Remote Sensing Time Series

The face of our planet is changing at an unprecedented rate. Forest ecosystems diminish at alarming speed, urban and agricultural areas expand into the surrounding natural space, aquaculture is spreading, sea level rise leads to changes in coastal ecosystems, and even without obvious land cover change, land use intensity may change and complex ecosystems may undergo transient changes in composition. Satellite based earth observation is a powerful means to monitor these changes, and especially time series analysis holds the potential to reveal long term land surface dynamics. Whereas in past decades time series analysis was an elaborate undertaking mostly performed by a limited number of experts using coarse resolution data, attention shifts nowadays to open source tools and novel techniques for analyzing time series and the utilization of the same for numerous environmental applications. The reasons are the pressing call for climate-relevant, long term data analyses and value added products revealing past land surface dynamics and trends, the growing demand for global data sets, and the opening up of multidecadal remote sensing data archives, all at a time of considerably-improved hardware power, computer literacy, and a general trend towards cloud solutions and available open source algorithms and programming languages. This chapter presents a comprehensive overview of time series analysis. We introduce currently orbiting optical, radar, and thermal infrared sensors and elucidate which of them are suitable for long term monitoring tasks based on remote sensing time series analysis. We briefly summarize the theoretical concept of time series components and important seasonal statistical features and list the types of variables usually analyzed as time series. Furthermore, we address data related, sensor related, location related, and processing related challenges of time series analysis. Lastly, we assess current developments and upcoming opportunities.

Dense time series of optical remote sensing data have long been the domain of broad-scale sensors with daily near-global coverage, such as the Advanced Very High Resolution Radiometer (AVHRR), the Medium Resolution Imaging Spectrometer (MERIS), the Moderate Resolution Imaging Spectrometer (MODIS) or the Satellite Pour l’Observation de la Terre (SPOT) VEGETATION. More recently, satellite data suitable for fine-scale analyses are becoming attractive for time series approaches. The major reasons for this development are the opening of the United States Geological Survey (USGS) Landsat archive along with a standardized geometric pre-processing including terrain correction. Based on such standardized products, tools for automated atmospheric correction and cloud/cloud shadow masking advanced the capabilities to handle cloud-contamination effectively. Finally, advances in information technology for mass data processing today allow analysing thousands of satellite images with comparatively little effort. Based on these major advancements, time series analyses have become feasible for solving questions across different research domains, while the focus here is on land systems. While early studies focused on better characterising forested ecosystems, now more complex ecosystem regimes, such as shrubland or agricultural system dynamics, come into focus. Despite the evolution of a wealth of novel time series-based applications, coherent analysis schemes and good practice guidelines are scarce. This chapter accordingly strives to structure the different approaches with a focus on potential applications or user needs. We end with an outlook on forthcoming sensor constellations that will greatly advance our opportunities concerning time series analyses.

Since the early 1980s, the German Remote Sensing Data Centre (DFD) of the German Aerospace Centre (DLR) has received archived and processed Advanced Very High Resolution Radiometer (AVHRR) data from the Polar Orbiting Environmental Satellites (POES) of the National Oceanic and Atmospheric Administration (NOAA). By December 2013, over 237,000 paths over Europe have since been archived at DLR. Based on these High Resolution Picture Transmission (HRPT) raw datasets, an operational pre-processing and value-adding chain has been developed (Dech et al., Aerosp Sci Technol 2(5):335–346, 1998; Tungalagsaikhan et al., Proc. 23th DGPF (12), 2003). In this chapter, the series of AVHRR sensors is introduced, and information on calibration and system correction procedures is given. Next, the pre-processing part of DLR’s processing chain is described, where focus is set on the calibration aspects. Time series examples are provided to show the influence of changes in calibration over time, and to illustrate the need for consistent pre-processing and data harmonization. According to these requirements DLR’s multi-decadal archive of AVHRR data will be re-processed in the frame of the TIMELINE project, providing consistent and well-calibrated time series data.

