Discharge is an important proxy to quantify the amount of freshwater and the related hazards, such as droughts and floods. However the observation of this variable remains complicated in poorly or ungauged basins. In the last two decades many studies have focused on indirect approach to estimate the discharge from satellite observations. Our study lies in this context and aims at providing fine-scale estimations of the discharge in the Middle Niger Basin.
We present here the two main insights of our study. First a novel method has been developed that estimates the cross-sections geometry by combining altimetric data from the Sentinel 3 SRAL altimeter and estimations of widths from water masks derived from Sentinel 2 Level 2 images. These estimations are computed at the location where every Sentinel 3 tracks crosses the river, the so-called virtual stations. These estimations provide the “observable” part of the cross-sections geometry which are above the lowest water level. Then a simple steady-state hydraulic model is used to estimate the cross-sectional area above this lowest water level. Finally, dedicated interpolation techniques are used to fill the gaps between these virtual stations to obtain a representative 1D model of the river geometry (mesh).
Then an hydrology-hydraulics model is used to estimate the discharge. First, the hydrology model MGB (Modelo de Grande Bacias [1]) with calibrated parameters provides a prior estimation of the discharge along the river and for the main tributaries. Then the discharges are corrected by assimilating the SRAL altimetric data into the 1D shallow water model DassFlow-1D [2] using 4D variationnal data assimilation thecnics and the mesh previously computed. Finally the accuracy of the estimated discharge is assessed at some in-situ stations where data is available.
The results obtained are encouraging and show that this new methods can provide more accurate estimations of the discharge when compared to the estimations computed using the hydrology model only. Some preliminary results on an experiment where the discharges of the tributaries are also corrected by the assimilation will also be presented.
References:
[1] PONTES, P. ; FAN, F. M. ; FLEISCHMANN, AYAN SANTOS ; PAIVA, R. ; BUARQUE, D. C. ; SIQUEIRA, V. A. ; JARDIM, P. ; SORRIBAS, M. ; COLLISCHONN, WALTER . MGB-IPH model for hydrological and hydraulic simulation of large floodplain river systems coupled with open source GIS. Environmental Modelling & Software, v. 94, p. 1-20, 2017
[2] K. Larnier, J. Monnier, P.-A. Garambois, J. Verley. "River discharge and bathymetry estimations from SWOT altimetry measurements". Inverse Problems Sc. Eng. (IPSE), 29(6), pp 759-789, 2020
Environmental conditions, including rainfall variability, frequency of droughts and floods, as well as land degradation, are known to influence human migration patterns in West Africa. The most recent IPCC report predicts an increase in extreme weather events in the region, leading to an increase in flooding, as well as an increase in drying and agricultural droughts. To cope with adverse climatic conditions, poverty and food security, migration has been a strategy for centuries in this region.
Within our work in the project MIGRAWARE, we aim to identify regions that favor migration decisions (areas of origin) as well as regions that attract people based on certain positive attributes (destination areas). In addition to environmental factors, we characterize these regions also by economic, demographic, social, and political factors that are relevant to migration. Thus, we define regions where migration could take place based on geospatial data.
We use historical rainfall indices based on Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) data and historical temperature indices by means of the ERA5 dataset both covering the time period from 1981 to 2020. Moreover, the Normalized Difference Vegetation Index (NDVI) from 2002 to 2020 based on the MODIS NDVI dataset is used to examine spatial and temporal changes in vegetation for providing conclusive information on food security. To quantify temporal trends, time series analyses are performed using Mann-Kendall tests.
To account for the complexity of the phenomena, environmental factors are integrated into the network of factors and processes of migration. Additionally, to investigate the interaction between the different drivers of migration, data on armed conflict and population growth are also involved in the analysis. Subsequently, on-site surveys will provide information on social networks and land tenure issues. Furthermore, the analyzed geospatial data will be validated by obtaining information on individuals' perceptions of environmental change. This multi-method approach, combining spatial data and qualitative interviews, is expected to provide valid information on current migration patterns and a better understanding of the linkages between food security, climate change and migration.
Our work is conducive to achieving the overall aim of the MIGRAWARE project (WASCAL Wrap 2.0), to compile information on rural-urban and cross-border migration in West Africa. In this context, the findings of this study will help formulate recommendations for stakeholders to better target climate change adaptation measures and develop governance tools and policy briefs that will be tailored to the local, national, and intergovernmental level. Concurrently, this study contributes to the development of strategies and solutions within the food-climate-migration nexus in the region.
