Application of genetic programming to project climate change impacts on the population of Formosan Landlocked Salmon
Introduction
Formosan Landlocked Salmon (Oncorhynchus masou formosanus) is a very important species as it is the salmon living at the lowest latitude. Although widely distributed in the upper Dajia Creek in early 1900s, it is currently only found in the uppermost reaches of Dajia Creek, Chichiawan Creek and Gaoshan Creek tributaries as in Fig. 1. Human-induced land use changes, hydraulic structures, poor water quality, increased water temperature and flooding are principal factors reducing habitat and endangering Formosan Landlocked Salmon. Conversion of land uses from forest to agriculture has been forbidden in the upper Dajia Creek, and some hydraulic structures have been removed and new construction prohibited. The factor endangering Formosan Landlocked Salmon most will be climate change. Climate change may influence streamflows, water temperature and habitats.
Increased water temperature, the primary effect of climate change, affects growth and survival rates of fish and other aquatic creatures, as well as migration patterns, breeding, and competitive ability of fish. Moreover, transmission of pollutants and resulting chemical reactions are also influenced by water temperature. These landlocked salmon only live in a water temperature range of 9–17 °C and, during spawning, the temperature must be below 12 °C (Tzeng, 1999). Generally, water temperature increases as it flows downstream. According to recent surveys (1985–1997), the 12 °C isotherm has moved 1.56 km upstream in the upper section of Dajia creek (Yang, 1997). This reduction in suitable spawning stream habitats can seriously influence the distribution of these landlocked salmon. Additionally, high flow rates also impact Formosan salmon. Sediment and increased flow generated by storm events degraded habitat and flushed fish downstream into inhospitable reaches. However, increased streamflow can also benefit habitats by reducing any increases in water temperature.
Numerous researchers, who have investigated the relationship between fish populations and environmental conditions, have identified decreasing populations of salmon and trout. Bradford and Irvine (2000) evaluated the impact of land use change, climate change and over-fishing on the population density of salmon. Shaw and Richardson (2001) assessed the direct and indirect effects of sediment pulse duration on rainbow trout survival. Cattaneo et al. (2002) concluded that flooding is the predominant factor influencing salmon and trout populations. Developed by Chen et al. (2000), the fuzzy logic model can assess the functional relationship between the number of spawners and sea surface temperature, reconstruct historical fish recruitment time series and predict the future fish populations.
According to IPCC 2007, ambient CO2 concentrations will increase mean global air temperature by 1.1–6.4 °C by 2100 (IPCC WG I, 2007). Moreover, recent local research concluded that climate change may result in increased stream temperatures (Tung and Lee, 2006) and flows (Tung and Lee, 2001) in Formosan Landlocked Salmon's habitat. Thus, identifying response functions for assessing the impacts of climate change on Formosan Landlocked Salmon is necessary.
Differing environmental conditions impact ecosystems differently; thus, a predefined response function will likely limit predictability. In this study, environmental conditions are classified into several categories and a response function is then optimized by genetic programming (GP) for each category. GP is utilized here because it can find functions of various forms, not merely those involving a pre-determined polynomial with unknown coefficients. GP enhances the predictability particularly for this unknown situation. The study's goal is to identify environmental impact functions and their parameters for Formosan Landlocked Salmon. Section 2 describes the method employed to optimize response functions and Section 3 presents an analysis of critical environmental factors. Section 4 then addresses the method of evaluating climate change impacts on specific environmental factors. Simulation results and conclusions are presented in Sections 5 Results and discussions, 6 Climate change impacts on Formosan Landlocked Salmon, respectively.
Section snippets
Development of response functions
Different environmental conditions can have varying impacts of varying magnitudes on an ecosystem; thus, different functions may be required to predict the effects of various environmental conditions. GP generating functions helps us to know the response of a fish population to various environmental conditions. Furthermore, the functions can be applied to predict possible fish populations in the future. Environmental conditions in this study are categorized prior to optimizing response
Analysis of environmental factors
Considerable researchers have concluded that the most important factors affecting fish populations include stream temperature, flow, sediment and nutrient levels (Yang, 1997, Shaw and Richardson, 2001, Cattaneo et al., 2002). The nutrient concentration meets water quality standards in the study area and is not considered to be a factor limiting fish population stability or growth. Furthermore, based on monitoring data from the study area, three factors were selected for further analysis: water
Climate change impact assessment
This section describes the procedure employed to assess the impact of climate change. First, future climate change scenarios derived from GCM simulations must be input into a weather generation model to produce weather data. Second, average daily air temperature and maximal streamflow in different climate conditions require evaluation. Then, the resulting data is input into the identified response functions to assess the impact of climate change on the population of Formosan Landlocked Salmon.
Results and discussions
The response functions identified are addressed first. These functions are then applied to assess the impact of climate changes on Formosan Landlocked Salmon.
Climate change impacts on Formosan Landlocked Salmon
Table 7, Table 8 show the impacts on Formosan Landlocked Salmon under High Temperature/High Flow and High Temperature/Low Flow for different climate conditions. Fig. 12 presents a plot of the 25th, 50th, and 75th percentiles of the fish population for the two categories under different climate conditions. As the response function of Low Temperature/Low Flow did not successfully provide predictability, it was not further applied in the climate change impact evaluation.