Snow cover is one of the most dynamic land cover parameters that can be monitored from space and plays an important role for the Earth’s climate system and hydrological circle. While the spatial extent can be limited to narrow mountain ridges during summer, the snow cover percentage on the Northern Hemisphere may exceed 50 % (Lemke et al., Observations: changes in snow, ice and frozen ground. In: Solomon S, Qin D, Manning M, Chen Z, Marquis MC, Averyt K, Tignor M, Miller HL (eds) Climate change 2007: the physical science basis. Contributions of Working Group 1 to the fourth assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge and New York, pp 337–383, 2007) of the total land surface (~45 million km²) during winter seasons (Barry et al., Global outlook for ice & snow. United Nations Environment Programme, Hertfordshire, 2007). Remote sensing has been used since the early 1970s to map terrestrial snow cover (Brown, J Clim 13:2339–2355, 2000) and both – sensors as well as retrieval algorithms – have undergone a substantial development since that time. This chapter will give a short introduction on how snow cover can be monitored from space. Furthermore, techniques will be outlined that show how time series analyses can be applied to remotely sensed snow cover products to reduce the compromising effect of cloud cover and to investigate the fundamental characteristics of snow. Time series of snow cover data allow for various analyses covering the fields of hydrology, climate research, flood prediction and management, and economy. Short term variations and extreme events can be analysed as well as long term climatological trends, constituting time series of snow cover data a valuable tool for a large bandwidth of applications.

The knowledge and understanding of intra- and inter annual characteristics of inland water bodies, such as natural lakes and artificial reservoirs are crucial for many reasons. Inland water bodies are sensitive to environmental variations and human impact which is reflected in spatial and temporal dynamics of surface extent. A time-series of areal surface extent of lakes and reservoirs might be a helpful dataset to understand the complex system and the spatio-temporal patterns of natural lakes and artificial reservoirs. In this study, we describe an approach to detect water bodies based on dynamical thresholding on daily basis and utilizing high frequency observations. Daily MODIS (Moderate Resolution Imaging Spectrometer) products were used to generate water masks for the year 2013 on global scale. The results indicate that time series of water bodies’ extent are important especially for those inland water bodies which are dominated by temporal changes and fluctuation through the year. In combination with ancillary data, our understanding of environmental and human interaction and the reaction of water bodies will be improved. Such information is critical to support sustainable water management, as well as for climate change discussion since many inland water bodies are sensitive to short- and long term environmental alterations.

Land surface temperature (LST) is an important parameter in the climate system, impacting vegetation development, snow cover, runoff, and human livelihoods. Knowledge of LST dynamics can furthermore be used as an indicator for climate variability and change. LST is regularly measured from satellite sensors on a broad spatial scale and with a high temporal resolution. In this research, MODIS (Moderate Resolution Imaging Spectroradiometer) sensor data are used to assess the spatial and temporal patterns of LST in the Mekong Basin (MB) including its temporal variability. The dataset contains 13 years of MODIS LST data, a unique measurement time series in terms of resolution, accuracy, and homogeneity. In the analysis a temporal granularity of 8-days was used. The MB was divided into six physiographically homogenous regions. The height and magnitude of annual LST curves differ between the regions and prove to be strongly dependent on the topography. Large intra-annual magnitudes and low temperatures (daytime/nighttime annual regional means are 14 °C/−7 °C) are found in the northern areas, mainly in the high-lying Tibetan Plateau. The more southern areas are characterized by low LST seasonality and high temperatures (daytime/nighttime annual regional means of these regions range from 25 °C to 30 °C/19 °C to 25 °C). The year-to-year variability of LST is similar in all regions (regional weekly daytime/nighttime deviations lower than 4 °C/6 °C, except for the Tibetan Plateau, where regional weekly daytime/nighttime deviations reach 6 °C/18 °C). In summer, 42 % of daytime LST could be explained by topographic height. In winter and in nighttime scenes, topography explained 89–97 % of the LST distribution. Land use and land use change further influence the LST pattern, mainly in the daytime. An example of rising LST due to deforestation is given. This study allows for an improved understanding of temperature dynamics in one of the world’s largest river basins.

Large volumes of data from satellite sensors with high time-resolution exist today, e.g. Advanced Very High Resolution Radiometer (AVHRR) and Moderate Resolution Imaging Spectroradiometer (MODIS), calling for efficient data processing methods. TIMESAT is a free software package for processing satellite time-series data in order to investigate problems related to global change and monitoring of vegetation resources. The assumptions behind TIMESAT are that the sensor data represent the seasonal vegetation signal in a meaningful way, and that the underlying vegetation variation is smooth. A number of processing steps are taken to transform the noisy signals into smooth seasonal curves, including fitting asymmetric Gaussian or double logistic functions, or smoothing the data using a modified Savitzky-Golay filter. TIMESAT can adapt to the upper envelope of the data, accounting for negatively biased noise, and can take missing data and quality flags into account. The software enables the extraction of seasonality parameters, like the beginning and end of the growing season, its length, integrated values, etc. TIMESAT has been used in a large number of applied studies for phenology parameter extraction, data smoothing, and general data quality improvement. To enable efficient analysis of future Earth Observation data sets, developments of TIMESAT are directed towards processing of high-spatial resolution data from e.g. Landsat and Sentinel-2, and use of spatio-temporal data processing methods.

