Quantification of vegetation response to climate anomalies through remote sensing: methodological aspects and applications

Ecosystems provide many crucial services, among which climate, water and erosion regulation, pollination, genetic resource conservation, wildlife habitat and products such as timber and fresh water. These services depend on ecosystem processes, such as biomass production and organic matter decomposition, which are on their turn dependent on biodiversity. A continuous and stable delivery of these ecosystem services is, however, threatened by the expected increase in average temperatures, the increased frequency and magnitude of climate extremes, and also by ecosystem degradation, land use conversion or change and biodiversity decline in many parts of the world. Within this context, it is of utmost importance to assess and monitor the stability of ecosystems and to understand the factors that may mediate ecosystem stability at large spatial scales. Ecosystem stability is often quantified through metrics defined on the ecosystem state variable, i.e. ecosystem resistance, resilience or variance. Resistance denotes the ability of the ecosystem to withstand a disturbance and resilience is related to the ability to recover after the disturbance has occurred. Resilience is often subdivided into two stability metrics: engineering resilience is defined as the speed at which the ecosystem returns to its original state, whereas ecological resilience denotes the amount of disturbance that is needed for the system to switch state. Finally, the variance is defined as the total variability of the ecosystem variable in face of environmental anomalies. The frequent and large scale assessment of satellite imagery provides an interesting asset to obtain and monitor ecosystem stability at large spatial scales. Time series of vegetation indices, such as the Normalised Difference Vegetation Index (NDVI) or Normalised Difference Water Index (NDWI), can be derived from these images and provide indicators of the vegetation condition, e.g. the vegetation greenness or water content. Yet, several problems, e.g. noise or the spatial heterogeneity of the disturbances, still hamper the reliable quantification of ecosystem stability. Therefore, the first part of this PhD provides methodologies to enhance the reliable assessment of ecosystem stability. Several types of noise and data characteristics, e.g. temporal resolution, time series length, and noise introduced by variable atmospheric conditions, may affect the stability metrics obtained from remote sensing time series. Therefore, a methodology has been described to quantify the impact of these noise and data characteristics on stability metrics, and assess the reliability of the time series. Next to noise, the spatial variability of disturbances may hamper the spatial comparison of stability metrics. A solution to this problem has been proposed by explicitly modelling the vegetation state indicator anomaly (e.g. NDVI anomaly) in function of past anomalies and the disturbance. Finally, the response of vegetation on environmental anomalies may differ over time. Vegetation may, for example, become more or less resistant or resilient to climate anomalies. This problem has first been theoretically described and subsequently been quantified using the magnitude and direction of stability change for several stability metrics in the case of Australia. Consequently, when comparing the stability of ecosystems, the proposed methodology allows to take into account not only data quality, and heterogeneity of the disturbances, but equally so their non-stationarity. The second part of the PhD provides insight in the factors that govern ecosystem stability. More specifically, it quantifies the effect of plant diversity and management on grassland stability. First, differences in stability between species-rich semi-natural and species-poor intensively managed agricultural grasslands in the Netherlands were assessed, and subsequently the relationship between plant diversity and stability of dune grasslands along the Dutch coast was studied. Both studies demonstrate that highly plant diverse grasslands tend to be less sensitive to droughts compared to lowly plant diverse grasslands. Yet, the semi-natural grasslands show to be less resilient than the intensively managed agricultural grasslands. Both studies highlight the importance of grassland plant diversity in the Netherlands, where it concerns maintaining a high resistance to droughts. Overall, this dissertation proposes three approaches or methodologies to enhance the reliability of large scale ecosystem stability assessment using remote sensing. In addition, these methodologies are used to study grassland stability in the Netherlands, and to upscale the relationship between diversity and grassland stability to droughts. This work therefor contributes to the reliable quantification, monitoring and understanding of ecosystem stability.nrpages: 161status: publishe