Search

Deep learning (DL) has recently emerged as an innovative tool to downscale climate variables from large-scale atmospheric fields under the perfect-prognosis (PP) approach. Different convolutional neural networks (CNNs) have been applied under present-day conditions with promising results, but little is known about their suitability for extrapolating future climate change conditions. Here, we analyze this problem from a multi-model perspective, developing and evaluating an ensemble of CNN-based downscaled projections (hereafter DeepESD) for temperature and precipitation over the European EUR-44i (0.5∘) domain, based on eight global circulation models (GCMs) from the Coupled Model Intercomparison Project Phase 5 (CMIP5). To our knowledge, this is the first time that CNNs have been used to produce downscaled multi-model ensembles based on the perfect-prognosis approach, allowing us to quantify inter-model uncertainty in climate change signals. The results are compared with those corresponding to an EUR-44 ensemble of regional climate models (RCMs) showing that DeepESD reduces distributional biases in the historical period. Moreover, the resulting climate change signals are broadly comparable to those obtained with the RCMs, with similar spatial structures. As for the uncertainty of the climate change signal (measured on the basis of inter-model spread), DeepESD preserves the uncertainty for temperature and results in a reduced uncertainty for precipitation. To facilitate further studies of this downscaling approach, we follow FAIR principles and make publicly available the code (a Jupyter notebook) and the DeepESD dataset. In particular, DeepESD is published at the Earth System Grid Federation (ESGF), as the first continental-wide PP dataset contributing to CORDEX (EUR-44).

The increasing demand for high-resolution climate information has attracted growing attention to statistical downscaling (SDS) methods, due in part to their relative advantages and merits as compared to dynamical approaches (based on regional climate model simulations), such as their much lower computational cost and their fitness for purpose for many local-scale applications. As a result, a plethora of SDS methods is nowadays available to climate scientists, which has motivated recent efforts for their comprehensive evaluation, like the VALUE initiative (http://www.value-cost.eu, last access: 29 March 2020). The systematic intercomparison of a large number of SDS techniques undertaken in VALUE, many of them independently developed by different authors and modeling centers in a variety of languages/environments, has shown a compelling need for new tools allowing for their application within an integrated framework. In this regard, downscaleR is an R package for statistical downscaling of climate information which covers the most popular approaches (model output statistics – including the so-called “bias correction” methods – and perfect prognosis) and state-of-the-art techniques. It has been conceived to work primarily with daily data and can be used in the framework of both seasonal forecasting and climate change studies. Its full integration within the climate4R framework (Iturbide et al., 2019) makes possible the development of end-to-end downscaling applications, from data retrieval to model building, validation, and prediction, bringing to climate scientists and practitioners a unique comprehensive framework for SDS model development. In this article the main features of downscaleR are showcased through the replication of some of the results obtained in VALUE, placing an emphasis on the most technically complex stages of perfect-prognosis model calibration (predictor screening, cross-validation, and model selection) that are accomplished through simple commands allowing for extremely flexible model tuning, tailored to the needs of users requiring an easy interface for different levels of experimental complexity. As part of the open-source climate4R framework, downscaleR is freely available and the necessary data and R scripts to fully replicate the experiments included in this paper are also provided as a companion notebook.

Deep learning techniques (in particular convolutional neural networks, CNNs) have recently emerged as a promising approach for statistical downscaling due to their ability to learn spatial features from huge spatiotemporal datasets. However, existing studies are based on complex models, applied to particular case studies and using simple validation frameworks, which makes a proper assessment of the (possible) added value offered by these techniques difficult. As a result, these models are usually seen as black boxes, generating distrust among the climate community, particularly in climate change applications. In this paper we undertake a comprehensive assessment of deep learning techniques for continental-scale statistical downscaling, building on the VALUE validation framework. In particular, different CNN models of increasing complexity are applied to downscale temperature and precipitation over Europe, comparing them with a few standard benchmark methods from VALUE (linear and generalized linear models) which have been traditionally used for this purpose. Besides analyzing the adequacy of different components and topologies, we also focus on their extrapolation capability, a critical point for their potential application in climate change studies. To do this, we use a warm test period as a surrogate for possible future climate conditions. Our results show that, while the added value of CNNs is mostly limited to the reproduction of extremes for temperature, these techniques do outperform the classic ones in the case of precipitation for most aspects considered. This overall good performance, together with the fact that they can be suitably applied to large regions (e.g., continents) without worrying about the spatial features being considered as predictors, can foster the use of statistical approaches in international initiatives such as Coordinated Regional Climate Downscaling Experiment (CORDEX).

