Improving the prediction of wildfire susceptibility on Hawai'i Island, Hawai'i, using explainable hybrid machine learning models.

This study presents a comparative analysis of four Machine Learning (ML) models used to map wildfire susceptibility on Hawai'i Island, Hawai'i. Extreme Gradient Boosting (XGBoost) combined with three meta-heuristic algorithms - Whale Optimization (WOA), Black Widow Optimization (BWO), and Butterfly Optimization (BOA) - were employed to map areas susceptible to wildfire. To generate a wildfire inventory, 1408 wildfire points were identified within the study area from 2004 to 2022. The four ML models (XGBoost, WOA-XGBoost, BWO-XGBoost, and BOA-XGBoost) were run using 14 wildfire-conditioning factors categorized into four main groups: topographical, meteorological, vegetation, and anthropogenic. Six performance metrics - sensitivity, specificity, positive predictive values, negative predictive values, the Area Under the receiver operating characteristic Curve (AUC), and the average precision (AP) of Precision-Recall Curves (PRCs) - were used to compare the predictive performance of the ML models. The SHapley Additive exPlanations (SHAP) framework was also used to interpret the importance values of the 14 influential variables for the modeling of wildfire on Hawai'i Island using the four models. The results of the wildfire modeling indicated that all four models performed well, with the BWO-XGBoost model exhibiting a slightly higher prediction performance (AUC = 0.9269), followed by WOA-XGBoost (AUC = 0.9253), BOA-XGBoost (AUC = 0.9232), and XGBoost (AUC = 0.9164). SHAP analysis revealed that the distance from a road, annual temperature, and elevation were the most influential factors. The wildfire susceptibility maps generated in this study can be used by local authorities for wildfire management and fire suppression activity.

Climate change impacts shifting landscape of the dairy industry in Hawai'i.

Proper knowledge and understanding of climatic variability across different seasons are important in farm management. To learn more about the potential effects of climate change on dairying in Hawaii, we conducted a study on site-specific climate characterization using several variables including rainfall, wind speed (WS), solar radiation, and temperature, at two dairy farms located on Hawai`i Island, Hawai`i, in Ookala named "OK DAIRY" and in Upolu Point named "UP DAIRY." Temperature-humidity index (THI) and WS variations in the hottest four months (June to September) were analyzed to determine when critical thresholds that affect animal health are exceeded. Rainfall data were used to estimate the capacity of forage production in 6-mo wet (November to April) and dry (May to October) seasons. Future projections of temperature and rainfall were assessed using mid- and end-century gridded data products for low (RCP 4.5) and high emissions (RCP 8.5) scenarios. Our results showed that the "OK DAIRY" site received higher rainfall than the "UP DAIRY" site, favoring grass growth and forage availability. In addition, the "UP DAIRY" site was more stressful for animals during the summer (THI 69 to 73) than the "OK DAIRY" site (THI 67 to 70) as the THI exceeded the critical threshold of 68, which is conducive for high-lactating cattle. On the "UP DAIRY" site, the THI did not drop below 68 during the summer nights, which created fewer opportunities for cattle to recover from heat stress. Future projections indicated that air temperature would increase 1.3 to 1.8 °C by mid-century and 1.6 to 3.2 °C by the end-century at both farms, and rainfall will increase at the "OK DAIRY" site and decrease at the "UP DAIRY" site by the end-century. The agriculture and livestock industries, particularly the dairy and beef subsectors in Hawai`i, are vulnerable to climate changes as higher temperatures and less rainfall will have adverse effects on cattle. The findings in this study demonstrated how both observed and projected changes in climate support the development of long-term strategies for breeding and holistic livestock management practices to adapt to changing climate conditions.

Climate sensitive size-dependent survival in tropical trees.

Survival rates of large trees determine forest biomass dynamics. Survival rates of small trees have been linked to mechanisms that maintain biodiversity across tropical forests. How species survival rates change with size offers insight into the links between biodiversity and ecosystem function across tropical forests. We tested patterns of size-dependent tree survival across the tropics using data from 1,781 species and over 2 million individuals to assess whether tropical forests can be characterized by size-dependent life-history survival strategies. We found that species were classifiable into four 'survival modes' that explain life-history variation that shapes carbon cycling and the relative abundance within forests. Frequently collected functional traits, such as wood density, leaf mass per area and seed mass, were not generally predictive of the survival modes of species. Mean annual temperature and cumulative water deficit predicted the proportion of biomass of survival modes, indicating important links between evolutionary strategies, climate and carbon cycling. The application of survival modes in demographic simulations predicted biomass change across forest sites. Our results reveal globally identifiable size-dependent survival strategies that differ across diverse systems in a consistent way. The abundance of survival modes and interaction with climate ultimately determine forest structure, carbon storage in biomass and future forest trajectories.

