As a cradle of ancient Chinese civilization, the Yellow River Basin has a very long human-environment interrelationship, where early anthropogenic activities re- sulted in large scale landscape modifications. Today, the impact of this relationship has intensified further as the basin plays a vital role for China’s continued economic development. It is one of the most densely-populated, fastest growing, and most dynamic regions of China with abundant natural and environmental resources providing a livelihood for almost 190 million people. Triggered by fundamental economic reforms, the basin has witnessed a spectacular economic boom during the last decades and can be considered as an exemplary blueprint region for contemporary dynamic Global Change processes occurring throughout the country, which is currently transitioning from an agrarian-dominated economy into a modern urbanized society. However, this resourcesdemanding growth has led to profound land use changes with adverse effects on the Yellow River social-ecological systems, where complex challenges arise threatening a long-term sustainable development. Consistent and continuous remote sensing-based monitoring of recent and past land cover and land use change is a fundamental requirement to mitigate the adverse impacts of Global Change processes. Nowadays, technical advancement and the multitude of available satellite sensors, in combination with the opening of data archives, allow the creation of new research perspectives in regional land cover applications over heterogeneous landscapes at large spatial scales. Despite the urgent need to better understand the prevailing dynamics and underlying factors influencing the current processes, detailed regional specific land cover data and change information are surprisingly absent for this region. In view of the noted research gaps and contemporary developments, three major objectives are defined in this thesis. First (i), the current and most pressing social-ecological challenges are elaborated and policy and management instruments towards more sustainability are discussed. Second (ii), this thesis provides new and improved insights on the current land cover state and dynamics of the entire Yellow River Basin. Finally (iii), the most dominant processes related to mining, agriculture, forest, and urban dynamics are determined on finer spatial and temporal scales. The complex and manifold problems and challenges that result from long-term abuse of the water and land resources in the basin have been underpinned by policy choices, cultural attitude, and institutions that have evolved over centuries in China. The tremendous economic growth that has been mainly achieved by extracting water and exploiting land resources in a rigorous, but unsustainable manner, might not only offset the economic benefits, but could also foster social unrest. Since the early emergence of the first Chinese dynasties, flooding was considered historically as a primary issue in river management and major achievements have been made to tame the wild nature of the Yellow River. Whereas flooding is therefore largely now under control, new environmental and social problems have evolved, including soil and water pollution, ecological degradation, biodiversity decline, and food security, all being further aggravated by anthropogenic climate change. To resolve the contemporary and complex challenges, many individual environmental laws and regulations have been enacted by various Chinese ministries. However, these policies often pursue different, often contradictory goals, are too general to tackle specific problems and are usually implemented by a strong top-down approach. Recently, more flexible economic and market-based incentives (pricing, tradable permits, investments) have been successfully adopted, which are specifically tailored to the respective needs, shifting now away from the pure command and regulating instruments. One way towards a more holistic and integrated river basin management could be the establishment of a common platform (e.g. a Geographical Information System) for data handling and sharing, possibly operated by the Yellow River Basin Conservancy Commission (YRCC), where available spatial data, statistical information and in-situ measures are coalesced, on which sustainable decision-making could be based. So far, the collected data is hardly accessible, fragmented, inconsistent, or outdated. The first step to address the absence and lack of consistent and spatially up-to-date information for the entire basin capturing the heterogeneous landscape conditions was taken up in this thesis. Land cover characteristics and dynamics were derived from the last decade for the years 2003 and 2013, based on optical medium-resolution hightemporal MODIS Normalized Differenced Vegetation Index (NDVI) time series at 250 m. To minimize the inherent influence of atmospheric and geometric interferences found in raw high temporal data, the applied adaptive Savitzky-Golay filter successfully smoothed the time series and substantially reduced noise. Based on the smoothed time series data, a large variety of intra-annual phenology metrics as well as spectral and multispectral annual statistics were derived, which served as input variables for random forest (RF) classifiers. High quality reference data sets were derived from very high resolution imagery for each year independently of which 70 % trained the RF models. The accuracy assessments for all regionally specific defined thematic classes were based on the remaining 30 % reference data split and yielded overall accuracies of 87 % and 84 % for 2003 and 2013, respectively. The first regional adapted Yellow River Land Cover Products (YRB LC) depict the detail spatial extent and distribution of the current land cover status and dynamics. The novel products overall differentiate overall 18 land cover and use classes, including classes of natural vegetation (terrestrial and aquatic), cultivated classes, mosaic classes, non-vegetated, and artificial classes, which are not presented in previous land cover studies so far. Building on this, an extended multi-faceted land cover analysis on the most prominent land cover change types at finer spatial and temporal scales provides a better and more detailed picture of the Yellow River Basin dynamics. Precise spatio-temporal products about mining, agriculture, forest, and urban areas were examined from long-trem Landsat satellite time series monitored at annual scales to capture the rapid rate of change in four selected focus regions. All archived Landsat images between 2000 and 2015 were used to derive spatially continuous spectral-temporal, multi-spectral, and textural metrics. For each thematic region and year RF models were built, trained and tested based on a stablepixels reference data set. The automated adaptive signature (AASG) algorithm identifies those pixels that did not change between the investigated time periods to generate a mono-temporal reference stable-pixels data set to keep manual sampling requirements to a minimum level. Derived results gained high accuracies ranging from 88 % to 98 %. Throughout the basin, afforestation on the Central Loess Plateau and urban sprawl are identified as most prominent drivers of land cover change, whereas agricultural land remained stable, only showing local small-scale dynamics. Mining operations started in 2004 on the Qinghai-Tibet Plateau, which resulted in a substantial loss of pristine alpine meadows and wetlands. In this thesis, a novel and unique regional specific view of current and past land cover characteristics in a complex and heterogeneous landscape was presented by using a multi-source remote sensing approach. The delineated products hold great potential for various model and management applications. They could serve as valuable components for effective and sustainable land and water management to adapt and mitigate the predicted consequences of Global Change processes.

