Low Data Availability Research Articles

West African reservoirs and their fisheries: An assessment of harvest potential

A major constraint to science-based fish stock management in West Africa is the lack of reliable data on target stocks. This especially holds true for inland fisheries, such as those that operate in reservoirs. Due to the low availability of resources and population data, and the limited number of fisheries experts in the region, state institutions and investigators rely heavily on simple catch statistics and empirical models for their estimations of fish production and potential yields. This paper reviews data from the FAO, and published articles and reports on West African reservoirs, with special reference to their morphometric and environmental features in relation to fish catch. In addition, we analyse primary data on three focus reservoirs. First, to improve and update available models of potential harvests from reservoirs, we regress fish catch data against reservoir surface area data for 30 reservoirs in West Africa, yielding the following equation: Catch (tonnes/year) = 17.3 × Area (km2)0.8626. The equation accounts for 95.7% of the variation observed in the fish catches. Analysis of covariance of small (<2 km2) and large (>2 km2) reservoirs shows no significant difference (F = 0.5895, p = 0.45) in the slopes of the two groups. Second, we apply multiple regressions to a sub dataset of 15 reservoirs with surface area and mean depth as predictors; and we also explore reservoir age as a further variable. We find that fisheries productivity is inversely correlated with both mean depth (r = -0.49) and surface area (r = -0.32), but there is no significant correlation found with reservoir age (r = 0.03).

NASA/POWER and DailyGridded weather datasets-how good they are for estimating maize yields in Brazil?

The low availability of high-quality meteorological data resulted in the development of synthetic meteorological data generated by satellite or data interpolation, which are available in grids with varying spatio-temporal resolution. Among these different data sources, NASA/POWER and DailyGridded databases have been applied for crop yield simulations. The objective of this study was to evaluate the performance of these two datasets, in different time scales (daily, 10-day, monthly, and annual), as input data for estimating potential (YP) and attainable (YA) maize yields, using the FAO Agroecological Zone crop simulation model (FAO-AEZ), properly calibrated and validated. For that, daily weather data from ten Brazilian locations were collected and compared to the data extracted from NASA/POWER and DailyGridded systems and later applied to estimate the potential and attainable maize yields. DailyGridded data showed a better performance than NASA/POWER for all weather variables and time scales, with confidence index (C) ranging from 0.52 to 0.99 for the former and from 0.09 and 0.99 for the latter. As a consequence of that, DailyGridded data was better than NASA/POWER to estimate maize yields with estimates close to those obtained with observed data, with a lower mean absolute errors (< 30 kg ha-1) and a higher confidence index (C = 0.99).

Testing approaches to determine relative stock abundance priors when setting catch recommendations using data-limited methods

Data-limited methods for managing stocks have expanded greatly over the last decade due to the necessity of quantitatively assessing exploited populations with limited information. A special category of such approaches is based on stock reduction analysis. These “catch-only” methods provide a way to handle low data availability, but also require as an input relative stock status (e.g., current biomass/initial biomass), a difficult to determine value that leads to large sensitivity in method output and performance. Published methods have been developed to devise informative priors for this quantity, but have not been evaluated together with the assessment methods. Here, relative stock abundance priors derived from elicited expert knowledge, vulnerability analysis and catch trends are compared to the common assumption of a stock being at B40% (40% of the initial biomass). The performance of each prior source is evaluated both in the degree of bias in estimating stock status and in the estimation procedure of catches for ten data-rich stocks with six stock assessment models that require stock abundance input. The results from both performance metrics show that these alternative sources can provide more accurate priors than assuming current biomass equals B40%, with priors elicited from stock assessment experts performing best. Finally, based on the findings of this work and the data requirements to construct a stock abundance prior, we make recommendations on how to navigate the options for devising a relative stock status prior.

Intersectoral and integrated approaches in achieving the right to health for refugees on resettlement: a scoping review

BackgroundBetter understanding, documentation and evaluation of different refugee health interventions and their means of health system integration and intersectoral collaboration are needed.ObjectivesExplore the barriers and facilitators to the integration of...

