While future changes are presented as extrapolations of the patterns quantified in the past, policy changes might cause deviation from the projections presented here.
A major challenge is to produce socio-economic and policy scenarios to inform projections that will differ from current landscape management. Given that urban sprawl is affecting many land surfaces globally, the approach used here could be generalized to other countries in similar situations. Global landscapes have been changing at an accelerating pace over the last decades, potentially threatening both the natural environment and human well-being Vitousek et al.
Such land use changes have reshaped—and will continue to reshape—semi-natural and artificial environments occupied by the human population Vitousek et al. Changes in the use of landscapes are generally driven by the growing demands for natural resources in a developing human society to foster an increase in standard of living for growing populations Foley et al.
Quantifying the drivers underlying observed landscape transformations may provide a better understanding of the consequences of human activities for future landscapes Rutherford et al. Furthermore, knowledge of the drivers of past land use change makes it possible to project potential future land use through land change models Verburg, ; Verburg and Overmars, Models quantifying statistical relationships between landscape variables and land use change make it possible to project expected shifts in the provision of services within future landscapes Pellissier et al.
Model projections can be used to inform management, which can then potentially buffer adverse effects of change through appropriate mitigating policies Lawler et al.
Abundant and detailed information on past land use exists for many countries and documents general past trends. For instance, recent land use change in the European Union has been dominated by urban growth and by agricultural land abandonment followed by spontaneous reforestation, to the detriment of cropland and grasslands, thus translating into the loss of agricultural production Falcucci et al. Statistical analyses enable investigation of the underlying drivers of past land change trends Verburg, ; Brown et al.
Rutherford et al. Furthermore, land use change models make it possible to project future landscape arrangements under specified scenarios of changes Sterk et al. Hence, with a single analysis, one can couple the two purposes of land use change models, explanation and projection Brown et al.
Accepted Open Panels - 4S NEW ORLEANS
For example, Lawler et al. The spatial nature of land use change models facilitates mapping of where future changes are expected to happen, which can guide regional decision making Lawler et al. The concept of ecosystem services ESS arose to assign values in terms of societal gains to natural systems that are not directly captured by market prices Kareiva, , while enabling the quantification of past and future landscape changes consequences on human well-being, such as material needs, social relations and security MEA, ; Kienast et al.
Nevertheless, the concept of ESS has not fully supported decision making and management as expected Wallace, ; de Groot et al. Landscape services LSS has been suggested as an alternative term to quantify services from both semi-natural and artificial environments within landscapes and express the idea that many ESS cannot be enjoyed at the plot level but rather in the context of neighboring plots, i.
Furthermore, Cumming et al. Our aim was to broaden the understanding of perspective on the provisioning of services by analyzing the two types of services simultaneously. Thus, we use the term LSS instead of ESS to emphasis the considered combination of services from natural, semi-natural, and artificial landscapes. Combining land use data with context information to quantify services improves the quality of service estimates and enables better assessments of trade-offs between the classical services sensu CICES 5. Simple approaches use look-up tables to link habitat or land use with a specific service.
Such look-up tables can either be binary Kienast et al. An approach that includes additional indicators e. For instance, Chan et al. The appropriate approach for assessing ESS depends on the time and data available, as well as the scale and goal of a given study Kareiva, ESS assessments are usually trade-offs between accuracy and time invested. Quick links between land use and provided services based on expert knowledge e. Complementing links based on land use with more static environmental parameters can improve the accuracy of the links Chan et al. However, the data necessary for that step are often not available at an appropriate spatial resolution Burkhard et al.
- To Catch a Thief.
- The Social Sciences of Quantification | Archambault;
- Disinformation by Quantification | Centre for Free Expression!
- Meeting Difficulties in Literacy: Research, Policy and Practice!
- World Crisis and British Decline, 1929–56;
- The Perils of Quantification | Boston Review.
In this study, we use statistical models to quantify the main trends of land use change in Switzerland and forecast the consequences of future changes on LSS. Over the last decades, settlement areas have encroached on agricultural land in the Swiss lowlands, while forests have expanded in the Swiss alpine regions because of agricultural abandonment Price et al. Nevertheless, how land use change will affect future LSS provisioning in the different regions of Switzerland is still unknown, in particular because the classic quantification of services does not consider built-up areas.
To provide a more comprehensive picture of how future land use change will reshape LSS provisioning, we consider a more integrative quantification of services. We combine new formulations for the quantification of LSS with land use change models and their drivers. This approach allows us to explore future consequences of land use change in Switzerland with the following expectations:.
Land use transitions should be largely driven by the landscape state at the starting time point. Therefore, we expect that predictors describing this state, such as the distance to established urban centers or to similar land use types, primarily determine the transition locations for land development or abandonment. Given the considerable heterogeneity of the Swiss landscape, the future dynamics of land use change are likely to be region-specific, with an increase in urban areas in the Swiss lowlands and a continuation of land abandonment in alpine regions.
Trade-offs between the LSS provisioning are likely to increase in the future, especially between those LSS based primarily on built-up areas and agricultural services. Our analysis provides a clearer spatial picture of potential future land use changes across the Swiss landscape and the associated LSS. Linking land use change with its potential consequences provides an information basis for policy interventions Nelson et al. The complete Swiss Land Use Statistics data sets are available for three flying periods with a periodicity of 12 years: —, —, and — We aggregated the 72 categories as presented in Table 1.