Compilation of climate data from heterogeneous networks across the Hawaiian Islands.

Long-term, accurate observations of atmospheric phenomena are essential for a myriad of applications, including historic and future climate assessments, resource management, and infrastructure planning. In Hawai'i, climate data are available from individual researchers, local, State, and Federal agencies, and from large electronic repositories such as the National Centers for Environmental Information (NCEI). Researchers attempting to make use of available data are faced with a series of challenges that include: (1) identifying potential data sources; (2) acquiring data; (3) establishing data quality assurance and quality control (QA/QC) protocols; and (4) implementing robust gap filling techniques. This paper addresses these challenges by providing: (1) a summary of the available climate data in Hawai'i including a detailed description of the various meteorological observation networks and data accessibility, and (2) a quality controlled meteorological dataset across the Hawaiian Islands for the 25-year period 1990-2014. The dataset draws on observations from 471 climate stations and includes rainfall, maximum and minimum surface air temperature, relative humidity, wind speed, downward shortwave and longwave radiation data.

Erratum to "Impact of uncertainty in soil, climatic, and chemical information in a pesticide leaching assessment".

A simple mobility index, when combined with a geographic information system, can be used to generate rating maps which indicate qualitatively the potential for various organic chemicals to leach to groundwater. In this paper we investigate the magnitude of uncertainty associated with pesticide mobility estimates as a result of data uncertainties. Our example is for the Pearl Harbor Basin, Oahu, Hawaii. The two pesticides included in our analysis are atrazine (2-chloro-4-ethylamino-6-isopropylamino-s-triazine) and diuron [3-(3,4-dichlorophenyl)-1,1-dimethylarea]. The mobility index used here is known as the Attenuation Factor (AF); it requires soil, hydrogeologic, climatic, and chemical information as input data. We employ first-order uncertainty analysis to characterize the uncertainty in estimates of AF resulting from uncertainties in the various input data. Soils in the Pearl Harbor Basin are delineated at the order taxonomic category for this study. Our results show that there can be a significant amount of uncertainty in estimates of pesticide mobility for the Pearl Harbor Basin. This information needs to be considered if future decisions concerning chemical regulation are to be based on estimates of pesticide mobility determined from simple indices.

Water relations and microclimate around the upper limit of a cloud forest in Maui, Hawai'i.

The goal of this study was to determine the effects of atmospheric demand on both plant water relations and daily whole-tree water balance across the upper limit of a cloud forest at the mean base height of the trade wind inversion in the tropical trade wind belt. We measured the microclimate and water relations (sap flow, water potential, stomatal conductance, pressure-volume relations) of Metrosideros polymorpha Gaudich. var. polymorpha in three habitats bracketing the cloud forest's upper limit in Hawai'i to understand the role of water relations in determining ecotone position. The subalpine shrubland site, located 100 m above the cloud forest boundary, had the highest vapor pressure deficit, the least amount of rainfall and the highest levels of nighttime transpiration (EN) of all three sites. In the shrubland site, on average, 29% of daily whole-tree transpiration occurred at night, while on the driest day of the study 50% of total daily transpiration occurred at night. While EN occurred in the cloud forest habitat, the proportion of total daily transpiration that occurred at night was much lower (4%). The average leaf water potential (Ψleaf) was above the water potential at the turgor loss point (ΨTLP) on both sides of the ecotone due to strong stomatal regulation. While stomatal closure maintained a high Ψleaf, the minimum leaf water potential (Ψleafmin) was close to ΨTLP, indicating that drier conditions may cause drought stress in these habitats and may be an important driver of current landscape patterns in stand density.

Transpiration characteristics of a rubber plantation in central Cambodia.