Comparative analysis of the accuracy of surface soil moisture estimation from the C- and L-bands

Surface soil moisture (SSM) estimation is of great importance in several areas, such as hydrology, agriculture and risk assessment. C-band SAR (synthetic aperture radar) data have been widely used to estimate SSM, whereas few studies have been performed using L-band SAR due to the low availability of L-band SAR data. In this context, the objective of the present paper is to compare the SSM estimation potentials of the C- (Sentinel-1) and L-bands (PALSAR) for wheat and grassland plots. The inversion approach developed in this study uses neural networks to invert the SAR signal and estimate the SSM. For each radar frequency, the developed neural networks were trained using the following as an input vector: SAR incidence angle, SAR polarization (VV for the C-band and HH for the L-band), and NDVI from optical images. Artificial Neural networks (ANNs) were developed and validated using synthetic and real databases. The results showed that the L-band provided slightly less accurate SSM estimates than the C-band. Moreover, the results showed that the accuracies of the SSM estimates for both frequencies strongly depended on the soil roughness (Hrms) and SSM values. From the synthetic database at SSM values less than 25 vol.%, the ANNs underestimated the SSM for Hrms values less than 1.5 cm and overestimated the SSM for Hrms values greater than 1.5 cm. In addition, the ANNs underestimated the SSM value regardless of the Hrms value when the SSM value was greater than 25 vol.%. An RMSE analysis of the SSM estimates showed that the highest RMSE values were observed for the L-band regardless of the SSM value, and high RMSE values were observed for the C-band only in very wet soil conditions (SSM>25 vol.%). From the real database at NDVI values less than 0.7, the RMSE (root mean square error) of the SSM estimates was 4.6 vol.% for the C-band and 5.3 vol.% for the L-band. Most importantly, the L-band enabled the estimation of the SSM under a well-developed vegetation cover (NDVI > 0.7) with an RMSE of 6.7 vol.%, whereas the C-band SAR signal became completely attenuated for some crops when the NDVI value was greater than 0.7, and thus the estimation of SSM was impossible using the C-band.

Current status and perspectives on "watch and wait" strategy for rectal cancer patients after clinical complete response following neoadjuvant chemoradiation

Neoadjuvant chemoradiation has been accepted as a standard of care for local advanced middle to low rectal cancer. Patients with clinical complete response (cCR) or near cCR following neoadjuvant chemoradiation may benefit from watch and wait strategy or organ-preserving surgery with good short- and long-term outcome and quality of life (QOL). Yet the criteria of cCR varies and cCR is not consistent with pCR. Therefore, the obstacle to the strategy lies on whether its failure can be salvaged and the complexity of follow-up. Available studies demonstrated that local recurrence or regrowth can be salvaged by surgery without compromising the survival. So, the key is appropriate follow-up schedule and timely salvage. The strategy has not drawn much attention until recently, and relevant studies go slowly because of low data availability, patient awareness, and peer acceptance. We still believe that more and more patients might benefit from this strategy, along with the increasing attention of QOL from the patients. That may be obtained through screening of the right patients and optimizing treatment modality, evaluation methods, and protocol of follow-up.

Acute toxicity of four metals to three tropical aquatic invertebrates: The dragonfly Tramea cophysa and the ostracods Chlamydotheca sp. and Strandesia trispinosa

The relatively low availability of toxicity data for indigenous tropical species has often been discussed. In addition, several taxonomic groups of invertebrates are understudied, such as dragonflies and ostracods. The aim of the present study was therefore to evaluate the acute toxicity of four metals (cadmium - Cd, copper - Cu, manganese - Mn, and mercury - Hg) to the tropical dragonfly nymphs of Tramea cophysa and two tropical ostracod species (Chlamydotheca sp. and Strandesia trispinosa). Toxicity data for other invertebrates were also mined to allow comparing the sensitivity of the three test species with that of other (temperate and tropical) invertebrates. The order of metal sensitivity was different for the three test species: T. cophysa: Cu > CdHg > Mn, Chlamydotheca sp.: Cd > Cu > Hg > Mn, and S. trispinosa: Cd > Hg > Cu > Mn. However, manganese was the least toxic metal tested for all three species, which is hypothesized to be due to a possible metal transfer to the cuticle of the moulting test species. The sensitivity ranking of the three test species to the metals was S. trispinosa > Chlamydotheca sp.>T. cophysa (except for Cu for which the ranking was Chlamydotheca sp.>T. cophysa > S. trispinosa). Overall, the test species are concluded to be suitable test organisms for tropical toxicity evaluations. Future studies should also evaluate the chronic toxicity and include other important metal exposure routes such as sediment and food.