The use of Sentinel-3 Ocean and Land Color Instrument (OLCI) images in estimating chlorophyll-a (total and class-differentiated)a concentration is promising owing to Sentinel-3’s 21 bands. This was investigated for the case of Laguna de Bay (or Laguna Lake), Philippines. Field surveys were conducted on 13–17 November 2018 using FluoroProbe, a submersible fluorimeter capable of quantifying concentrations of spectral classes of microalgae. These were regressed with reflectance data obtained from 10-day composite Sentinel-3 reflectance images as well as ten empirical algorithms (indices) for OLCI. Compared to band reflectance, the 10 indices yielded stronger correlations, especially with R665/R709, R674/R709, and (1/R665-1/R709)xR754 with the following respective correlation values: −0.623, −0.646, and 0.628. Multiple regression results indicates that 48% of the variability of total chl-a concentration is explained by five explanatory (reflectance) variables (R412, R443, R560, R681, and R754) with RMSE of 2.814 μg/l. In contrast, the two indices R674/R754 and (1/R665-1/R709)xR754 accounted for about 46% of the variability of total chl-a concentration with RMSE of 2.475 μg/l. For diatoms and bluegreen microalgae, R560/R665 and (1/R665-1/R709)xR754 constitute the models with R2 of 0.21 and 0.435, and RMSE of 2.516 and 2.163 ug/l, respectively. Green microalgal concentration is jointly described by three indices: R560/R665, R674/R754, and R709-R754, with R2 = 0.182 and RMSE = 1.219 μg/l. From cryptophytes, the model comprising of R560/R665, (1/R665-1/R709)xR754, and R709-R754 produced an R2 = 0.289 and RMSE = 0.767 μg/l. It can be said that the empirical algorithms can be used for Sentinel-3 OLCI data providing acceptable estimations of total and spectral class-differentiated chl-a concentration.

Land use and land cover monitoring is an important component in the management of Laguna Lake watershed due to its impacts on the lake’s water quality. Due to limitations caused by cloud cover, satellite systems with limited revisit capability fail to provide sufficient data to more effectively monitor the land surface. Normalized difference vegetation index (NDVI) derived from MODIS image data were used to generate land cover maps for the years 2001, 2005, 2009, 2013, and 2017. These were produced by classifying ISODATA classes using annual NDVI profiles, which resulted in land cover classes, namely, agricultural land, built-up, forest, rangeland, water, and wetland. The resulting maps were post-processed using multi-variate alteration detection (MAD), resulting in multi-temporal land cover maps with improved overall accuracies and kappa coefficients that indicate moderate agreement with ground truth data. Spatiotemporal hot spot analysis was also performed using NDVI data from 2001 to 2017 to identify vegetation hot spot areas, where clustering of low NDVI values were observed over the years. Results showed an increasing trend in built-up areas accompanied by decreasing trends in water and wetland areas, indicating impacts caused by land reclamation and expansion of residential subdivisions near the lakeshore. The decrease in total vegetation area from 2001 to 2017 could be attributed to conversion of land to built-up surface. Vegetated areas in identified hot spots decreased from 41% in 2001 to 19% in 2017. This suggests that vegetation cover in these hot spots was converted to non-vegetated surface during the time period studied.

An indirect effect of global warming is a reduction in the depth of the upper mixed layer (UML) causing organisms to be exposed to higher levels of ultraviolet (UVR, 280–400 nm) and photosynthetically active radiation (PAR, 400–700 nm). This can affect primary and bacterial production as well as the commensalistic phytoplankton–bacteria relationship. The combined effects of UVR and reduction in the depth of the UML were assessed on variables related to the metabolism of phytoplankton and bacteria, during in situ experiments performed with natural pico- and nanoplankton communities from two oligotrophic lakes with contrasting UVR transparency (high-UVR versus low-UVR waters) of southern Spain. The negative UVR effects on epilimnetic primary production (PP) and on heterotrophic bacterial production (HBP), intensified under increased stratification, were higher in the low-UVR than in the high-UVR lake, and stronger on the phytoplanktonic than on the heterotrophic bacterial communities. Under UVR and increased stratification, the commensalistic phytoplankton–bacteria relationship was strengthened in the high-UVR lake where excretion of organic carbon (EOC) rates exceeded the bacterial carbon demand (BCD; i.e., BCD : EOC(%) ratio < 100). This did not occur in the low-UVR lake (i.e., BCD : EOC(%) ratio > 100). The greater UVR damage to phytoplankton and bacteria and the weakening of their commensalistic interaction found in the low-UVR lake indicates that these ecosystems would be especially vulnerable to UVR and increased stratification as stressors related to global climate change. Thus, our findings may have important implications for the carbon cycle in oligotrophic lakes of the Mediterranean region.