The rapid and widespread expansion of rubber plantations in Southeast Asia necessitates a greater understanding of tree physiology and the impacts of water consumption on local hydrology. Sap flow measurements were used to study the intra- and inter-annual variations in transpiration rate (Et) in a rubber stand in the low-elevation plain of central Cambodia. Mean stand sap flux density (JS) indicates that rubber trees actively transpire in the rainy season, but become inactive in the dry season. A sharp, brief drop in JS occurred simultaneously with leaf shedding in the middle of the dry season in January. Although the annual maxima of JS were approximately the same in the two study years, the maximum daily stand Et of ∼2.0 mm day(-1) in 2010 increased to ∼2.4 mm day(-1) in 2011. Canopy-level stomatal response was well explained by changes in solar radiation, vapor pressure deficit, soil moisture availability, leaf area, and stem diameter. Rubber trees had a relatively small potential to transpire at the beginning of the study period, compared with average diffuse-porous species. After 2 years of growth in stem diameter, transpiration potential was comparable to other species. The sensitivity of canopy conductance (gc) to atmospheric drought indicates isohydric behavior of rubber trees. Modeling also predicted a relatively small sensitivity of gc to the soil moisture deficit and a rapid decrease in gc under extreme drought conditions. However, annual observations suggest the possibility of a change in leaf characteristics with tree maturity and/or initiation of latex tapping. The estimated annual stand Et was 469 mm year(-1) in 2010, increasing to 658 mm year(-1) in 2011. Diagnostic analysis using the derived gc model showed that inter-annual change in stand Et in the rapidly growing young rubber stand was determined mainly by tree growth rate, not by differences in air and soil variables in the surrounding environment. Future research should focus on the potentially broad applicability of the relationship between Et and tree size as well as environmental factors at stands different in terms of clonal type and age.

Moisture status during a strong El Niño explains a tropical montane cloud forest's upper limit.

Growing evidence suggests short-duration climate events may drive community structure and composition more directly than long-term climate means, particularly at ecotones where taxa are close to their physiological limits. Here we use an empirical habitat model to evaluate the role of microclimate during a strong El Niño in structuring a tropical montane cloud forest's upper limit and composition in Hawai'i. We interpolate climate surfaces, derived from a high-density network of climate stations, to permanent vegetation plots. Climatic predictor variables include (1) total rainfall, (2) mean relative humidity, and (3) mean temperature representing non-El Niño periods and a strong El Niño drought. Habitat models explained species composition within the cloud forest with non-El Niño rainfall; however, the ecotone at the cloud forest's upper limit was modeled with relative humidity during a strong El Niño drought and secondarily with non-El Niño rainfall. This forest ecotone may be particularly responsive to strong, short-duration climate variability because taxa here, particularly the isohydric dominant Metrosideros polymorpha, are near their physiological limits. Overall, this study demonstrates moisture's overarching influence on a tropical montane ecosystem, and suggests that short-term climate events affecting moisture status are particularly relevant at tropical ecotones. This study further suggests that predicting the consequences of climate change here, and perhaps in other tropical montane settings, will rely on the skill and certainty around future climate models of regional rainfall, relative humidity, and El Niño.

Changing climate and the altitudinal range of avian malaria in the Hawaiian Islands - an ongoing conservation crisis on the island of Kaua'i.

Transmission of avian malaria in the Hawaiian Islands varies across altitudinal gradients and is greatest at elevations below 1500 m where both temperature and moisture are favorable for the sole mosquito vector, Culex quinquefasciatus, and extrinsic sporogonic development of the parasite, Plasmodium relictum. Potential consequences of global warming on this system have been recognized for over a decade with concerns that increases in mean temperatures could lead to expansion of malaria into habitats where cool temperatures currently limit transmission to highly susceptible endemic forest birds. Recent declines in two endangered species on the island of Kaua'i, the 'Akikiki (Oreomystis bairdi) and 'Akeke'e (Loxops caeruleirostris), and retreat of more common native honeycreepers to the last remaining high elevation habitat on the Alaka'i Plateau suggest that predicted changes in disease transmission may be occurring. We compared prevalence of malarial infections in forest birds that were sampled at three locations on the Plateau during 1994-1997 and again during 2007-2013, and also evaluated changes in the occurrence of mosquito larvae in available aquatic habitats during the same time periods. Prevalence of infection increased significantly at the lower (1100 m, 10.3% to 28.2%), middle (1250 m, 8.4% to 12.2%), and upper ends of the Plateau (1350 m, 2.0% to 19.3%). A concurrent increase in detections of Culex larvae in aquatic habitats associated with stream margins indicates that populations of the vector are also increasing. These increases are at least in part due to local transmission because overall prevalence in Kaua'i 'Elepaio (Chasiempis sclateri), a sedentary native species, has increased from 17.2% to 27.0%. Increasing mean air temperatures, declining precipitation, and changes in streamflow that have taken place over the past 20 years are creating environmental conditions throughout major portions of the Alaka'i Plateau that support increased transmission of avian malaria.