Turkey OpenStreetMap Dataset - Spatial Analysis of Development and Growth Proxies

Abstract Number of studies covering major data aspects of OpenStreetMap (OSM) for developed cities and countries are available in scientific literature. However, this is not the case for developing ones mainly because of low data availability in OSM. This study presents a time-series spatial analysis of Turkey OSM dataset, a developing country, between the year 2007 and 2015 to understand how the dataset has developed with time and space. Five different socio-economic factors of the region are tested to find their relationship, if any, with dataset growth. An east-west spatial trend in data density is observed within the country. Population Density and Literacy Level of the region are found be the factors controlling it. It has also been observed that the street network of the region has followed the Exploration and Densification evolutionary model. High participation inequality is found within the OSM mappers, with only 5 of them responsible for the country’s 50% geo-data upload. Furthermore, it is found that these mappers use other Volunteered Geographic Information (VGI) and government open-dataset to feed into OSM. This study is believed to bring some high level insights of OSM for a developing country which would be useful for geographers, open-data policy makers, VGI projects planners and data-curators to structure and deploy similar future projects.

Evolution of rotavirus C in humans and several domestic animal species.

Rotavirus C (RVC) causes enteric disease in multiple species, including humans, swine, bovines, and canines. To date, the evolutionary relationships of RVC populations circulating in different host species are poorly understood, owing to the low availability of genetic sequence data. To address this gap, we sequenced 45 RVC complete genomes from swine samples collected in the United States and Mexico. A phylogenetic analysis of each genome segment indicates that RVC populations have been evolving independently in human, swine, canine, and bovine hosts for at least the last century, with inter-species transmission events occurring deep in the phylogenetic tree, and none in the last 100years. Bovine and canine RVC populations clustered together nine of the 11 gene segments, indicating a shared common ancestor centuries ago. The evolutionary relationships of RVC in humans and swine were more complex, due to the extensive genetic diversity and multiple RVC clades identified in pigs, which were not structured geographically. Topological differences between trees inferred for different genome segments occurred frequently, including at nodes deep in the tree, indicating that RVC's evolutionary history includes multiple reassortment events that occurred a long time ago. Overall, we find that RVC is evolving within host-defined lineages, but the evolutionary history of RVC is more complex than previously recognized due to the high genetic diversity of RVC in swine, with a common ancestor dating back centuries. Pigs may act as a reservoir host for RVC, and a source of the lineages identified in other species, including humans, but additional sequencing is needed to understand the full diversity of this understudied pathogen across multiple host species.

Recent Trends in the Modeling and Quantification of Non-probabilistic Uncertainty

This paper gives an overview of recent advances in the field of non-probabilistic uncertainty quantification. Both techniques for the forward propagation and inverse quantification of interval and fuzzy uncertainty are discussed. Also the modeling of spatial uncertainty in an interval and fuzzy context is discussed. An in depth discussion of a recently introduced method for the inverse quantification of spatial interval uncertainty is provided and its performance is illustrated using a case studies taken from literature. It is shown that the method enables an accurate quantification of spatial uncertainty under very low data availability and with a very limited amount of assumptions on the underlying uncertainty. Finally, also a conceptual comparison with the class of Bayesian methods for uncertainty quantification is provided.

Características e procedência da lenha usada na cocção no Brasil

RESUMO A lenha vem sendo substituída ao longo dos anos por novas formas de energia, como o gás liquefeito de petróleo (GLP) e a eletricidade. Porém, esse combustível ainda representa uma fração significativa da matriz energética brasileira, muito embora as pesquisas referentes ao seu uso ainda sejam limitadas e regionais. O objetivo deste estudo foi realizar um levantamento da produção, consumo e características da lenha usada com a finalidade de cocção. De acordo com os resultados, em 2016, 26,5% da lenha produzida no Brasil foram empregados com fins residenciais, ou seja, em torno de 2x107 toneladas. A lenha é proveniente tanto da silvicultura, sendo o Paraná o maior produtor, quanto do extrativismo, sendo a Bahia a maior produtora. O estado que apresenta maior dependência desse combustível para a cocção é o Pará, enquanto o Rio de Janeiro, praticamente, não a usa para esse fim. O consumo per capita varia muito de uma região a outra do país. Com base nos dados de 2016, foi estimado um consumo de 1,7 kg/pessoa/dia. No entanto, estudos in loco mostraram variação de 0,7 a 8,5 kg/pessoa/dia. O levantamento aponta que uma parte considerável da lenha é proveniente de matas nativas que têm reflexos no desmatamento de uma dada região. A pouca disponibilidade de dados e a variação entre eles impede uma avaliação mais precisa do uso deste combustível no país e suas implicações para a saúde das florestas. A implantação de políticas públicas deveria ser prioridade do governo, pois o uso da lenha de forma inadequada afeta o manejo da vegetação nativa, além de ser uma questão socioambiental, econômica e de saúde pública.

