Climate change impact on recreation in heathland areas: the case of the National Park Hoge Kempen Anne Nobela, Sebastien Lizina, Nele Wittersa, Francois Rineaub and Robert Malinaa a Research Group Environmental Economics, Hasselt University, Belgium b Centre for Environmental Sciences, Hasselt University, Belgium Presentation prepared for the EAERE-FEEM-VIU Summer School Venice, 5 July 2018
Heathland Ecosystem with typical purple flower, on dry acidic soils. The EU expects a cost of 280mio annually to protect heathland against climate change, land use change. The question is: what are the benefits of doing so? I quantify the recreational value, which is one of the benefits, and how this changes under climate change 2/05/2019
Where can we find heathland? Dry heath generally occurs in regions with an Atlantic climate 24000 km² in total across Europe Decreasing in quantity (-20% in past 50 years) Decreasing in quality due to land use change, climate change and nitrogen input 2/05/2019
Last week Droughts are expected to be more frequent and more severe in the future across Europe. droughts are expected to increase the risk of uncontrolled, unwanted wildfires in heathlands 2/05/2019
Context – so why heathland? Dry heathland is one of Europe’s ecosystems that provides important ecosystem goods and services Of all ecosystems in Europe, heathlands are the most expensive to preserve, and also the most vulnerable to wildfires Heathland provides society with a variety of ecosystem goods and services, such as biodiversity conservation, carbon sequestration and recreation (Fagúndez, 2013; Webb, 1998). 2/05/2019
Context (2) Primary valuations of heathland ecosystem services are rare (n=2), and no estimates of potential welfare losses from climate change impact on heathland exist My objective is to value recreational opportunities provided by heathland, and how climate change affects this value My study provides the first economic estimates of climate change impact on heathland 2/05/2019
Overview Introduction Methods Expected results 2/05/2019
Methods For the estimation of recreational value we use the Travel Cost Method (TCM) For the estimation of the change of this value we use a Contingent Behavior (CB) model 2/05/2019
Methods – theoretical framework TCM is a method to infer the consumer surplus generated by visits to a recreational destination based on the cost incurred to visit that destination TCM is informed by data on past visitation rates ( 𝑦 𝑖 ): a revealed preference method The intuition is that if people incur high cost to pay a visit, it must be worth something 2/05/2019
Methods – theoretical framework The theoretical model for TCM is: 𝑦 𝑖 =𝑓 𝑇𝐶 𝑖 + 𝑋 𝑖 where 𝑦 𝑖 denotes the number of trips to the recreational destination by individual I, 𝑇𝐶 𝑖 the travel cost and 𝑋 𝑖 individual’s characteristics. If we want to incorporate site quality changes: 𝑦 𝑖𝑞 =𝑓 𝑇𝐶 𝑖 + 𝑿 𝒒 + 𝑋 𝑖 where 𝑋 𝑞 denotes site quality characteristics affected by climate change For the valuation of recreational opportunities we start from the proposition that individuals aim to maximize their utility within a budget constraint (Whitehead, Haab, & Huang, 2000). In addition, we assume that the qualities of a recreational site influence the utility that individuals expect to derive from their visits (J. Whitehead, Haab, & Huang, 2011). The expected utility, in turn, determines the number of trips to the site, and real or hypothetical changes of these qualities lead to changes to this number. In order to investigate the impact of wildfires on the recreational value of heathland we need to obtain visitation rates under current (no wildfire) conditions and hypothetical (after a wildfire) conditions. 2/05/2019
Methods – theoretical framework We will inform the theoretical model with current visitation rates (TCM), and contingent visitation rates under hypothetical conditions (Contingent Behavior) analyse the difference in visitation rates For the valuation of recreational opportunities we start from the proposition that individuals aim to maximize their utility within a budget constraint (Whitehead, Haab, & Huang, 2000). In addition, we assume that the qualities of a recreational site influence the utility that individuals expect to derive from their visits (J. Whitehead, Haab, & Huang, 2011). The expected utility, in turn, determines the number of trips to the site, and real or hypothetical changes of these qualities lead to changes to this number. In order to investigate the impact of wildfires on the recreational value of heathland we need to obtain visitation rates under current (no wildfire) conditions and hypothetical (after a wildfire) conditions. 2/05/2019
