Biocomplexity Project Summary

Project Schematic

The goal of this research is to develop and test the analytical tools needed to understand and predict the interactions and feedbacks among humans and aquatic species across complex landscapes. Our central organizing principle is that landscape patterns and changes in network structure and function are explained by energy and time optimizations of water flows, biota and humans. Although we understand many ways that each subsystem individually optimizes time and/or energy, each optimization imposes constraints on related subsystems, which can then change the rules by which each operates. Through these multiple feedbacks and interactions, even individually well-understood subsystems can produce unpredictable dynamics. Thus, we propose an integrated modeling framework to incorporate feedback and interactions among biotic and abiotic systems so as to accurately predict the effects of interactions of human and aquatic populations. These multiple feedbacks and resulting emergent properties are hallmarks of a complex adaptive system that can be simulated using an individual- (agent) based simulation model.

Typically, adding more complexity increases the variance and uncertainty of modeling predictions. Our main overarching hypothesis is that an integrated individual-based model will more accurately predict environmental effects than any single physical, biotic or social model by reducing unexplained variation. The mechanism generating this reduction in variance results from including cross-disciplinary connections and corresponding agent feedbacks to what otherwise would be missing in the individual disciplinary models ( e.g. , human use as an explanatory variable in biotic systems). In order to test this hypothesis we will develop an integrated individual-based model of physical, biotic and social networks using energy and time optimization as a unifying principle with multiple feedbacks between networks as a framework to capture different hierarchical levels of complexity. In our case study, river and road intersections bring about interaction of natural and social agents, with mutual feedbacks (Figure 1). Dynamic interactions are captured by interdependent functions of each individual agent and each subsystem ( e.g. , nearest neighbor dynamic). Thus, our approach integrates multidisciplinary field studies into a unified individual-based model that addresses fundamental problems of biocomplexity to quantify and evaluate different dimensions of interacting human and natural systems. The model will be developed and applied to test our main hypothesis along a land use continuum from coastal commercial, residential, agricultural and forested reserves. The experimental design relies upon an existing natural continuum in the level of road densities and road types within a network structure that reflects a hierarchy of land uses, geomorphic and riverine features. As detailed below, we will test the accuracy of predictions using all categories of interacting agents versus predictions based on sequential removal of one class of agents.

We hypothesize that there is an inverse relationship between the influences of physical and social factors on food-web organization and function at river-road junctions (common nodes) within entire river networks. In high-elevation, high-energy (erosive) streams, we expect that social effects ( e.g. , fish and shell-fish harvesting, introductions of non-native species) dominate changes in the species distributions because of easy access for people to intensively harvest river-based food and for frequent recreational uses near numerous small roads. In low-elevation, low-energy (depositional) streams, physical effects ( e.g. , channel and hydrologic modification) dominate changes in the species distributions although land uses and other types of human impacts may become important in watersheds with high densities of roads. We will test a series of explicit hypotheses to compare and contrast the hydrology, geomorphology, visitor behavior, and food web structure at river segments and river-road junctions within these overlapping networks. These hypotheses have policy relevance in terms of the level of development that is allowed in currently unroaded watersheds.

The tropics provides a natural laboratory to study biocomplexity due to high biodiversity and rapidly increasing human density. Our case study site offers both of these features in a spatially compact set of adjacent, well-studied watersheds. We will conduct our field studies and calibrate our sub-models in three adjacent watersheds in northeastern Puerto Rico that have land uses that range from high-density urban to pristine tropical rain forests. The integrated model will then be tested in a fourth watershed that contains the same mix of land uses found in the adjacent watersheds. Northeastern Puerto Rico is a good natural laboratory due to rapid spreading of urbanization and sub-urbanization with an associated hierarchy of road networks that modifies the natural landscape and places people closer to natural resources. Being in a coastal environment in proximity to rain forests, there is a steep land rent gradient that compresses the transitions from one land use to another. Similarly, the topographic steepness results in numerous, distinct transitions from one biological community to another in a short distance.