Research on improved privacy publishing algorithm based on set cover

With the invention of big data era, data releasing is becoming a hot topic in database community. Meanwhile, data privacy also raises the attention of users. As far as the privacy protection models that have been proposed, the differential privacy model is widely utilized because of its many advantages over other models. However, for the private releasing of multi-dimensional data sets, the existing algorithms are publishing data usually with low availability. The reason is that the noise in the released data is rapidly grown as the increasing of the dimensions. In view of this issue, we propose algorithms based on regular and irregular marginal tables of frequent item sets to protect privacy and promote availability. The main idea is to reduce the dimension of the data set, and to achieve differential privacy protection with Laplace noise. First, we propose a marginal table cover algorithm based on frequent items by considering the effectiveness of query cover combination, and then obtain a regular marginal table cover set with smaller size but higher data availability. Then, a differential privacy model with irregular marginal table is proposed in the application scenario with low data availability and high cover rate. Next, we obtain the approximate optimal marginal table cover algorithm by our analysis to get the query cover set which satisfies the multi-level query policy constraint. Thus, the balance between privacy protection and data availability is achieved. Finally, extensive experiments have been done on synthetic and real databases, demonstrating that the proposed method preforms better than state-of-the-art methods in most cases.

Soft tissue sarcomas management guidelines

Soft tissue sarcomas (STS) are group of rare tumors arising from mesenchymal tissues. It can develop in soft tissues like skeletal muscle, fibrous tissue, fat, nerves, blood vessels, lymphatic vessels, and the deep layers of the skin. The disease can arise in any part of the body, but mostly found in arms and legs, and are also found in internal organs and abdominal cavity. However, soft tissue sarcomas are still uncommon even with more than hundred different types of sarcomas of soft tissues. The disease comprises less than one percent of adult malignancies and twelve percent of pediatric malignancies. As a rare disease with heterogeneity of its subtypes, there is low availability of data to recommend evidencebased management guidelines. STS management guidelines are recent recommendations for surgeons for better diagnosis and treatment of different types of sarcomas, aiming at improving quality of care in patient, and giving them the proper decision for management. This review will summarize clinical presentation, diagnosis, investigation, staging and multidisciplinary treatment of different subtypes of STS based on most recent published guidelines.

A generic pixel-to-point comparison for simulated large-scale ecosystem properties and ground-based observations: an example from the Amazon region

Abstract. Comparing model output and observed data is an important step for assessing model performance and quality of simulation results. However, such comparisons are often hampered by differences in spatial scales between local point observations and large-scale simulations of grid cells or pixels. In this study, we propose a generic approach for a pixel-to-point comparison and provide statistical measures accounting for the uncertainty resulting from landscape variability and measurement errors in ecosystem variables. The basic concept of our approach is to determine the statistical properties of small-scale (within-pixel) variability and observational errors, and to use this information to correct for their effect when large-scale area averages (pixel) are compared to small-scale point estimates. We demonstrate our approach by comparing simulated values of aboveground biomass, woody productivity (woody net primary productivity, NPP) and residence time of woody biomass from four dynamic global vegetation models (DGVMs) with measured inventory data from permanent plots in the Amazon rainforest, a region with the typical problem of low data availability, potential scale mismatch and thus high model uncertainty. We find that the DGVMs under- and overestimate aboveground biomass by 25 % and up to 60 %, respectively. Our comparison metrics provide a quantitative measure for model–data agreement and show moderate to good agreement with the region-wide spatial biomass pattern detected by plot observations. However, all four DGVMs overestimate woody productivity and underestimate residence time of woody biomass even when accounting for the large uncertainty range of the observational data. This is because DGVMs do not represent the relation between productivity and residence time of woody biomass correctly. Thus, the DGVMs may simulate the correct large-scale patterns of biomass but for the wrong reasons. We conclude that more information about the underlying processes driving biomass distribution are necessary to improve DGVMs. Our approach provides robust statistical measures for any pixel-to-point comparison, which is applicable for evaluation of models and remote-sensing products.

