the method requires excessive field work. In the physical approach the RapidEye reflectances are used as input data to an inverted RTM (PROSAIL), which is parameterized with leaf and canopy properties collected in the field. The LAI derived by the RTM have a RMSE between 2.02 and 2.28 for the different dates. Both methods capture the general LAI pattern. However, due to the broad parameterization of the RTM used to cover the heterogeneous grassland conditions, resulting LAI values are generally higher than the statistically derived LAI values.

Assessment of urban extent and imperviousness of Cape Town using TerraSAR-X and Landsat images

The worldwide urban growth leads to an increase in impervious surface, which in turn has many negative consequences for the environment. For an assessment of these phenomena in this study first TerraSAR-X radar data are classified using a knowledge-based approach to detect the extent of urban areas. Subsequently within this area the percentage of imperviousness is estimated by using a Support Vector Regression model with optical Landsat images and high resolution aerial pictures. These methods were developed for urban areas in Germany and transferred to Cape Town, South Africa. The overall accuracy of the settlement detection is 82.3 % and the mean error of the percentage of imperviousness is 14.1 % with a local regression model. It was also possible to apply a model generated for the German city of Munich for Cape Town but the absolute mean error increased to 32.8 %, indicating the necessity to further improve the radiometric adjustment.