Detection and attribution of large spatiotemporal extreme events in Earth observation data

Highlights

• We present a method to detect spatiotemporal extremes in Earth observation data.

• We show how to find causes for extremes in variables monitoring the biosphere.

• Hotspots of fAPAR extremes are in NE Brazil, SE Australia, Kenya and Tanzania.

• Extremes in fAPAR are most often driven by water scarcity.

Abstract

Latest climate projections suggest that both frequency and intensity of climate extremes will be substantially modified over the course of the coming decades. As a consequence, we need to understand to what extent and via which pathways climate extremes affect the state and functionality of terrestrial ecosystems and the associated biogeochemical cycles on a global scale. So far the impacts of climate extremes on the terrestrial biosphere were mainly investigated on the basis of case studies, while global assessments are widely lacking. In order to facilitate global analysis of this kind, we present a methodological framework that firstly detects spatiotemporally contiguous extremes in Earth observations, and secondly infers the likely pathway of the preceding climate anomaly. The approach does not require long time series, is computationally fast, and easily applicable to a variety of data sets with different spatial and temporal resolutions. The key element of our analysis strategy is to directly search in the relevant observations for spatiotemporally connected components exceeding a certain percentile threshold. We also put an emphasis on characterization of extreme event distribution, and scrutinize the attribution issue. We exemplify the analysis strategy by exploring the fraction of absorbed photosynthetically active radiation (fAPAR) from 1982 to 2011. Our results suggest that the hot spots of extremes in fAPAR lie in Northeastern Brazil, Southeastern Australia, Kenya and Tanzania. Moreover, we demonstrate that the size distribution of extremes follow a distinct power law. The attribution framework reveals that extremes in fAPAR are primarily driven by phases of water scarcity.

Introduction

Understanding the role of climate extremes is of major interest for global change assessments, because their effects on the terrestrial biosphere have the potential to substantially modify regional carbon budgets (Ciais et al., 2005, Schwalm et al., 2012). However, it has yet to be investigated whether these effects can influence the global carbon cycle climate feedback system. Hence, a global analysis of extreme events with an emphasis on data streams describing the state of the vegetation and estimating the potential impact on land–atmosphere fluxes of CO₂ is of uttermost importance (IPCC, 2012).

Many recent studies on extreme events focus on either air temperature extremes (Anderson and Kostinski, 2011, Barriopedro et al., 2011, Rahmstorf and Coumou, 2011, amongst others) and/or water scarcity (Mueller and Seneviratne, 2012, Sheffield et al., 2012). Analyses linking the effects of climate extremes to impacts on the biosphere and the associated biogeochemical cycles, instead, have primarily revolved around effects of specific climate extremes of limited geographical extent (Ciais et al., 2005, Kurz et al., 2008, Page et al., 2002, Zeng et al., 2009). In the context of disturbance analysis the direct investigation of extreme events in the terrestrial biosphere based on remotely sensed data is well established (e.g. Forkel et al., 2012, Kennedy et al., 2007, Potter et al., 2003, Sun et al., 2012). However, most of these studies focus on individual disturbance types, for example storms (Sun et al., 2012) and fires (Forkel et al., 2012), or investigate specific regions and their ecosystems (e.g. Kennedy et al., 2007). To the best of our knowledge, the question how to attribute these disturbances to either synchronous or antecedent climate extremes has been only treated marginally in the literature (Potter et al. (2003) provides some ideas in that direction).

As a consequence, it is difficult to draw a generalized and global picture on the impact of climate extremes on terrestrial ecosystems and the corresponding land–atmosphere fluxes (but see Zhao and Running (2010)). Especially the question how to detect and quantify extreme events in the terrestrial biosphere spanning over the three dimensions latitude, longitude and time on a global scale has still not been addressed systematically. Here we can learn from methodological advances in hydrology where a series of ideas have been proposed that consider both the spatial and temporal extent of large-scale droughts. For instance, Andreadis et al. (2005) use a true space–time representation of droughts, but eventually mainly focus on their areal extent. Corzo Perez et al. (2011) perform a “non-contiguous and contiguous drought area analysis”, where they specifically aim at capturing both spatial and temporal information on droughts, but without investigating the corresponding climate trajectories corresponding to the detected events. Recently, Lloyd-Hughes (2012) presented the first analysis strategy for droughts that is truly giving full consideration to both space and time. In fact, in his approach space and time are exchangeable and a drought event is characterized by its three-dimensional extent. Thus, a further analysis of the drought's size, shape, temporal evolution and other interesting quantities can be performed.

The common notion of all the aforementioned studies is that extreme events have to be understood as spatiotemporal phenomena and detection algorithms have to be tailored to this peculiarity. This property gains additional relevance when we consider that Earth observations monitoring the terrestrial biosphere are often relatively short compared to the periods of relevance. For instance, satellite remote sensing observations are currently spanning up to maximal three decades, which prevents the application of classical extreme value theory (e.g. Coles, 2001).

In this paper we present an approach, firstly, to detect large spatiotemporal contiguous extreme events in Earth observation data, and secondly, to attribute these events to ancillary spatiotemporal climate observations. While the core ingredient is similar to the recent large-scale drought detection independently developed by Lloyd-Hughes (2012) we complement this work with a comprehensible preprocessing of the data and a new algorithm for attribution and by exploring a range of percentiles in data anomalies. Our goal is to obtain a generic method that is widely usable for a large class of variables, in particular gridded Earth observations. We illustrate the effects of the various steps of data preprocessing based on a synthetic data set. Then, we use the methodology for exploring extreme events in a composite of the fraction of absorbed photosynthetically active radiation (fAPAR, Jung et al., 2011), as an example for a remotely sensed data set. Because fAPAR can be related to the primary productivity of photosynthesis (Gobron et al., 2010), it is often used in diagnostic models that estimate the assimilation of carbon dioxide by the terrestrial vegetation. Finally, we attribute the detected extreme events in fAPAR to extreme anomalies in likely driver variables, i.e. air temperature, water availability, and fire.