Impact network analysis and the <scp>ina r</scp> package: Decision support for regional management interventions

Success in interventions ecological systems often depends on understanding which system components are limiting factors. Successful regional management of species how effective methodologies are, whether a critical mass decision-makers adopts good management, and the resulting efficacy geographical landscape choices. Achieving this type ‘management landscape’ is common challenge for intervention projects across applied ecology—including invasive endangered restoration, agricultural development public health programmes, illustrated prominently by COVID-19 pandemic—with opportunities synergies developing concepts subdisciplines (Carvajal-Yepes et al., 2019; Chadès 2011; Hobbs Hulme 2020; Lenzner Ostrom, 2009). Invasive dispersal key threat to while connectivity reserves conservation (Hilty 2012). Sustainable technologies managing spread pathogens, arthropod pests weeds, supporting improved crop genotypes (Henry & Vollan, 2014; McEwan 2021). Public supported communicating about disease using methods such as vaccination slow its (Manfredi d'Onofrio, 2013). Developing strategies these related requires integration three components: (a) quality research underlying them, (b) socioeconomic networks technology use, land managers or farmers (c) biophysical dispersal, pathogen invasion where decisions use influence establishment probabilities. Here, ‘impact networks’ defined multilayer networks, composed linked through may have effect. This paper introduces framework scenario analysis (Garrett 2018), network analysis’ (INA; Figure 1) an r package that implements scenarios ecology impact can provide decision support formulating project (Garrett, 2021a, 2021b). The first component technology, biocontrol agents, biocides, burning regimens, models indicating best timing activities some combination technologies. These all products scientific be thought terms ‘information’ (in broad sense) from experiments, with associated uncertainty their effect (Klerkx 2010). ‘value information’—the improvement outcomes when take into account information, versus not having information—is useful concept strategies. Analyses information’ been incorporated in, example, medical decision-making (Bartell 2000; Claxton Sculpher, 2006; Tappenden 2004), (Canessa 2015; Tallis Polasky, Wiles, 2004) adaptive resource (Williams 2011). reproducibility science being critically evaluated multiple disciplines, questioning information generated experiments (Ioannidis, 2005; Kenett Shmueli, Leek Peng, 2015). And even if very high quality, system-level will minimal unaware them. Impact (INA) evaluation realized value landscapes. second network, nodes farmers, other water managers, individuals families’ (Burgess Garcia-Figuera 2021; Rebaudo Dangles, 2011, 2013)—and potentially also include agents scientists (Ekboir, 2003), extension policymakers, consumers institutions. Links between indicate ideas, and/or money. Individual adopt new plays out context available individuals’ 2012; Rogers, 2003). Agricultural limited lack (Parsa 2014), general heuristics well developed (Ascough 2008; Gigerenzer Gaissmaier, effects without options, create influences success failure network. In third component, locations or, more generally, ‘bioentities’ (including functional groups, biotypes, genotypes, varieties genes) become established (Calabrese Fagan, 2004; Galpern Nodes might people (as hosts human pathogens), farms, habitat patches units. potential undesirable bioentities, antibiotic-resistant pathogens (Epanchin-Niell 2010; Margosian 2009; Sutrave Xing 2020), desirable varieties, example orange-fleshed sweetpotatoes vitamin A consumption (evaluated Andersen al. (2019)). (In cases, same model usefully abiotic components, pollutants, soil erosion provisioning fresh water; Baron 2002). corresponding layer probability bioentity at node modified (Figure 1). Combining provides perspective based analyses, evaluate investments. INA used likely degree adaptation pulse (intermittent) press (continual) stressors, introduction climate change (Harris operationalize sustainability, resilience economic viability. Some considered together more-or-less explicitly (Funk 2009, Garrett, Harwood Manfredi 2013; Sahneh 2012) natural (Bodin Prell, Burgess Conley Udry, Epanchin-Niell Hastings, Hernandez Nopsa Magnan Mills INA, agent-based model, help bridge interactions. policy (Fealing 2011) evaluating interactions among engaged results implementing results. overall goal integrates ecology, enhance lessons learned community practice, platform viability management. Network compared aggregated models, allow consideration role social structures likelihood technological innovations. designed implementers, funders policymakers prioritizations they must consider, complement traditional approaches monitoring evaluation. specific objectives introduce ina (github.com/GarrettLab/INA, 2021b), illustrate identification smart surveillance strategies, application before, during after (d) under global introducing ‘adaptation functions’ response required sustainability resilience. hypothetical systems, observed blend observed. Data limitations always but quantification methods, here, inform investments interventions. Because importance both most platforms like valuable current Many applications would data along simulated represent part understand varying parameters difficult estimate. Three simulation presented here purposes, experiment smartsurv function simpler, structure describing bioentity. experiment, implicit: location weighted enter two INAscene analyses bioentity, cultivar, species. Technical details functions user guide 2021b) several vignettes github.com/GarrettLab/INA 2021a). Surveillance informed knowledge spread. hubs (high degree) bridges betweenness) tend important sampling (e.g. 2019), complex traits (Holme, 2017, 2018). relative risk point decision-maker node's communication (Buddenhagen 2017). R each detect (Table evaluates find, turn, many remain free time it detected node. uninvaded detection, identifying there still manage invasion. Sampling considering smartsurv.weight uses output introduced vary one another. illustration, differences commonly studied types evaluated. users want own estimate system. identified set types, shown vignette (V1 github.com/GarrettLab/INA). 1 nine simple scenarios, representing weighting entry points invasives. random (Erdős Rényi, 1960), small world (Watts Strogatz, 1998) scale-free (Barabasi Albert, 1999). unweighted, weights proportional inversely degree. 2 3, was perform simulations programming environment, igraph (Csárdi Nepusz, 2006) generating figures. Details Supplemental Information shows work, V2 github.com/GarrettLab/INA. First, consider luxurious case lot high-confidence promoted (V3 judged, share region present. 2A 2), performed planning stage project, distribution project. 2B, planners ask size increased, so range possible values mean size. cases 1, addition ring contrast. 2C, suppose observations consistent initial conceptualization performing low end frequency outcomes. If efforts increased adoption, perhaps subsidies policies increase uptake, what adoption rates necessary keep progress track? 2D, conclusion status long benefits last successful decline inputs (such educational campaigns), happens over time? Can sizes make up reductions adoption? varied. Suppose less Uncertainty change. For study, parameter varied inverse power law describe movement distance. particular parameters, planned changes system, implemented similarly. relevant collect amenable decision-makers. illustrates modify managers’ control compensate outside control, establishment—due considered. analysis, represented environmental conduciveness reflected establishment. (V4 Sustainability system's ability maintain incidence desired levels probabilities sustainability) jumps plunges needs adjusted back previous level resilience). 3A 3), sustainability’ scenario, increases remains steady time, stressor. absence management) baseline 0.5 0.9. 3B, sustainability’, manageable unmanageable changing. higher due bioentity), compensate, is, rate ‘observed’ before. proportion below 0.2 sustainable no than conditions. 3C, resilience’ stressor, starting 0.05 has leapt 0.50. What needed bring down before leap occurred? 