Methods - Study area: the High Campines National Park I will look into the case of NPHK - 6000 hectares 2/05/2019
The High Campines National Park is.. Methods – Study area The High Campines National Park is.. .. an urban protected area of 6,000 hectares that consists for 15% of heathland .. a recreational destination (> 1 mio visitors annually) .. the only National Park in Belgium 2/05/2019
Survey design - data collection To inform the theoretical model with data we hold a questionnaire among visitors of the High Campines National Park Visitors at the access point to the heathland-dominated part are (and will be) intercepted between May-September 2018 The questionnaire retrieves data about: Travel cost determinants Socio-economic characteristics Past visitation behavior and intended visitation behavior under contingent scenarios Scenarios are based on the premise that wildfires make heathland less attractive 2/05/2019
Survey design – Contingent scenarios Two scenarios, each with two variants: 1. A year with a wildfire: Variant 1: Intermediate wildfire (50% burned) Variant 2: Extreme wildfire (100% burned) 2. One year after a year in which wildfire occurred: Variant 1: Intermediate wildfire (50% grass cover) Variant 2: Extreme wildfire (100% grass cover) How many times would you visit in that year if.. First pair of scenarios have a wildfire in April 2/05/2019
Methods - Estimation of recreation demand model Specification of expected demand 𝑦 𝑖𝑡 : ln 𝑦 𝑖𝑡 = 𝛽 0 + 𝛽 𝑇𝐶 𝑇𝐶 𝑖 + 𝛽 𝑡 𝑞 𝑋 𝑡 𝑞 + 𝛽 𝑖 𝑋 𝑖 + 𝜀 How to estimate the coefficients? Travel cost % burned socio-economic % grass cover characteristics We take the log of y to equal .. The collection of RP and SP data from the same individual in the form of quasi-panel data enables researchers to include a correlation structure that models unobserved, individual heterogeneity, while this is not possible with independent RP observations alone (J. Whitehead et al., 2011). The models that are combined in this study will now be discussed. 2/05/2019
Estimation of recreation demand model – Data structure Individual T=1 current T=2 contingent T=3 T=4 T=5 1 2 3 .. i This means that our model should allow for correlation between the prob. distribution of the treatments Correlation between scenarios for each individual 2/05/2019
Estimation of recreation demand model – Data structure (2) Interrelated error terms 𝑦 =𝑓(𝛽 𝑋) + 𝜀 𝑦 =𝑓(𝛽 𝑋) + 𝜀 𝑦 =𝑓(𝛽 𝑋) + 𝜀 𝑦 =𝑓(𝛽 𝑋) + 𝜀 𝑦 =𝑓(𝛽 𝑋) + 𝜀 Individual T=1 T=2 T=3 T=4 T=5 1 2 3 .. i Solved via error terms of expected demands 2/05/2019
Estimation of recreation demand model (3) Discrete Factor Method-Generalized Negative Binomial relates error terms to each other by means of discrete factors Multivariate Poisson with Log-Normal error terms related to each other by means of a variance-covariance matrix I will estimate the 𝜷 for which these specifications return the highest likelihood value DFM-GNB is based on Negative Binomial MVPLN on Poisson Which are both count data models 2/05/2019
Expected results 𝑄= 𝑄 0 + exp(𝛽 𝑇𝐶 𝑃) 𝑄= 𝑄 1 + exp(𝛽 𝑇𝐶 𝑃) ∆𝐶𝑆 is a welfare measure of the impact of wildfires on heathland Key discussion point: time duration of scenarios and the probabilities remain unknown P added to travel cost ∆𝐶𝑆 Current contingent Q = visits per year of individual Welfare measure can be compared with current preservation cost and also with additional wildfire mitigation cost 2/05/2019
Thank you anne.nobel@uhasselt.be
DFM-GNB The conditional probability of the DFM-GNB specification is given as (Cameron & Trivedi, 2009): Pr 𝑦 𝑖𝑡 𝑥 𝑖𝑡 = Γ 𝑦 𝑖𝑡 + 𝑎 𝑡 −1 𝝁 𝑖𝑡𝑘 2−𝑝 𝑎 𝑡 𝑦 𝑖𝑡 𝝁 𝑖𝑡𝑘 𝑝 𝑦 𝑖𝑡 −2 𝑦 𝑖𝑡 1+ 𝑎 𝑡 𝝁 𝑖𝑡𝑘 𝑝−1 − (𝑦 𝑖𝑡 + 𝑎 𝑡 −1 𝜇 𝑖𝑡𝑘 2−𝑝 ) Γ 𝑦 𝑖𝑡 +1 Γ 𝑎 𝑡 −1 𝝁 𝑖𝑡𝑘 2−𝑝 𝝁 𝒊𝒕 = exp 𝑥 𝑖𝑡 ′ 𝛽+ 𝜀 𝑖𝑡 =exp 𝑥 𝑖𝑡 ′ 𝛽+ 𝜸 𝒕 𝝀 𝒌 discrete factors 2/05/2019
MVPLN 2/05/2019
Survey design – Scenarios Current situation cc cc cc cc The heathland is brownish in winter, greens up in spring, is purple in summer, and becomes brown in autumn 2/05/2019