Multi-Objective Genetic Algorithm (MOGA) As a Feature Selecting Strategy in the Development of Ionic Liquids' Quantitative Toxicity-Toxicity Relationship Models.

Quantitative toxicity-toxicity relationship (QTTR) models have a great potential for improving the meaning of toxicological tests conducted on simple organisms. These models allow predicting the toxicological effect of a chemical based on its known toxicological effect in different toxicity tests, even against a different organism. This fact poses a great potential for predicting the toxicity of chemicals against higher organisms based on the results against lower ones. However, the possibility of developing such models is often restricted due to the low availability of data. We present a case study of developing the QTTR model for ionic liquids in different toxicological tests against the same species, in the face of insufficient experimental data (an additional confirmation for a different species is provided in the Supporting Information). In the presented case, we use a series of quantitative structure-activity relationship (QSAR) models developed to deliver the data concerning the toxicity of ionic liquids against human HeLa and MCF-7 cancer cell lines. We use these data to develop a QTTR model with an R2 as high as 0.8. The benefit of applying the multi-objective genetic algorithm (MOGA-a genetic algorithm allowing for selection of the best set of explanatory features for several different dependent variables at the same time) as a QSAR model feature selecting strategy is presented and discussed.

Functional emulation of high resolution tsunami modelling over Cascadia

The rarity of tsunamis impels the scientific community to rely on numerical simulation for planning and risk assessment purposes because of the low availability of actual data from historic events. Numerical models, also called simulators, typically produce time series of outputs. Due to the large computational cost of such simulators, statistical emulation is required to carry out uncertainty quantification tasks, as emulators efficiently approximate simulators. There is thus a need to create emulators that respect the nature of time series outputs. We introduce here a novel statistical emulation of the input-output dependence of these computer models. We employ the Outer Product Emulator with two enhancements. Functional registration and Functional Principal Components techniques improve the predictions of the emulator. Our phase registration method captures fine variations in amplitude. Smoothness in the time series of outputs is modelled, and we are thus able to select more representative, and more parsimonious, regression functions than a fixed basis method such as a Fourier basis. We apply this approach to the high resolution tsunami wave propagation and coastal inundation for the Cascadia region in the Pacific Northwest. The coseismic representation in this analysis is novel, and more realistic than in previous studies. With the help of the emulator, we can carry out sensitivity analysis of the maximum wave elevation with respect to the source characteristics, and we are able to propagate uncertainties from the source characteristics to wave heights in order to issue probabilistic statements about tsunami hazard for Cascadia.

Spatio-temporal downscaling of gridded crop model yield estimates based on machine learning

Global gridded crop models (GGCMs) are essential tools for estimating agricultural crop yields and externalities at large scales, typically at coarse spatial resolutions. Higher resolution estimates are required for robust agricultural assessments at regional and local scales, where the applicability of GGCMs is often limited by low data availability and high computational demand. An approach to bridge this gap is the application of meta-models trained on GGCM output data to covariates of high spatial resolution. In this study, we explore two machine learning approaches – extreme gradient boosting and random forests - to develop meta-models for the prediction of crop model outputs at fine spatial resolutions. Machine learning algorithms are trained on global scale maize simulations of a GGCM and exemplary applied to the extent of Mexico at a finer spatial resolution. Results show very high accuracy with R2>0.96 for predictions of maize yields as well as the hydrologic externalities evapotranspiration and crop available water with also low mean bias in all cases. While limited sets of covariates such as annual climate data alone provide satisfactory results already, a comprehensive set of predictors covering annual, growing season, and monthly climate data is required to obtain high performance in reproducing climate-driven inter-annual crop yield variability. The findings presented herein provide a first proof of concept that machine learning methods are highly suitable for building crop meta-models for spatio-temporal downscaling and indicate potential for further developments towards scalable crop model emulators.