Rapid population growth in West Africa has led to expansion in croplands due to the need to grow more food to meet the rising food demand of the burgeoning population. These expansions negatively impact the sub-region's ecosystem, with implications for water and soil quality, biodiversity and climate. In order to appropriately monitor the changes in croplands and assess its impact on the ecosystem and other environmental processes, accurate and up-to-date information on agricultural land use is required. But agricultural land use mapping (i.e. mapping the spatial distribution of crops and croplands) in West Africa has been challenging due to the unavailability of adequate satellite images (as a result of excessive cloud cover), small agricultural fields and a heterogeneous landscape. This study, therefore, investigated the possibilities of improving agricultural land use mapping by utilizing optical satellite images with higher spatial and temporal resolution as well as images from Synthetic Aperture Radar (SAR) systems which are near-independent of weather conditions. The study was conducted at both watershed and regional scales. At watershed scale, classification of different crop types in three watersheds in Ghana, Burkina Faso and Benin was conducted using multi-temporal: (1) only optical images (RapidEye) and (2) optical plus dual polarimetric (VV/VH) SAR images (TerraSAR-X). In addition, inter-annual or short term (2-3 years) changes in cropland area in the past ten years were investigated using historical Landsat images. Results obtained indicate that the use of only optical images to map different crop types in West Africa can achieve moderate classification accuracies (57% to 71%). Overlaps between the cropping calendars of most crops types and certain inter-croppings pose a challenge to optical images in achieving an adequate separation between those crop classes. Integration of SAR images, however, can improve classification accuracies by between 8 and 15%, depending on the number of available images and their acquisition dates. The sensitivity of SAR systems to different crop canopy architectures and land surface characteristics improved the separation between certain crop types. The VV polarization of TerraSAR-X was found to better discrimination between crop types than the VH. Images acquired between August and October were found to be very useful for crop mapping in the sub-region due to structural differences in some crop types during this period. At the regional scale, inter-annual or short term changes in cropland area in the Sudanian Savanna agro-ecological zone in West Africa were assessed by upscaling historical cropland information derived at the watershed scale (using Landsat imagery) unto a coarse spatial resolution, but geographically large, satellite imagery (MODIS) using regression based modeling. The possibility of using such regional scale cropland information to improve government-derived agricultural statistics was investigated by comparing extracted cropland area from the fractional cover maps with district-level agricultural statistics from Ghana The accuracy of the fractional cover maps (MAE between 14.2% and 19.1%) indicate that the heterogeneous agricultural landscape of West Africa can be suitably represented at the regional or continental scales by estimating fractional cropland cover on low resolution Analysis of the results revealed that cropland area in the Sudanian Savanna zone has experienced inter-annual or short term fluctuations in the past ten years due to a variety of factors including climate factors (e.g. floods and droughts), declining soil fertility, population increases and agricultural policies such as fertilizer subsidies. Comparison of extracted cropland area from the fractional cover maps with government's agricultural statistics (MoFA) for seventeen districts (second administrative units) in Ghana revealed high inconsistencies in the government statistics, and highlighted the potential of satellite derived cropland information at regional scales to improve national/sub-national agricultural statistics in West Africa. The results obtained in this study is promising for West Africa, considering the recent launch of optical (Landsat 8) and SAR sensors (Sentinel-1) that will provide free data for crop mapping in the sub-region. This will improve chances of obtaining adequate satellite images acquired during the cropping season for agricultural land use mapping and bolster opportunities of operationalizing agricultural land use mapping in West Africa. This can benefit a wide range of biophysical and economic models and improve decision making based on their results.