3D, indicates modifying direct stressor unusually then resilient. steps considered? brought resilience, probability? improve influenced outcome experiment. could group testing deciding promote them not. Depending effort invested estimated greater lesser precision. When threshold, communicated, realizations does technology. threshold ranges 0 (communication occurs regardless effect) cannot occur unless complete effectiveness management). analysed standard deviation fairly high, 0.5, (1). Additional examples V4 Note explore costs There overestimated, underestimation overestimation problem science. detection parts periphery 2). degree, shifts, github.com/GarrettLab/INA, described paragraph). similar importance, though link different somewhat little because roles nodes. clearly again only slight driving important. stochastic considered, clear seen deterministic decreased 2A, half V3 responsive generally benefit increasing least 3). weak changing 3) certain (for technology), 0.2. persist push quantification, (determining gradient estimate) 0.7 1.2 2.0 4). However, true 1.2, obtaining precise 3A, 0.9, 5, 3 changes. (given size) 5). pushes Increasing 5), actions reliability lower. recovers stress 0.50 responding number increases, expected 0.8, adequate ascertain worthy 6). studies package. examples, distributions mixture data. section presents ideas expanding address missing information. intended expand future versions incorporate follow applications. above couple INA. One thresholds discussed. Another flat responses Projects stall until factors understood, insensitive readily managed components. identify adaptations expanded integrating layers, maps satellite images species, detailed window opportunity efficient target effectively. strategy (using smartsurv) patterns typical random, surveillance, implicit, Buddenhagen (2017) reliable sources points. Other reasons port, weather conditions resources Beyond archetypal unique properties Users demographic lower surveillance. host populations 2020) characterize landscapes smartsurv. ongoing priority given updated becomes available. succeed INAscene) 2. show options singly succeed, responsiveness Changes attainable further experimentation, subsidies. INAscene, success, updating categories nodes, hypotheses manager gender wealth, illustration challenging precision, gradients, It convenient not, helpful know particularly better. combined estimates cost prioritize next rounds collection ecosystem services operationalizing additional (Biggs Clark Howden 2007; Standish 2014). Experiment illustrations functions’. Adaptation outcomes, return study magnitude adaptation. change, trade, lead links. Variations issues research, integrated optimize design benefits. Agroecological seed Layers formal informal pest seed, (Thomas-Sharma 2016, maintenance (Labeyrie 2016; Pautasso, Pautasso 2013) minimizing grain (Andersen Onofre 2017; Linked breeders who exchange genetic material Priorities defining differ another 7). immediate consideration, try achieve greatest precision interventions, theory priorities trade-offs realism generality (Gross, Levins, 1966). focus theories including combinations Networks characterized exponential graph (ERGMs), test ERGMs (Lusher basis expansions better systems. function, incorporating performance sets specified tracking population multi-scale finer-resolution helps determine move coarser-resolution generation heterogeneous landscape, addressing ‘big data’ form throughout, (Cui 2016). temporal spatial trends parameters. implementation, upscaling formation dissolution bioentities Operationalizing challenges (Standish 2014) option limits feasible within those Major biodiversity meeting food production (Leclère communities problems interests addressed integrate (De Domenico broader practice wide contexts questions, providing spill-over cross-disciplinary learned. As approaches, artificial intelligence support, layers rapidly aim periods likely. Development undertaken of, funded by, CGIAR Research Program Roots, Tubers Bananas (RTB), Trust Fund contributors, USDA NIFA grants 2015-51181-24257 2020-51181-32198. I appreciate US NSF Grant EF-0525712 joint NSF-NIH Ecology Infectious Disease program; DEB-0516046; Climate Change Food Security (CCAFS); APHIS grant 11–8453–1483-CA; USAID Feed Future Haiti Appui à la Recherche au Développement Agricole (AREA) AID-OAA-A-15-00039; Foundation Agriculture FF-NIA19-0000000050; University Florida. contents responsibility author do necessarily reflect views funders, United States Government. Thanks Y. Xing, K. F. Onofre, R. I. Alcalá-Briseño, A. Choudhury P. Garfinkel, reviewers Methods Evolution, stimulating input. work dedicated memory C. B. Garrett J. Garrett. peer review history article https://publons.com/publon/10.1111/2041-210X.13655. code illustrating https://github.com/GarrettLab/INA. An archived version 1.0.0 Zenodo Please note: publisher responsible content functionality any supplied authors. Any queries (other content) should directed article.