Coastal cities and Small Islands DevelopingStates are perpetually at the edge of high-impact environmental hazards such as coastal flooding, storm surges, coastal erosion and hurricanes. In the West African coasts, these hazards are exacerbated by the additional constraints of the high vulnerability and inadequacy of adaptation capabilities of the local population, institutional tools and infrastructures for improving resilience and accessing relevant, accurate and in-time data. These are the realities in West African coasts that the Enhancing Adaptation and Resilience against Coastal Multi-hazards along the West African Coasts (EARWAC) project posits to address by adopting a mixed-method approach that integrates ecological, socioeconomic and spatial data. This approach enabled the development of a co-design and analytic dashboard that helps enhance adaptation, resilience and response coordination for coastal multi-hazards along West Africa’s coasts. The project was implemented in three steps: i. conceptualisation and needs assessments, ii. user-assessment, and iii. need-based geospatial analytics dashboard development. In the first phase of the EARWAC project, desk research and literature review were conducted to understand the contextual underpinnings of coastal hazards in West Africa, the strengths and weaknesses of existing projects and initiatives on coastal hazards in West Africa. The need and users assessment was conducted through a desktop review and participatory survey process with community members (667 participants from 24 districts), 26 subject matter experts from the academia and regulatory bodies. The need assessment revealed no existing coastal multi-hazards dashboards in the region and that existing initiatives are devoid of stakeholders inputs. Besides 80% of subject matter experts agreeing that coastal flooding is endemic in the region, 41% of community members surveyed highlighted coastal flooding as the most critical coastal hazard in the West African region, followed by marine pollution (37%) and coastal erosion (28%). This outcome guided the development of the EARWAC dashboard (https://earwac.com/.)to prioritise coastal flooding and seasonal flash flood events in the region up to 2023. In the lead up to the dashboard development, we analysed and computed the Coastal Flood Vulnerability Index -(CFVI) -from 2020 to 2022, the historic flash flood (2015 and 2022) and a forecast of the seasonal flash flood (from 2022 till 2023) all disaggregated to the state level. Currently, the EARWAC dashboard is a first of its kind one-stop-shop infrastructure in the region that provides necessary information on coastal vulnerability and seasonal flood events to local dwellers, first responders, governmental institutions and SMEs. Upon resources availability and support, the dashboard has the capacity to function as a multi-hazard infrastructure, including hazards such as erosion, marine pollution, sea-level rise, etc.
The Niger Delta region of Nigeria has been battling with an unprecedented number of oil spills since the discovery of oil in 1950. As a recommendation given by the International Union for Conservation of Nature (IUCN–Niger Delta Panel), in response to its oil spill, there is the need to identify key sensitive and pristine areas that would need special protection in the event of their being threatened by oil spills. Hence this study seeks to explore why these oil spills happen, where they occur, and which areas are most affected.
From the result, the majority of the spills occurred within 1 km distance from the pipelines with the level of accessibility of oil facilities influencing their chances of being sabotaged. From the Chi-square analysis, a p-value of 0.001 implied that there is a statistically significant relationship between the accessibility of an oil facility and its chances of being sabotaged. Oil spill cold and hotspot regions within the Niger Delta were identified, with protected areas being affected by the oil spills. This raises the question: How protected are the protected areas in the Niger Delta? Especially the already threatened biodiversity, faced with extinction due to the rampant bush hunting and fragmentation from logging.
With knowledge of where the oil spills occurred, the study assesses the potential and significance of remote sensing vegetation indices in assessing the environmental impact of oil spills on land cover. A pixel-based classification using Random Forest classifier in Google Earth Engine was performed to obtain the land use land cover information which was assigned to the 2704 oil spill incidence data obtained from the National Oil Spill Detection and Response Agency (NOSDRA). Oil spills within vegetated cover areas were analysed in comparison to the selected non-spill sites using boxplot. Five vegetation indices: Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Soil-Adjusted Vegetation Index (SAVI), Normalized Green Red Difference Index (NGRDI) and Green Leaf Index (GLI) were used to measure the quantity and vigour of plants within the selected oil-polluted and non-polluted vegetated areas. All five indices were statistically significant at p-value ≤ 0.0001. However, indices that incorporated the near-infrared band (NIR) in their calculation performed better, as they were able to differentiate spill and non-spill sites across the different vegetation cover areas.
Wood charcoal is the main cooking fuel across rural and urban sub-Saharan Africa (SSA) and is a major source of forest degradation in the region. Charcoal production is either directly driven by domestic demand for fuel or can, as in Somalia, be linked to illegal exports in support of war regimes. The monitoring of forest degradation associated with charcoal production is often addressed as an indirect approach by detecting charcoal production sites on satellites imagery. Recent advances in computer vision could improve operational methods to map and monitor such charcoal production. This should lead to better estimates of forest degradation and associated carbon emissions.