Agriculture is mankind’s primary source of food production and plays the key role for cereal supply to humanity. One of the future challenges will be to feed a constantly growing population, which is expected to reach more than nine billion by 2050. The potential to expand cropland is limited, and enhancing agricultural production efficiency is one important means to meet the future food demand. Hence, there is an increasing demand for dependable, accurate and comprehensive agricultural intelligence on crop production. The value of satellite earth observation (EO) data for agricultural monitoring is well recognized. One fundamental requirement for agricultural monitoring is routinely updated information on crop acreage and the spatial distribution of crops. With the technical advancement of satellite sensor systems, imagery with higher temporal and finer spatial resolution became available. The classification of such multi-temporal data sets is an effective and accurate means to produce crop maps, but methods must be developed that can handle such large and complex data sets. Furthermore, to properly use satellite EO for agricultural production monitoring a high temporal revisit frequency over vast geographic areas is often necessary. However, this often limits the spatial resolution that can be used. The challenge of discriminating pixels that correspond to a particular crop type, a prerequisite for crop specific agricultural monitoring, remains daunting when the signal encoded in pixels stems from several land uses (mixed pixels), e.g. over heterogeneous landscapes where individual fields are often smaller than individual pixels. The main purposes of the presented study were (i) to assess the influence of input dimensionality and feature selection on classification accuracy and uncertainty in object-based crop classification, (ii) to evaluate if combining classifier algorithms can improve the quality of crop maps (e.g. classification accuracy), (iii) to assess the spatial resolution requirements for crop identification via image classification. Reporting on the map quality is traditionally done with measures that stem from the confusion matrix based on the hard classification result. Yet, these measures do not consider the spatial variation of errors in maps. Measures of classification uncertainty can be used for this purpose, but they have attained only little attention in remote sensing studies. Classifier algorithms like the support vector machine (SVM) can estimate class memberships (the so called soft output) for each classified pixel or object. Based on these estimations, measures of classification uncertainty can be calculated, but it has not been analysed in detail, yet, if these are reliable in predicting the spatial distribution of errors in maps. In this study, SVM was applied for the classification of agricultural crops in irrigated landscapes in Middle Asia at the object-level. Five different categories of features were calculated from RapidEye time series data as classification input. The reliability of classification uncertainty measures like entropy, derived from the soft output of SVM, with regard to predicting the spatial distribution of error was evaluated. Further, the impact of the type and dimensionality of the input data on classification uncertainty was analysed. The results revealed that SMVs applied to the five feature categories separately performed different in classifying different types of crops. Incorporating all five categories of features by concatenating them into one stacked vector did not lead to an increase in accuracy, and partly reduced the model performance most obviously because of the Hughes phenomena. Yet, applying the random forest (RF) algorithm to select a subset of features led to an increase of classification accuracy of the SVM. The feature group with red edge-based indices was the most important for general crop classification, and the red edge NDVI had an outstanding importance for classifying crops. Two measures of uncertainty were calculated based on the soft output from SVM: maximum a-posteriori probability and alpha quadratic entropy. Irrespective of the measure used, the results indicate a decline in classification uncertainty when a dimensionality reduction was performed. The two uncertainty measures were found to be reliable indicators to predict errors in maps. Correctly classified test cases were associated with low uncertainty, whilst incorrectly test cases tended to be associated with higher uncertainty. The issue of combining the results of different classifier algorithms in order to increase classification accuracy was addressed. First, the SVM was compared with two other non-parametric classifier algorithms: multilayer perceptron neural network (MLP) and RF. Despite their comparatively high classification performance, each of the tested classifier algorithms tended to make errors in different parts of the input space, e.g. performed different in classifying crops. Hence, a combination of the complementary outputs was envisaged. To this end, a classifier combination scheme was proposed, which is based on existing algebraic operators. It combines the outputs of different classifier algorithms at the per-case (e.g. pixel or object) basis. The per-case class membership estimations of each classifier algorithm were compared, and the reliability of each classifier algorithm with respect to classifying a specific crop class was assessed based on the confusion matrix. In doing so, less reliable classifier algorithms were excluded at the per-class basis before the final combination. Emphasis was put on evaluating the selected classification algorithms under limiting conditions by applying them to small input datasets and to reduced training sample sets, respectively. Further, the applicability to datasets from another year was demonstrated to assess temporal transferability. Although the single classifier algorithms performed well in all test sites, the classifier combination scheme provided consistently higher classification accuracies over all test sites and in different years, respectively. This makes this approach distinct from the single classifier algorithms, which performed different and showed a higher variability in class-wise accuracies. Further, the proposed classifier combination scheme performed better when using small training set sizes or when applied to small input datasets, respectively. A framework was proposed to quantitatively define pixel size requirements for crop identification via image