This chapter summarizes approaches to the detection of dryland vegetation change and methods for observing spatio-temporal trends from space. An overview of suitable long-term Earth Observation (EO) based datasets for assessment of global dryland vegetation trends is provided and a status map of contemporary greening and browning trends for global drylands is presented. The vegetation metrics suitable for per-pixel temporal trend analysis is discussed, including seasonal parameterisation and the appropriate choice of trend indicators. Recent methods designed to overcome assumptions of long-term linearity in time series analysis (Breaks For Additive Season and Trend(BFAST)) are discussed. Finally, the importance of the spatial scale when performing temporal trend analysis is introduced and a method for image downscaling (Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM)) is presented.

This chapter summarizes methods of inferring information about drivers of global dryland vegetation changes observed from remote sensing time series data covering from the 1980s until present time. Earth observation (EO) based time series of vegetation metrics, sea surface temperature (SST) (both from the AVHRR (Advanced Very High Resolution Radiometer) series of instruments) and precipitation data (blended satellite/rain gauge) are used for determining the mechanisms of observed changes. EO-based methods to better distinguish between climate and human induced (land use) vegetation changes are reviewed. The techniques presented include trend analysis based on the Rain-Use Efficiency (RUE) and the Residual Trend Analysis (RESTREND) and the methodological challenges related to the use of these. Finally, teleconnections between global sea surface temperature (SST) anomalies and dryland vegetation productivity are illustrated and the associated predictive capabilities are discussed.

Phenological change and variation have become increasingly relevant topics in global change science due to recognition of their importance for ecosystem functioning and biogeophysical processes. Remote sensing time series offer great potential for assessing phenological dynamics at landscape, regional and global scales. Even though a number of studies have investigated phenology, mostly with a focus on climatic variability, we do not yet have a detailed understanding of phenological cycles and respective biogeographical patterns. This is particularly true for biomes like the tropical savannas, which cover approximately one eighth of the global land surface. Savannas are often characterized by high human population density and growth, one example being the West African Sudanian Savanna. The phenological characteristics in these regions can be assumed to be particularly influenced by agricultural land use and fires, in addition to climatic variability. This study analyses the spatio-temporal patterns of land surface phenology in a Sudanian Savanna landscape of southern Burkina Faso based on time series of the Moderate Resolution Spectroradiometer (MODIS), and on multi-temporal Landsat data. The analyses focus on influences of fire, land use, and vegetation structure on phenological patterns, and disclose the effects of long-term fire frequency, as well as the short-term effects of burning on the vegetation dynamics observed in the following growing season. Possibilities of further improvements for remote sensing based analyses of land surface phenology are seen in using earth observation datasets of increased spatial and temporal resolution as well as in linking phenological metrics from remote sensing with actual biological events observed on the ground.

Aboveground net primary productivity (ANPP) is limited by water availability especially in dry and desert regions, and many studies have linked ANPP to current and previous “effective” rainfall events. In this study a distributed lag model (DLM) was used to assess the impact of current and previous 16 day rainfall anomalies on the Enhanced Vegetation Index (EVI) as a proxy for ANPP in the Okavango catchment (South Africa). The two important aspects in using DLMs are the explained total ANPP variability by the rainfall regime and the duration of that dependency. The results indicate that more than 50 % of the Okavango Basin are sensitive towards current and previous rainfall anomalies. These regions are mainly restricted to the southern semi-arid parts of the catchment, whereas in the humid and sub-humid northern areas significant correlations were observed only locally. Here, the dominant land cover classes are shrub- and grassland, thornbush savannahs and mixed woodlands. The duration of significant rainfall-EVI dependencies ranges from concurrent anomalies to a time-shift of 3.5 months. A logistic regression model was applied to discriminate among the sensitive and non-sensitive areas in the basin in terms of possible physiogeographic covariates. The model was able to correctly classify ~80 % of the available pixels. Most relevant explanatory covariates were evaporation, elevation and land cover.

Dry regions such as arid southern Africa are strained by unfavourable climatic conditions. Intensive land use as rangeland and for livestock farming leads to additional encroachment of these ecosystems. The consequence of this long-time stress is degradation in terms of loss of the vegetative cover and productivity. Albeit these are known facts there is still a lack of objectiveness in the long term assessment of degradation on a larger scale. We present a method of applying remote sensing time-series in a vegetation model that helps to fill this gap. The approach is based on time-series of the vegetative productivity computed by our vegetation model BETHY/DLR (Biosphere Energy Transfer Hydrology Model). The used data included SPOT-VGT LAI (Leaf area index) and ECMWF meteorology time-series for the period of 1999–2010. The trend-analysis of model output and climatic input results in a new land degradation index (LDI) that distinguishes between climatic and human-induced reduction of vegetative productivity.