Conceptulization and Theory

We will integrate physical, biological and social systems by incorporating the results of field studies into a unified individual-based model. In this framework agents interact by a set of rules or functions. These rules describe how each type of agent (physical, biotic and human) operates. These rules include not only the direct actions of the agents, but also interactions of the agents with each other. In the past, these individual-based models were applied separately to biotic and human systems [ 28 ]. Using an established individual-based modeling protocol called Swarm , [ 28 , 123 ] we will integrate the physical , biotic and human agents as interacting components of a single ecosystem. We will model operation of physical processes of river networks through energy minimization [ 177 , 178 , 179 ]. We will also model operation of biotic organisms within these rivers using time and energy as our common currency. Furthermore, we will model human agents' behavior through time and energy trade-offs in attainment of human objectives. A common linkage between biota and humans is the road network that provides humans access to river networks and influences land-use development.

1. Research Integration and Project Management Plan. Figure 1 below illustrates the interconnections between models and field studies by each PI, who will coordinate data collection and development of particular submodels. Ramirez, Wohl and a graduate student, will lead analysis of stream channel networks using Optimal Channel Network models (OCN) that combine climatic inputs ( e.g. , rainfall), topography (Digital Elevation Models or DEMs) and concepts of optimal energy expenditure to derive channel and network geometry [ 177 , 178 , 179 ]. Tomlin and his graduate students will lead the development of road networks over time using raster based techniques and inputs of topography, land-use, and population centers developed from Xplorah by Gutierrez and Santiago at the Graduate School of Planning, University of Puerto Rico. Outputs from both the OCN models ( e.g. , discharge, stream power, slope) and road densities from the Optimal Road Network (ORN) model will be combined by Scatena and Wohl using empirical field data and physical habitat models (PHABSIM) developed by the U.S. Geological Survey and recently modified by Scatena and Johnson [233], to model the quality of stream habitat for aquatic organisms ( e.g. , shrimp) and recreation ( e.g. , swimming). In this Stream Habitat model, habitat quality will be evaluated on the basis of physical features ( e.g. , depth, volume, velocity, riparian conditions etc.) conducive to both recreation and aquatic organisms. Covich and Crowl and their graduate students will model use of stream habitats by aquatic species such as fishes and shrimps that are harvested by local fishermen. Recreational visitors to rivers who fish and swim in the rivers will be interviewed by Loomis and Gonzalez-Caban along with graduate student at UPR. They will develop and implement visitor surveys to estimate a Travel Cost Method recreation demand model utilizing the output from the Stream Habitat model ( e.g. , stream depth, velocity), diversity and abundance of aquatic organisms, and Optimal Road Network ( e.g. , travel times and costs) to parameterize the attraction of various points in the stream network for various types of recreational use ( e.g. , swimming, fishing, etc.). The spatial and temporal distributions of the aquatic organisms will be derived from field data by Covich and Crowl, as well as data on road infrastructure from Laituri's GIS analysis. Likewise, output from Stream Habitat model combined with information on visitor use and the location of the pools within the Optimal Channel Network models will be used to compare the predicted and measured population structure of aquatic organisms at various locations within the landscape. Land-use changes at the parcel level will be modeled using the Von Thunen land-use transitions from population centers, as determined by road access and associated economic returns per acre [ 78 , 227 ]. CO-PI's at UPR will use the Xplorah spatial decision support system that has been calibrated for Puerto Rico to link changes in social and economic processes associated with changes in the road network to land use changes. The Xplorah model uses economic forces such as dollar returns per acre, spatial interaction models, and cellular automata to model land use and their transitions.

All modeling efforts will initially be developed from field studies in four adjacent watersheds that range in land use from urban (high road density) to moderately developed (medium road density), to relatively pristine tropical rain forest (low road density). The aquatic and human populations will be examined using individual-based models that work within the physical framework of pool habitats and are dependent on the configuration of the stream and road network. Rules for all sub-models will be implemented as integrated components of an individual-based model, where dynamic populations and dynamic landscape processes will run concurrently, using Swarm [ 176 ] and the Kenge libraries [ 27 ] to integrate agent-based models, cellular automata, and GIS in a common framework. After individual models are developed and integrated they will be tested by predicting outcomes in a fourth, relatively unroaded watershed, the Rio Fajardo. We will then use our integrated models to simulate future landscape development with various roading options.