Global change, together with human activities, has resulted in increasing amounts of organic material (including nutrients) that water bodies receive. This input further attenuates the penetration of solar radiation, leading to the view that opaque lakes are more "protected" from solar ultraviolet radiation (UVR) than clear ones. Vertical mixing, however, complicates this view as cells are exposed to fluctuating radiation regimes, for which the effects have, in general, been neglected. Furthermore, the combined impacts of mixing, together with those of UVR and nutrient inputs are virtually unknown. In this study, we carried out complex in situ experiments in three high mountain lakes of Spain (Lake Enol in the National Park Picos de Europa, Asturias, and lakes Las Yeguas and La Caldera in the National Park Sierra Nevada, Granada), used as model ecosystems to evaluate the joint impact of these climate change variables. The main goal of this study was to address the question of how short-term pulses of nutrient inputs, together with vertical mixing and increased UVR fluxes modify the photosynthetic responses of phytoplankton. The experimentation consisted in all possible combinations of the following treatments: (a) solar radiation: UVR &plus; PAR (280–700 nm) versus PAR (photosynthetically active radiation) alone (400–700 nm); (b) nutrient addition (phosphorus (P) and nitrogen (N)): ambient versus addition (P to reach to a final concentration of 30 μg P L−1, and N to reach N:P molar ratio of 31); and (c) mixing: mixed (one rotation from surface to 3 m depth (speed of 1 m 4 min−1, total of 10 cycles)) versus static. Our findings suggest that under ambient nutrient conditions there is a synergistic effect between vertical mixing and UVR, increasing phytoplankton photosynthetic inhibition and excretion of organic carbon (EOC) from opaque lakes as compared to algae that received constant mean irradiance within the epilimnion. The opposite occurs in clear lakes where antagonistic effects were determined, with mixing partially counteracting the negative effects of UVR. Nutrient input, mimicking atmospheric pulses from Saharan dust, reversed this effect and clear lakes became more inhibited during mixing, while opaque lakes benefited from the fluctuating irradiance regime. These climate change related scenarios of nutrient input and increased mixing, would not only affect photosynthesis and production in lakes, but might also further influence the microbial loop and trophic interactions via enhanced EOC under fluctuating UVR exposure.

Freshwater bodies like Taal Lake play a pivotal role in providing essential resources like fresh drinking water and in supporting local livelihoods. This study aimed to examine the environmental conditions of Taal Lake by quantifying chlorophyll-a (chl-a) and turbidity levels with Water Colour Simulator (WASI) and water quality indices, specifically with Sentinel-2 and PRISMA imagery. The results from the satellite image-derived data revealed discernible variations in chlorophyll-a and turbidity concentrations across different regions of Taal Lake. Higher chlorophyll-a was consistently observed in the western regions (1.3–5.9 μg/L for Sentinel-2; 1.4–5.4 μg/L for PRISMA), while lower concentrations were found in the south (1.7–3.4 μg/L for Sentinel-2; 1.9–3.3 μg/L for PRISMA). Meanwhile, turbidity values were higher in the eastern and northeastern parts (0.15–0.59 mg/L for Sentinel-2; 0.13–0.51 mg/L for PRISMA). Water quality indices (NDTI & NDCI) also supported these findings. The findings of this comparative analysis between PRISMA hyperspectral and Sentinel-2 multispectral imagery demonstrate the potential of both satellite systems in providing valuable insights into the spatial distribution of water quality parameters in Taal Lake. Nonetheless, discrepancies observed in the scatter plots and inversion failures in PRISMA underscore the need for further research and refinement in utilizing PRISMA data with the WASI model. This study highlights the significance of freshwater bodies and the importance of monitoring their health for the well-being of communities and the environment.