Net primary productivity (NPP) is an important environmental indicator that provides information about vegetation productivity and carbon fluxes. Analyses of NPP time-series allow for understanding temporal patterns and changes in vegetation productivity. These are especially important in rapidly changing environments, such as China, the world’s third largest country. In this study, we use the model BETHY/DLR (Biosphere Energy Transfer Hydrology Model) for derivation of NPP time-series for China for 14 years from 1999–2012. We analyse spatial and temporal NPP distributions. These include mean annual NPP distribution and mean productivities for different land cover classes. Monthly data provide information about temporal patterns of vegetation productivity for different regions in China and different vegetation types. Analyses of interannual NPP variability revealed considerable differences in the development of annual vegetation productivity within the analysed time period for different provinces. The decrease in NPP for the district Shanghai shows the strong influence of one of Asia’s fastest growing megacities on the environment. The NPP time-series was additionally analysed for a forest region in North China, which has been affected by forest disturbances. Our results show that the NPP data are suitable for monitoring of forest disturbance and regrowth. The analyses and results presented in this study provide valuable information about spatial and temporal variation of vegetation productivity in the various regions within China.

The utility of satellite time series data for monitoring land surface change is well established. This chapter highlights recent Landsat research, product developments, and opportunities, for global long term Landsat monitoring, that are now evolving rapidly with the opening of the Landsat archive. Specifically, it introduces the NASA (National Aeronautics and Space Administration) funded global Web Enabled Landsat Data products, and overviews Landsat time series phenology and land cover change monitoring applications and research, and prospectives for Landsat time series monitoring.

Remote sensing is a core tool for forest monitoring. Landsat data has been widely used in forest change detection studies but many approaches lack capabilities such as assessing changes for long temporal sequences. Moreover, most methods are not capable of detecting gradual long term processes, such as post disturbance recovery. Following the open Landsat data policy implemented in 2008, but also due to the improved level 1 processing standards, Landsat remote sensing experienced considerable innovation, with many novel algorithms for automated preprocessing and also for change detection. Among these, trajectory based change detection methods provide new means for assessing forest cover changes using Landsat data. For example, disturbances can be assessed on a yearly basis and residual noise in the time series is effectively reduced, enabling the previously impossible detection of gradual changes (e.g. recovery, degradation). We here demonstrate the analytic power of an annual time series approach (using the Landsat based detection of trends in disturbance and recovery (LandTrendr) algorithm) by assessing forest cover dynamics for an area in Romania, Eastern Europe. Our results illustrate that trajectory-based time series approaches can successfully be applied in relatively data scarce regions. Annual disturbance patterns allow for improved process understanding, and provide valuable inputs to a range of applications, including resource management, climate modelling or socio-ecological systems understanding, as in the case of Romania.

Radar time series are powerful means to improve retrieval algorithms about land surface characteristics in the following ways: (i) as information for identification of land surface conditions, (ii) as source of multivariate statistics for mapping methodologies, (iii) to select the right scene(s) for dedicated retrieval procedures, or (iv) to train model parameters in physical retrievals. Albeit radar data from air- and spaceborne platforms have been investigated since 40 years, operational applications are limited – partly due to the non-intuitive handling of complex microwave backscatter signals, and partly due to restricted geometric and temporal resolutions or frequency and polarization constraints. This chapter gives an overview of 20 years of pilot projects performed by the authors and their collaborators with the goal of large-area radar data exploration. All studies lead to innovative pre-operational applications, several with promising discoveries that can now be realized with a new and expanding fleet of radar satellites. Four case studies for land cover, forest mapping, forest cover change and savannah monitoring conclude this chapter.

