Estimating patterns of vulnerability in a changing landscape: a case study of New Zealand's indigenous grasslands

Abstract

SUMMARYEffective conservation planning must anticipate the rates and patterns of dynamic threats to biodiversity, such as rapid changes in land use. Poor understanding and prediction of drivers and patterns of conversion of habitat can hinder assessments of the relative vulnerability of areas of remaining indigenous habitat to conversion, and identification of habitats in most immediate need of protection. Methods developed to model vulnerability to conversion vary in their complexity and applicability to conservation management. Generalized additive models provide a simple robust method to explore predictors and patterns of land-use conversion, and may be used to predict future patterns of conversion using recent land conversion data. This paper provides the first data-derived and statistically validated measurement of the vulnerability of New Zealand's indigenous grasslands to conversion. Higher altitude and more marginal (for agriculture and forestry) land showed greater conversion, and models based on earlier conversion patterns performed more poorly in predicting current patterns of conversion. Up-to-date land conversion data appear crucial for accurately predicting future vulnerability to habitat conversion.

Prioritising Land-Use Decisions for the Optimal Delivery of Ecosystem Services and Biodiversity Protection in Productive Landscapes

“…The predictive strength of empirical (observed) patterns of landuse change can be enhanced or diminished based on the combination of different. This also underlies the importance of validation in the modelling process [75].…”

“…Modelling patterns of land-use change is a fundamental component of conservation planning in productive landscapes [74][75][76][77][78]. Land-use change models can provide a tool for capturing the essence of where land-use change is likely to take place and what is driving these patterns [75,79] This information can then be used to assess the vulnerability of remaining indigenous habitats and help identify their relative urgency for protection.…”

“…Validation therefore requires testing the predictions of independent data (i.e. not those used in model parameterisation) to ensure the relationships inferred by a model are robust and the predictions are reliable [75].…”

“…Spatial statistical models provide a tool for predicting where land-use change is most likely to take place [75,79,91]. However, models based on patterns of past change will not necessarily provide reliable predictions of future change because over time exhaustion of formerly suitable areas and changes in global markets, technology, and crops can alter both the distribution and rate of habitat conversion [77,103].…”

“…However, models based on patterns of past change will not necessarily provide reliable predictions of future change because over time exhaustion of formerly suitable areas and changes in global markets, technology, and crops can alter both the distribution and rate of habitat conversion [77,103]. The reliability of land-use change predictions is therefore likely to decrease as they are projected further into the future, risking misallocation of scarce conservation resources [75]. However, though recent land-use change data will provide the best estimates of future land-use change, vulnerability estimates based on too narrow a time range may also provide less accurate forecasts, because they are based on a small sample of conversion events.…”