Boracay Island, renowned for its pristine beaches, faced challenges arising from haphazard practices on infrastructure and tourism management. A rehabilitation program was initiated which resulted in a significant reconstruction on its land. This study investigates the impact of Land Cover (LC) change caused by the rehabilitation program on water quality parameters (chl-a and TSM) across three key periods: pre-rehabilitation (2017), rehabilitation (2020), and post-rehabilitation (2022). Utilizing Sentinel-2 imagery, quarterly assessments of chl-a and TSM concentrations are conducted to evaluate water quality. The findings revealed significant variations in chl-a and TSM with significance values of 0.043 and 0.014, respectively, within zone 1 of Boracay Island. Post-hoc test revealed differences in chl-a concentrations between 2017 and subsequent years (2020, p = 0.038; 2021, p = 0.036 ; and 2022, p = 0.039) and between 2019 and subsequent years (2020, p = 0.043; 2021, p = 0.041; and 2022, p = 0.044); as well as in TSM concentrations between 2017 and subsequent years (2020, p = 0.006; 2021, p = 0.008; and 2022, p = 0.023) and between 2019 and subsequent years (2020, p = 0.014; 2021, p = 0.019; and 2022, p = 0.049). However, no significant correlation exists between LC and water quality parameters. It is recommended to refine LC classification through high-resolution satellite imagery, consider extended timeframes, and investigate additional water quality parameters. This study contributes valuable insights into Boracay Island’s environmental transformations, guiding sustainable coastal management in similar contexts.

The Philippines has imported 93% (550,000 MT) of its annual salt requirement in 2020. Such heavy reliance on salt importation may be attributed to the two decades of neglect and steady downward production trend of the local salt industry. There is a need to increase local salt production because Filipinos consume salt on a daily basis for over 14,000 uses, ranging from being a household commodity to being utilized for agricultural and industrial processes. It is therefore imperative to increase the number of areas allocated for local salt production. In this study, a suitability analysis for solar salt farm operations in Pangasinan was conducted using geospatial data and Analytic Hierarchy Process (AHP). Physical land variables (slope, land cover, distance from coastline, distance from river, and soil texture) and meteorological factors (temperature, humidity, rainfall, wind speed) were considered to assess the suitability of an area for salt production. Results showed that meteorological factors (62.5%) outweigh the physical land variables (37.5%) by a 25% margin, suggesting the heavy influence of climate on salt production. 4629.88 km2 (or 86.6% of the total land area of Pangasinan) was identified to be suitable for salt production with 496.01 km2 (9.5%) being moderately suitable and 3800.87 km2 (77.1%) being rated as low. The moderately suitable areas were validated using existing salt farm locations in two sites, Dasol and Alaminos City, and showed 62.13% and 28.06% overlapping areas, respectively. The main explanation for these relatively low percentage figures is the gaps in the output suitability map due to the slope and soil datasets. The concentration of the moderately suitable areas is situated along coastal municipalities of Pangasinan including Bolinao, Dasol, Alaminos City, Sual, Agno, Labrador, Infanta, San Fabian, Anda, and Bani. The findings of this study can help policymakers, relevant government agencies and stakeholders in the planning process of revitalizing the salt industry by expanding existing salt farms or constructing new ones in Pangasinan.

The need for trees in cities is rapidly increasing because of global warming, rapid population growth, and urbanization. To maximize the benefits of trees, areas of greatest need should be identified and prioritized. In this study, a geospatial framework for tree planting prioritization is developed for NCR using remotely sensed datasets, AHP, and GIS. A Planting Priority Index (PPI) for NCR is derived based on five criteria with corresponding weights calculated based on ratings from 18 experts – air quality (25.2%), tree cover (23.8%), land cover (17.8%), population (17.4%), and land surface temperature (15.8%). The AHP resulted in a consistency ratio (CR) value of 1.7% and a consensus value of 53.2%. PPI values were calculated and statistically significant prioritization clusters and outliers were identified using Anselin Local Moran’s I statistics. The resulting PPI and cluster maps showed that the high-priority areas for tree planting clustered near the region’s center and northwest portion covering the cities of Malabon, South Caloocan, Navotas, Quezon City, Marikina, San Juan, and Mandaluyong, while the low priority areas were found mostly along the region’s outskirts at cities of Pasay, Las Piñas, Muntinlupa, Taguig, Valenzuela, and North Caloocan. The generated maps showing PPI values across the region may aid local government agencies and environmental organizations in evaluating and recalibrating their local greening programs. The workflow presented in this study can also be adopted in other regions with localized variables and site-specific goals relevant to tree planting and greening programs.