The biogeographic regions differ in topography, climate and historical context. We opted for a hard regionalization i. We considered a set of 22 predictors for the land use change models based on previous studies Rutherford et al. This set includes the most relevant but by no means all potential predictors of land use change in Switzerland. The predictors can be grouped into five categories: twelve predictors are computed directly from the Swiss Land Use Statistics land use types see Electronic Supplementary Material 1.
Three predictors are assigned to each of the three categories topography elevation, slope, solar radiation , socio-economics employee density 1st economic sector, employee density 2nd and 3rd economic sectors, population change , and accessibility public transportation accessibility, distance to major roads, distance to economic centers. We also considered one climatic predictor mean annual temperature.
For more details on the predictors, their sources, and their temporal development see ESM 1. Predictors were selected for each transition individually based on a literature review and logical reasoning.
We considered different land use change transition models see Table S1 for details. We used generalized linear models to model the different land use transitions separately, ignoring transitions that happened on fewer than 50 pixels across the whole of Switzerland. We validated the models using an external validation: we calibrated generalized linear models with data from the first transition period of the Swiss Land Use Statistics — and validated them with data from the second transition period — We quantified the model performance using True-Skill-Statistics Allouche et al.
In addition, True-Skill-Statistics were calculated with the split-sample approach ratio , an internal evaluation method. To calibrate the final projection models, the separate data sets from both past transitions were combined into one transition data set and considered equally, ignoring their starting time point.
For rare transitions, this procedure increased the number of data points, while for frequent transitions samples of equal size were randomly selected from both past transition data sets. All calculations were conducted in R Statistical Software version 3. We projected the land use changes over Switzerland for the next six decades starting from a base map of the period —, followed by six future year time steps.
For each cell, we forecasted the probability that each land use category would transition into another category using the calibrated generalized linear models. Each cell was thus attributed a set of probabilities of change into other land use types. In addition, we defined the probability that a cell would retain its land use as one minus the maximum transition probability of the cell.
For each cell we sampled a single final projected transition 1 or stability 0 based on the vector of probabilities. We ran land use change projections separately for each biogeographic region using the regional generalized linear models. At each time step, the projected land use categories from the previous step were used to update the landscape predictors used for the next time step forecast. In order to quantify uncertainties, we ran an ensemble of independent projections.
If the required data were not available, we developed formulas adapted to our study. In addition, we considered two non-traditional LSS to expand the research field to services provided by built-up areas, i. The formulas for LSS quantification were developed based on a combination of methods from the literature and the available data Table 2.
A detailed description of the calculation of each service, including how the coefficients for the look-up tables were defined, can be found in ESM 1. When possible, we validated the quantification of those services using independent data sets. The spatial quantification of jobs was validated by comparing the Swiss Land Use Statistics with the statistics of the number of people employed in the secondary and tertiary economic sectors SFSO, c.
We validated the quantification of crop production with the cantonal numbers describing total agricultural production SFSO, No comparison data were available for the other five services, and thus no validation was conducted. To analyze the development of services over time, we looked at the total value of one service across Switzerland and in each biogeographical region individually.
Brain-Computer Technology Is Accelerating. Will We Soon Be Typing With Our Minds?
For that, we calculated the service provisions for each cell as described in Table 2 and summed them for each region. As the regions differ significantly in total area, we divided each sum by the area of the respective region to provide an average service provisioning per hectare, which enabled comparison of future trends between the different regions. We found no spatial auto-correlation in the residuals of the models. The explained deviance of the predictors D 2 was in the highest quartile of the deviance distribution 5. Details on the explained deviance of the predictors for individual transitions are provided in Table S1.
Transitions between forests and lower elevation grasslands and meadows also occurred frequently times for forests to grasslands and 12, times for the reverse. As expected, with a few exceptions, the trends observed in the past were projected to continue in the future. However, acceleration and deceleration of specific land use type transitions were observed.
In contrast, changes from grasslands and meadows to few-family houses and from arable land to manufacturing and service infrastructure were projected to continue increasing. In general, built-up areas were projected to continue to increase Fig. While grasslands and meadows showed an accelerating decrease, the decrease in arable land, as well as orchards, vineyards and horticulture, decelerating considerably compared with in the past Fig.
Spatial arrangement of aggregated land use categories in time steps of 24 years. Note the increase in urban area around existing city centers and in the valley bottoms of the Western Alps. Distinct differences in the changes were projected for the individual biogeographic regions see standard deviations in Fig.
The most prominent projected changes happened on the Plateau, where the built-up area was projected to encroach further into the lowland agricultural areas, and in the Western Central Alps, where an immense increase in urban area, largely containing few-family houses and manufacturing and service infrastructure, was projected. Orchards, vineyards and horticulture were projected to disappear in the Western Central Alps with accelerating speed, while forests were projected to shift upward in elevation Fig.
Average change in relative covered area for each land use category. The changes are for across all six biogeographic regions and all four year time periods. Error bars indicate the standard deviation across the regions. Clear distinctions between urban and non-urban areas are recognizable in the current distribution of the provisioning of housing, jobs and recreation. Differences are more prominent between the lowland and the alpine regions of Switzerland for biodiversity, reared animals and hazard protection.
Crop production occurs mostly in the lowlands. According to our formula, forest patches in the lowlands provide more forest products than those in the alpine regions.