This research reveals new insights into the weather drivers of interannual variation in land surface phenology (LSP) across the entire European forest, while at the same time establishes a new conceptual framework for predictive modelling of LSP. Specifically, the random-forest (RF) method, a multivariate, spatially non-stationary and non-linear machine learning approach, was introduced for phenological modelling across very large areas and across multiple years simultaneously: the typical case for satellite-observed LSP. The RF model was fitted to the relation between LSP interannual variation and numerous climate predictor variables computed at biologically relevant rather than human-imposed temporal scales. In addition, the legacy effect of an advanced or delayed spring on autumn phenology was explored. The RF models explained 81 and 62 % of the variance in the spring and autumn LSP interannual variation, with relative errors of 10 and 20 %, respectively: a level of precision that has until now been unobtainable at the continental scale. Multivariate linear regression models explained only 36 and 25 %, respectively. It also allowed identification of the main drivers of the interannual variation in LSP through its estimation of variable importance. This research, thus, shows an alternative to the hitherto applied linear regression approaches for modelling LSP and paves the way for further scientific investigation based on machine learning methods.
8 February 2017 in Journal of Coastal Conservation
Accurate digital elevation models of saltmarshes are crucial for both conservation and management goals. Light detection and ranging (LiDAR) is increasingly used for topographic surveys due to the ability to acquire high resolution data over spatially-extensive areas. This capability is ideally suited to saltmarsh environments, which are often vast, inaccessible systems where topographic variations can be very subtle. Derivation of surface (DSMs) (ground elevation plus vegetation) versus terrain (bare ground elevation) models (DTMs) relies on the ability of the LiDAR sensor to accurately record multiple returns. In saltmarshes however, the dense stands of low (< 1 m) vegetation commonly found precludes the acquisition of more than one return, and the resulting DTM is not different to the DSM. Establishing the offset between ground and vegetation surface in order to correct the LiDAR-derived DTM can be challenging due to the spatial variability in saltmarsh habitats. Here we show the development and application of a habitat-specific correction factor (HSCF) for the Odiel Saltmarshes using a combination of habitat object-based classification (82% overall accuracy) and ground control surveys that reduces the DTM error to within that associated with the LiDAR sensor (average error 0.1 m). We also show that the true accuracy of supplied (unmodified) DTMs can be >0.5 m in saltmarshes dominated by dense vegetation such as Spartina densiflora. In particular, global projections of sea-level rise across the next 80 years (0.18–0.59 m) significantly overlaps this accuracy margin, implying that assessments and modelling of sea-level impacts in saltmarsh systems will likely be erroneous if based on Lidar-derived DTMs. Erroneous assumptions and conclusions can result if the real accuracy of DTMs (bare ground) on vegetated saltmarshes is not considered, and the consequences of the propagation of this misinformation through to management decisions should not be over-looked.
2 January 2014 in International Journal of Remote Sensing
Saltmarshes provide high-value ecological services and play an important role in coastal ecosystems and populations. As the rate of sea level rise accelerates in response to climate change, saltmarshes and tidal environments and the ecosystem services that they provide could be lost in those areas that lack sediment supply for vertical accretion or space for landward migration. Predictive models could play an important role in foreseeing those impacts, and to guide the implementation of suitable management plans that increase the adaptive capacity of these valuable ecosystems. The SLAMM (sea-level affecting marshes model) has been extensively used to evaluate coastal wetland habitat response to sea-level rise. However, uncertainties in predicted response will also reflect the accuracy and quality of primary inputs such as elevation and habitat coverage. Here, we assessed the potential of SLAMM for investigating the response of Atlantic-Mediterranean saltmarshes to future sea-level rise and its application in managerial schemes. Our findings show that SLAMM is sensitive to elevation and habitat maps resolution and that historical sea-level trend and saltmarsh accretion rates are the predominant input parameters that influence uncertainty in predictions of change in saltmarsh habitats. The understanding of the past evolution of the system, as well as the contemporary situation, is crucial to providing accurate uncertainty distributions and thus to set a robust baseline for future predictions.