In Somalia specifically, the FAO-SWALIM team – operating under the UN Joint Programme for Sustainable Charcoal Reduction and Alternative Livelihoods (PROSCAL) – is monitoring production of charcoal by identifying kilns where the wood is piled up and heated under low oxygen conditions to identify the areas where charcoal related deforestation takes place. The kilns' burn scars have a distinct round shape often with a lighter colored center and a diameter of 3 up to 25 m in diameter.
The FAO-SWALIM team has created an expert-based kiln delineation on Very High Resolution (VHR) imagery that contains more than 69,500 objects for the years 2018-2019 over an area of about 37,700 km2 in southern Somalia.
Taking advantage of this unique dataset we design an AI approach to automatically identify kilns related to illegal charcoal production in Somalia on VHR images. The methodology relies on a subset of this large dataset of expert-labelled data and a collection of VHR for the years 2018 and 2019. We use Worldview imagery, both panchromatic and RGB, obtained through the DigitalGlobe NextView license. Expert delineations are paired with available VHR tiles.
A CNN (convolution neural network) model for object detection is used (Faster R-CNN with pre-trained COCO weights) to recognize charcoal kilns on both RGB and panchromatic VHR. A CNN model for each type of VHR is trained separately.
We followed a two-stage methodology to decrease the time spent on verification of already delineated kilns and any additional delineation by hand. Indeed, due to inconsistencies, the expert delineation had to undergo an additional visual check so that it could be used to train the CNN model. This visual validation is done with the CVAT tool. During the first stage, the model is trained on a small number of validated pairs of images and delineated kilns. This step includes a hyperparameter tuning of the model to learn the best hyperparameters for this problem. Once the best hyperparameters are found, we inferred more images to increase the subset selected to a maximum of 300 for panchromatic and RGB and validated them with CVAT. In the second stage, we used this new dataset (300 pairs for panchromatic and 300 pairs for RGB) split into train, test and validation sets to measure the performance of the method on panchromatic and RGB detection.
The training set is used to fine-tune the models defined at the previous stage. Some data augmentations are also done to train the model for a better generalization of all chromatic conditions of the imagery (contrast and brightness change and image flipping). The best model for each workflow (panchromatic and RGB) related to performance metrics is selected and used to infer all images available and all the objects detected are post-processed as a vector file.
The performance metrics computed on the test sets show very promising results with an F-1 score of 0.95 (precision: 0.98, recall: 0.93) for panchromatic images and 0.90 for RGB (precision: 0.93, recall: 0.91).
With this methodology we show how neural networks can generalize on the detection of kilns with a small number of images visually delineated and validated. This will allow local experts to focus only on the validation of the outputs of the net instead of the current time-consuming workflow that consists of visually searching and manually delineating kilns across large areas. The methodology developed here will be shared through open Jupyter notebooks to facilitate appropriation and reutilization.
Surface features like fumaroles (steaming ground), hot springs, mud pots or hydrothermal alteration are direct evidence for the geothermal activity within an area. Different remote sensing methods offer a range of application possibilities to map these surface manifestations in order to define areas of enhanced geothermal potential in early exploration phase and to contribute to the characterization of the geothermal reservoir.
The Djiboutian Office for Geothermal Development (ODDEG: Office Djiboutien de Développement de l’Énergie Géothermique) defined 22 sites in the country of Djibouti for further exploration in the near future in order to accomplish the long-term aim to cover a considerable amount of electricity production by renewable resources, and in this term to extend the geothermal sector.
One of these sites is the area North Ghoubbet in central Djibouti, which is thought to be a promising site for geothermal exploration due to the tectonic active Asal-Ghoubbet rift in the vicinity. A high topographic relief and difficult-to-access terrain limit the methods for early exploration. Different remote sensing methods are applied to promote the investigation in this exploration phase. Freely available multispectral Sentinel-2 and SRTM digital elevation data have been used to perform a structural analysis, which is of major importance as geological structures increase the permeability of the rock for hydrothermal fluids. Areas with a high density of structures and/or a high number of fault intersections are therefore areas with an enhanced geothermal potential. In addition, high-resolution multispectral Worldview-3 VNIR and SWIR data have been used for the mapping of alteration zones, where host rock is interacting with hydrothermal fluids, and whose locations therefore indicate fluid flow and enhanced potential as well. Band ratioing, Spectral Angle Mapping and Principal Component Analysis have been used to map the alteration zones that particularly are characterized by clay minerals, iron bearing minerals and carbonates.
Both methods lead to the identification of locations with an enhanced geothermal potential in the same area, where a large NW-SE striking fault zone seems to facilitate fluid migration as this area correlates with alteration zones. Fumaroles and one hot spring are the evidence for geothermal activity in this area.