Water quality monitoring is vital in ensuring the suitability of Lake Buhi for aquaculture and recreational activities. However, with the limitations of traditional monitoring methods, the Bureau of Fisheries and Aquatic Resources (BFAR) cannot comprehensively characterize the spatial and temporal water quality patterns in the lake. With the development of remote sensing methods, this study aimed to describe the spatial and temporal variability of turbidity in Lake Buhi in 2020 using Geographically Weighted Regression (GWR) and the Normalized Difference Turbidity Index (NDTI) applied to Sentinel-2 images. GWR provided accurate estimates of turbidity, with the results achieving an R2 value of 0.98 for the February image and 0.93 for October. GWR-derived turbidity was 2 used as an alternative to in situ data and then regressed with NDTI values to acquire turbidity models for dry and wet seasons. Upon validation, the best turbidity model for the dry (R2 = 0.59, RMSE = 0.60, MAE = 0.15) and wet (R2 = 0.49, RMSE = 1.22, MAE = 0.25) seasons produced acceptable results, hence, used to assess the spatial and temporal variability of turbidity in the lake throughout 2020. The analysis revealed that Lake Buhi is more turbid during the dry season than the wet season. Turbidity during the dry season is governed by its natural water flow, while it is heavily influenced by precipitation during the wet season.

Optical water type (OWT) classification provides a way to delineate surface waters according to spectral properties caused by optically active constituents. This classification is useful in mapping out spatial variability and describing changes in water quality over time as it uses inherent optical properties to discriminate areas of a water body. We used Sentinel-3 OLCI images from July 2017 to December 2020 to classify OWT in Manila Bay and to generate chlorophyll-a and total suspended matter (TSM) concentration layers using C2RCC algorithm. Nine coastal OWT classes for Manila Bay having consistent presence and spatial distribution seasonally were identified. The most dominant class, OWT 6, (3.50 µg/cm3 chlorophyll-a; 4.93 g/m3 TSM) was concentrated around the central area of the bay, covering as much as 79% of it. Chlorophyll-a and TSM concentrations were higher in the nearshore areas and were decreasing offshore. The class with the second highest concentrations, OWT 7, (9.11 µg/cm3 chlorophyll-a; 26.80 g/m3 TSM) were found near river outlets, where high nutrient and sediment loadings coming from the watershed enter the bay. Lowest concentrations (0.26–0.70 µg/cm3 chlorophyll-a; 0.64–1.10 g/m3 TSM) were found in OWT 1, OWT 2, and OWT 3, all located farthest from the shore, such as the mouth of the bay leading to the open sea. Validation using in situ observations showed that C2RCC produced relatively good estimates for chlorophyll-a concentrations of up to 12 µg/cm3 (RMSE=1.71 µg/cm3). No conclusion was drawn for TSM due to insufficient data.

Products and outputs for water quality monitoring are made available in the IM4ManilaBay MapABLE Web Portal, a web-based GIS platform that serves as a repository of processed maps for Manila Bay and linked systems. The Portal is also linked with the MASDAN mobile application that has been developed to allow the public to report different environmental issues. The web development involved the utilization of geospatial content management system, spatial database system, operating systems, programming, and style language. Services included frontend and backend development. The Web Portal serves as a repository for all outputs derived for water quality monitoring. It contains processed results, technical overview on the developed models and tools, historical data viewing and use of basic geoprocessing tools, and viewing of reported incidents from the MASDAN app. It is expected that the Web Portal will contribute significantly in the efforts of communities and environmental agencies to implement policies for sustainable development and management of Manila Bay and its watershed.

This paper presents the development of MASDAN, an integrated platform consisting of a mobile application and a web-GIS interface for monitoring the Manila Bay environment. The MASDAN mobile application uses citizen science and Volunteered Geographic Information (VGI) to collect data on specific environmental issues. It also serves as an information, education, and communication (IEC) tool for its users. The MASDAN Monitoring Interface is a centralized citizen report assessment tool for local authorities and environmental agencies to validate and monitor environmental reports within their area of jurisdiction. These two components create a two-way flow of information between citizens and authorities, ensuring prompt and information-driven actions to environmental issues affecting the Manila Bay and linked environments.

Indoor localization has gained considerable attention over the past decade because of the emergence of numerous location-aware services. Research works have been proposed on solving this problem by using wireless networks. Nevertheless, there is still much room for improvement in the quality of the proposed classification models. In the last years, the emergence of Visible Light Communication (VLC) brings a brand new approach to high quality indoor positioning. Among its advantages, this new technology is immune to electromagnetic interference and has the advantage of having a smaller variance of received signal power compared to RF based technologies. In this paper, a performance analysis of seventeen machine leaning classifiers for indoor localization in VLC networks is carried out. The analysis is accomplished in terms of accuracy, average distance error, computational cost, training size, precision and recall measurements. Results show that most of classifiers harvest an accuracy above 90 %. The best tested classifier yielded a 99.0 % accuracy, with an average error distance of 0.3 centimetres.