Sensitivity analysis of the system of wind speed distributions

The fitting of empirical wind speed distributions is an important component in wind turbine energy yield assessment. Its accuracy depends on both the adequacy of the fitting function and the quality of the fitted wind speed data. Therefore, the system of wind speed distributions was recently introduced to further improve the statistical estimation of wind turbine energy yield. The system of wind speed distributions incorporates the three complement distributions Kappa, Wakeby and Burr-Generalized Extreme Value. In contrast to the progress that has been made in the statistical description of wind speed distributions, there is little research on the effects of poor quality data on the estimation of wind turbine energy yield. Therefore, the goal of this study was to evaluate the robustness of the system of wind speed distributions against typical wind speed data shortcomings such as (1) measurement errors, (2) missing data, and (3) low temporal resolution. The database were quality-controlled wind speed time series from 187 measurement stations of the German Meteorological Service located in Germany with a 10-minute temporal resolution. The investigated time series cover the period from July 2016 to December 2017. The original wind speed time series were modified by (1) rounding them to 0.5 m/s and 1 m/s steps, (2) reducing data availability to 1.0% and 0.5%, (3) reducing the temporal resolution. Afterwards, the original and the modified wind speed time series were fitted to the system of wind speed distributions. Ten goodness-of-fit metrics were applied for comparison of the original wind speed distributions with the modified distributions fitted to the system of wind speed distributions. Overall, it was found that the system of wind speed distributions is robust against common quality issues in the wind speed time series. Applying a 2.5 MW wind turbine power curve and fitting the system of wind speed distributions to integer wind speed data, the mean percentage annual energy yield error was estimated at 3.3%. Fitting the system of wind speed distributions to wind speed data with very low data availability of 0.5% of the original data, the annual energy yield over all stations was underestimated at −8.2%. The results of this investigation will help adequately addressing the biases caused by poor quality wind speed data, and thus, help to improve wind turbine energy yield assessment.

Delineation of individual deciduous trees in plantations with low-density LiDAR data

ABSTRACTDelineation of individual deciduous trees with Light Detection and Ranging (LiDAR) data has long been sought for accurate forest inventory in temperate forests. Previous attempts mainly focused on high-density LiDAR data to obtain reliable delineation results, which may have limited applications due to the high cost and low availability of such data. Here, the feasibility of individual deciduous tree delineation with low-density LiDAR data was examined using a point-density-based algorithm. First a high-resolution point density model (PDM) was developed from low-density LiDAR point cloud to locate individual trees through the horizontal spatial distribution of LiDAR points. Then, individual tree crowns and associated attributes were delineated with a 2D marker-controlled watershed segmentation. Additionally, the PDM-based approach was compared with a conventional canopy height model (CHM) based delineation. The results demonstrated that the PDM-based approach produced an 89% detection accuracy to identify deciduous trees in our study area. The tree attributes derived from the PDM-based algorithm explained 81% and 83% of tree height and crown width variations of forest stands, respectively. The conventional CHM-based tree attributes, on the other hand, could explain only 71% and 66% of tree height and crown width, respectively. Our results suggest that the application of the PDM-based individual tree identification in deciduous forests with low-density LiDAR data is feasible and has relatively high accuracy to predict tree height and crown width, which are highly desired in large-scale forest inventory and analysis.

Machine learning predictions of trophic status indicators and plankton dynamic in coastal lagoons

Multivariate trophic indices provide an efficient way to assess and classify the eutrophication level and ecological status of a given water body, but their computation requires the availability of experimental information on many parameters, including biological data, that might not always be available. Here we show that machine learning techniques – once trained against a full data set – can be used to infer plankton biomass information from chemical and physical parameter only, so that trophic index can then be computed without using additional biological data. More specifically, we reconstruct plankton information from chemical and physical data, and this information together with chemical data was used to compute the TRIX, which was eventually used to assess the eutrophication status of the water body. The RF was also used to evaluate the prevailing mechanism (bottom-up versus top-down) controlling plankton dynamic. The case study was a Mediterranean lagoon, the Ghar El Melh Lagoon, which has been used as a natural laboratory to test the effectiveness of the proposed approach. Based on the resulting TRIX values (4.2 in April – 5.7 in December) the Ghar El Melh Lagoon can be classified an eutrophic ecosystem. This modeling process suggests that phytoplankton growth in Ghar El Melh Lagoon is mainly bottom-up control by nutrients availability, whereas the top-down control exerted by the zooplankton is relatively weak. Results highlight that in bottom up controlled lagoon machine learning technique can efficiently be used to compute ecological indicators even with low availability of biological data.