classification. That framework is based on simulating how agricultural landscapes, and more specifically the fields covered by one crop of interest, are seen by instruments with increasingly coarser resolving power. The concept of crop specific pixel purity, defined as the degree of homogeneity of the signal encoded in a pixel with respect to the target crop type, is used to analyse how mixed the pixels can be (as they become coarser) without undermining their capacity to describe the desired surface properties (e.g. to distinguish crop classes via supervised or unsupervised image classification). This tool can be modulated using different parameterizations to explore trade-offs between pixel size and pixel purity when addressing the question of crop identification. Inputs to the experiments were eight multi-temporal images from the RapidEye sensor. Simulated pixel sizes ranged from 13 m to 747.5 m, in increments of 6.5 m. Constraining parameters for crop identification were defined by setting thresholds for classification accuracy and uncertainty. Results over irrigated agricultural landscapes in Middle Asia demonstrate that the task of finding the optimum pixel size did not have a “one-size-fits-all” solution. The resulting values for pixel size and purity that were suitable for crop identification proved to be specific to a given landscape, and for each crop they differed across different landscapes. Over the same time series, different crops were not identifiable simultaneously in the season and these requirements further changed over the years, reflecting the different agro-ecological conditions the investigated crops were growing in. Results further indicate that map quality (e.g. classification accuracy) was not homogeneously distributed in a landscape, but that it depended on the spatial structures and the pixel size, respectively. The proposed framework is generic and can be applied to any agricultural landscape, thereby potentially serving to guide recommendations for designing dedicated EO missions that can satisfy the requirements in terms of pixel size to identify and discriminate crop types. Regarding the operationalization of EO-based techniques for agricultural monitoring and its application to a broader range of agricultural landscapes, it can be noted that, despite the high performance of existing methods (e.g. classifier algorithms), transferability and stability of such methods remain one important research issue. This means that methods developed and tested in one place might not necessarily be portable to another place or over several years, respectively. Specifically in Middle Asia, which was selected as study region in this thesis, classifier combination makes sense due to its easy implementation and because it enhanced classification accuracy for classes with insufficient training samples. This observation makes it interesting for operational contexts and when field reference data availability is limited. Similar to the transferability of methods, the application of only one certain kind of EO data (e.g. with one specific pixel size) over different landscapes needs to be revisited and the synergistic use of multi-scale data, e.g. combining remote sensing imagery of both fine and coarse spatial resolution, should be fostered. The necessity to predict and control the effects of spatial and temporal scale on crop classification is recognized here as a major goal to achieve in EO-based agricultural monitoring.

Remote sensing time series is the collection or acquisition of remote sensing data in a fixed equally spaced time period over a particular area or for the whole world. Near daily high spatial resolution data is very much needed for remote sensing applications such as agriculture monitoring, phenology change detection, environmental monitoring and so on. Remote sensing applications can produce better and accurate results if they are provided with dense and accurate time series of data. The current remote sensing satellite architecture is still not capable of providing near daily or daily high spatial resolution images to fulfill the needs of the above mentioned remote sensing applications. Limitations in sensors, high development, operational costs of satellites and presence of clouds blocking the area of observation are some of the reasons that makes near daily or daily high spatial resolution optical remote sensing data highly challenging to achieve. With developments in the optical sensor systems and well planned remote sensing satellite constellations, this condition can be improved but it comes at a cost. Even then the issue will not be completely resolved and thus the growing need for high temporal and high spatial resolution data cannot be fulfilled entirely. Because the data collection process relies on satellites which are physical system, these can fail unpredictably due to various reasons and cause a complete loss of observation for a given period of time making a gap in the time series. Moreover, to observe the long term trend in phenology change due to rapidly changing environmental conditions, the remote sensing data from the present is not just sufficient, the data from the past is also important. A better alternative solution for this issue can be the generation of remote sensing time series by fusing data from multiple remote sensing satellite which has different spatial and temporal resolutions. This approach will be effective and efficient. In this method a high temporal low spatial resolution image from a satellite such as Sentinel-2 can be fused with a low temporal and high spatial resolution image from a satellite such as the Sentinel-3 to generate a synthetic high temporal high spatial resolution data. Remote sensing time series generation by data fusion methods can be applied to the satellite images captured currently as well as the images captured by the satellites in the past. This will provide the much needed high temporal and high spatial resolution images for remote sensing applications. This approach with its simplistic nature is cost effective and provides the researchers the means to generate the data needed for their application on their own from the limited source of data available to them. An efficient data fusion approach in combination with a well planned satellite constellation can offer a solution which will ensure near daily time series of remote sensing data with out any gap. The aim of this research work is to develop an efficient data fusion approaches to achieve dense remote sensing time series.