The projected timing of climate departure from recent variability.

Ecological and societal disruptions by modern climate change are critically determined by the time frame over which climates shift beyond historical analogues. Here we present a new index of the year when the projected mean climate of a given location moves to a state continuously outside the bounds of historical variability under alternative greenhouse gas emissions scenarios. Using 1860 to 2005 as the historical period, this index has a global mean of 2069 (±18 years s.d.) for near-surface air temperature under an emissions stabilization scenario and 2047 (±14 years s.d.) under a 'business-as-usual' scenario. Unprecedented climates will occur earliest in the tropics and among low-income countries, highlighting the vulnerability of global biodiversity and the limited governmental capacity to respond to the impacts of climate change. Our findings shed light on the urgency of mitigating greenhouse gas emissions if climates potentially harmful to biodiversity and society are to be prevented.

Climate-associated population declines reverse recovery and threaten future of an iconic high-elevation plant.

Although climate change is predicted to place mountain-top and other narrowly endemic species at severe risk of extinction, the ecological processes involved in such extinctions are still poorly resolved. In addition, much of this biodiversity loss will likely go unobserved, and therefore largely unappreciated. The Haleakala silversword is restricted to a single volcano summit in Hawai'i, but is a highly charismatic giant rosette plant that is viewed by 1-2 million visitors annually. We link detailed local climate data to a lengthy demographic record, and combine both with a population-wide assessment of recent plant mortality and recruitment, to show that after decades of strong recovery following successful management, this iconic species has entered a period of substantial climate-associated decline. Mortality has been highest at the lower end of the distributional range, where most silverswords occur, and the strong association of annual population growth rates with patterns of precipitation suggests an increasing frequency of lethal water stress. Local climate data confirm trends toward warmer and drier conditions on the mountain, and signify a bleak outlook for silverswords if these trends continue. The silversword example foreshadows trouble for diversity in other biological hotspots, and illustrates how even well-protected and relatively abundant species may succumb to climate-induced stresses.

Simulating land-cover change in Montane mainland southeast Asia.

We used the conversion of land use and its effects (CLUE-s) model to simulate scenarios of land-cover change in Montane mainland southeast Asia (MMSEA), a region in the midst of transformation due to rapid intensification of agriculture and expansion of regional trade markets. Simulated changes affected approximately 10 % of the MMSEA landscape between 2001 and 2025 and 16 % between 2001 and 2050. Roughly 9 % of the current vegetation, which consists of native species of trees, shrubs, and grasses, is projected to be replaced by tree plantations, tea, and other evergreen shrubs during the 50 years period. Importantly, 4 % of this transition is expected to be due to the expansion of rubber (Hevea brasiliensis), a tree plantation crop that may have important implications for local-to-regional scale hydrology because of its potentially high water consumption in the dry season.

Reduction of stream sediment concentration by a riparian buffer: filtering of road runoff in disturbed headwater basins of montane mainland southeast Asia.

We determined the extent that a riparian buffer reduces stream suspended sediment concentrations by filtering road runoff during 18 rain events in a 2.5-ha, multi-use watershed in northern Thailand. The dominant buffer species was the perennial sedge Fimbristylis aphylla Zoll. ex Steud. (Cyperaceae). We monitored stream sediment concentration for situations where road runoff either flowed into the riparian buffer or was diverted directly into the stream (buffer and no buffer scenarios). These data were used to develop the following relationships between instantaneous stream sediment concentration (Ci) and discharge (Qi): Ci= 28.329Qi(0.851) (buffer scenario) and Ci= 22.265Qi(1.579) (no buffer scenario). Using these functions to calculate total event suspended concentrations, we determined that the buffer reduced suspended sediment concentration by 34 to 87%, for the range of events monitored. Removal of sediment from runoff generated on a 2.4-m-wide, 165-m-long unpaved road section was achieved principally via ponding, which reduced the transport capacity as flow entered the relatively flat, saturated buffer. Sediment deposition occurred primarily within the first 10 m of the buffer. Some sediment was also deposited on the fillslope leading to the buffer. Maximum road sediment concentration during the largest buffer event approached 100,000 mg L(-1). Meanwhile, the corresponding maximum stream suspended sediment concentration was <4000 mg L(-1). In contrast, maximum stream concentrations when flow bypassed the buffer during smaller events were commonly 4000 to 7000 mg L(-1). Naturally occurring buffers represent an economical means of mitigating road-related impacts in upland basins in Southeast Asia, particularly if combined with measures limiting sediment and runoff production on contributing road sections.