Radar remote sensing is beneficial for retrieval of hydrological information such as soil moisture and flood extents due to the strong influence of water on the radar signal. The proper monitoring and analysis of such temporally dynamic phenomena requires dense time series data. Radar time series data is also useful for mitigating uncertainties in individual images, e.g. for the mapping of permanent water bodies. This chapter reviews capabilities, potentials and challenges of spaceborne radar time series data for the mapping of permanent water bodies, the monitoring of floods, and the retrieval of soil moisture content. The focus is put on the Lower Mekong Basin (LMB) in Southeast Asia. Two thirds of the LMB’s population of 60 million people live directly from agriculture and fisheries. The Mekong River's resources are under pressure among others from an increasing population, intensified agriculture, and the expansion of hydropower. A thorough understanding of water resources in the LMB is therefore crucial to the sustainable development in the region. The chapter provides an outline of radar remote sensing for retrieval of hydrological information as well as an overview of the relevant operational capabilities of radar missions. A map of permanent water bodies of the entire Lower Mekong Basin derived from a time series of ENVISAT Advanced Synthetic Aperture Radar (ASAR) data is presented. Potentials and challenges of flood monitoring with SAR are illustrated with ASAR imagery showing the evolution of the floods that occurred around Tonle Sap Lake in Cambodia in 2011. Finally, the spatial and temporal dynamics of soil moisture across the LMB are analysed with the use of 14 years of scatterometer time series data acquired by the ERS-1, ERS-2, Metop-A and Metop-B satellites. The average seasonal soil moisture cycle was computed at the sub-catchment level. An anomaly analysis of the temporal soil moisture dynamics revealed large inter-annual variability across the Lower Mekong Basin.

Vegetation phenology tracks plants’ lifecycle events and reveals the response of vegetation to global climate change. Microwave backscatter is insensitive to signal degradation from solar illumination and atmospheric effects and thus provides a useful tool for phenology monitoring. In this chapter, we analyzed a time series of Ku-band radar backscatter measurements from the SeaWinds scatterometer on board the Quick Scatterometer (QuickSCAT) to examine its effectiveness for land surface phenology monitoring across eastern Asia. The spatial pattern of annual mean backscatter follows regional vegetation type distributions. The Start Of Season (SOS) and End Of Season (EOS) were derived from the backscatter time series and compared with MODIS (Moderate Resolution Imaging Spectroradiometer) phenology products from 2003 to 2007. The failure of phenology metric detection for backscatter time series is caused by snow coverage and limited vegetation activity in arid areas. For tropical and semi-arid areas where optical observation is unavailable, backscatter data can provide valid phenological information. Due to their sensitivity to different factors, temporal discrepancies were observed between phenology products from backscatter and MODIS time series. Overall, the results indicate that SeaWinds backscatter provides an alternative view of vegetation phenology that is independent of optical sensors and can be applied to global phenology studies.

With worldwide economic development and population increases, urban areas create significant stresses on the local, regional and global environment. Information about the spatial and temporal dynamics of the characteristics of urban areas is therefore needed to support sustainable urban development. Time series earth observation data obtained using radar satellites have provided effective data sources for monitoring urban areas. This chapter first describes the development of synthetic aperture radar as well as its important role in the detection and monitoring of urban areas. Then, the fundamental principle of time series radar data in monitoring urban areas is introduced and discussed. Next, to demonstrate the capacity of time series SAR (Synthetic Aperture Radar) imagery for monitoring urban areas using ENVISAT/ASAR (Environmental Satellite /Advanced Synthetic Aperture Radar) time series radar data, Beijing city in China was selected as a test site. Beijing has all of the typical problems of a megacity such as resource, environment and population problems arising from rapid urban expansion during recent decades. A C5.0 rulesets classifier and the Multi Temporal Interferometric Synthetic Aperture Radar (MTInSAR) method were used to map the urban expansion and the millimeter level urban subsidence, respectively and the results were validated via high resolution WorldView optical datasets and leveling benchmark measurement, respectively. The results demonstrate the effectiveness and high accuracy of the time series radar data for monitoring urban areas. Furthermore, the spatial-temporal characteristic of urban expansion and urban subsidence of Beijing city were analyzed. Finally, the mechanisms or driving factors for urban expansion and subsidence are addressed based on economic development, population growth and the impacts of recent Beijing government policy.

Earth Observation using radar remote sensing is a valuable tool for the monitoring large scale inundation over time. This study performs a time series analysis using 18 ENVISAT/ASAR Wide Swath Mode data sets for the year 2008 and 13 TerraSAR-X Stripmap data sets for the year 2013/2014 to characterize inundation patterns in the Yellow River Delta, located in Shandong Province of China. Water surfaces are automatically derived using the software package WaMaPro, developed at the German Remote Sensing Data Center (DFD), of the German Aerospace Center (DLR), which allows an automatic classification using empirical thresholding. The temporal analysis allows the separation of different types of water bodies such as rivers, water storage basins, aquaculture, brine ponds, and agricultural fields based on inundation frequencies. This supports the understanding of the water dynamics in this highly variable study region. As ENVISAT data is not available anymore since April 2012, and as access to TerraSAR-X data is limited, Sentinel-1 data of the European Space Agency, ESA, are eagerly expected for the region. The good spatial resolution between 40 up to 5 m, as well as a dense temporal coverage, which allow to generate “true” SAR time series, and will help to lift annual analyses to the next level.