There was a decrease for
Conclusions
This study proposes a procedure for identifying environmental response functions for Formosan Landlocked Salmon under different environmental conditions, to which GP is applied to optimize both the type and coefficients of response functions. Identified response functions were further applied to assess the impacts of climate change on Formosan Landlocked Salmon. Two principal environmental factors, average daily temperature and maximal flow, were used in categorizing data for the study area.
The
Acknowledgements
We sincerely thank the SHEI-PA National Park Management Office and Professor Tzeng from National Tsing Hua University for providing data on Formosan Landlocked Salmon. The authors would also like to thank the National Science Council Republic of China, Taiwan (Contract No. NSC 90-2313-B-002-328) and the Environmental Protection Agency of Taiwan for financially supporting this research.
References (15)
- et al.
Stream temperature/air temperature relationship: a physical interpretation
Journal of Hydrology
(1999) - et al.
Land use, fishing, climate change, and the decline of Thompson River, British Columbia, Coho salmon
Canadian Journal of Fisheries and Aquatic Sciences
(2000) - et al.
The influence of hydrological and biotic process on brown trout (Salmo trutta) population dynamics
Canadian Journal of Fisheries and Aquatic Sciences
(2002) - et al.
A fuzzy logic model with genetic algorithm for analyzing fish stock−recruitment relationships
Canadian Journal of Fisheries and Aquatic Sciences
(2000) - et al.
Generalized watershed loading functions for streamflow nutrients
Water Resource Bulletin
(1987) Genetic Programming: On the Programming of Computers by Means of Natural Selection
(1992)
Cited by (16)
Factors affecting the presence of Arctic charr in streams based on a jittered binary genetic programming model
2022, Ecological IndicatorsCitation Excerpt :To tackle these problems, this study introduces a novel ML technique that combines data augmentation and resampling techniques with a relatively new ML technique, state-of-the-art genetic programming (GP), which is empowered by a classifier root node. Despite the promising applications of GP for knowledge discovery around hydrological phenomena (Herath et al., 2021), current literature has produced few investigations on the potential use of GP for ecohydrological process modeling (e.g., Muttil and Chau 2006; Tung et al., 2009; Sanderson 2009; Wang et al., 2012). In this study, a new model was developed using a dataset extracted through in-situ and remote measurements and applied to describe the physical and ecological factors that influence the occurrence and utilization of various habitats in the large sub-Arctic Teno catchment.
Rice yield response forecasting tool (YIELDCAST) for supporting climate change adaptation decision in Sahel
2020, Agricultural Water ManagementCitation Excerpt :Gene-expression programming (GEP) is one of the most powerful machine-coding sources for solving nonlinear problem. Basically, gene-expression programming has demonstrated outstanding results in the subject of climate change and agriculture engineering, as reported in the work of Guven et al. (2007), Tung et al. (2009), Kisi and Guven (2010), Guven and Kisi (2011), Traore and Guven (2011, 2012) and Britaldo et al. (2013). Although the GEP has recently received a great deal of attention, note that Sahel is still lagging behind in finding yield related-codes to its typical weather condition.
A dry season streamflow reconstruction of the critically endangered Formosan landlocked salmon habitat
2018, DendrochronologiaCitation Excerpt :Despite great efforts to protect and restore the species over the past several decades, the natural population size is estimated to range from only a few hundred to perhaps a thousand individuals and has been listed as Critically Endangered by the IUCN (Kottelat, 1996). Because the Formosan landlocked salmon requires cold water (< 18 °C) to persist, the current focus in assessing the possible climate change impacts on the species is mainly stream water temperature change (Tung et al., 2006, 2009). One equally important but less understood factor is how climate change may impact the flow regime of CCWS.
Multi-genes programing and local scale regression for analyzing rice yield response to climate factors using observed and downscaled data in Sahel
2014, Agricultural Water ManagementCitation Excerpt :To explore yield response function to current and future climate variables, Gene-expression programing computation model (GEP) was employed in this study, as the program has been used with success in statistical downscaling and agriculture engineering research in the recent past. It was recently reported that GEP can generate local scale climate data information from the global circulation model outputs (Tung et al., 2009). Britaldo et al. (2013) used it to calibrate land-use change models, and Traore and Guven (2011, 2012) deployed it for modeling evapotranspiration for irrigation purposes.
A data-driven approach for modeling post-fire debris-flow volumes and their uncertainty
2011, Environmental Modelling and SoftwareCitation Excerpt :A summary of the evolutionary input used in this study is presented in Table 4. Because these parameters are well-described in the literature (Koza, 1992; Koza et al., 1999; Babovic and Keijzer, 2000; Tung et al., 2009), only a brief overview of selected evolutionary parameters is presented. Population model: The generational model uses an intermediate mating pool where the selected individuals reside to form the next generation of offspring.
Automated Discovery of Relationships, Models, and Principles in Ecology
2020, Frontiers in Ecology and Evolution