Water Quality Monitoring Service (WQMS)

Date modified: March 2025

Overview

Background

The Water Quality Monitoring Service (WQMS) delivers water quality information for surface water bodies across the African continent. WQMS builds directly on the existing Waterbodies Monitoring Service which identifies and maps the locations and extents of surface water across Africa. WQMS provides a suite of water quality properties for the identified water bodies, enabling more informed understanding, management, and use of these critical water resources.

Earth Observation (EO) data offers significant advantages for water resources monitoring: it provides continuous spatial coverage, captures variations within a body of water, tracks changes through time, and does so at a fraction of the cost and logistical complexity of traditional in-situ monitoring. WQMS leverages these advantages by applying a range of published water quality algorithms to multi-sensor EO data processed through the Digital Earth Africa datacube infrastructure.

The service is designed to be multi-sensor and sensor-agnostic from the outset, drawing on multiple DE Africa data sources to generate the best possible estimates of each water quality parameter. These properties may be constant over time, vary on an annual basis, or exhibit intra-annual variability. The parameters for the services are outlined below.

Parameters

Parameter	Description
Turbidity	Indicates water clarity based on the concentration of suspended particles in the water column.Turbidity is provided as an indicator on a relative scale, with values designed to correspond to g/m3 of suspended particles (i.e., total suspended material)
Chlorophyll-A (ChlA)	ChlA is a pigment required for photosynthesis. High levels of ChlA are an indicator of algal content and often indicate poor water quality. ChlA is provided as an indicator on a relative scale, with values intended to correspond to mg/m3. ChlA is readily transformed to the Trophic State Index used in SDG reporting.
Optical Water Type (OWT)	OWT (Spyrakos et al 2017) is a classification of water types used to support EO based monitoring of water quality. For inland waters 13 OWTs are identified based on their spectral characteristics. These can then be characterised as oligotrophic, eutrophic, or hypereutrophic water types. The OWT of a waterbody may change seasonally or over the longer term.
Water Colour	The Hue of the water, derived from the spectral reflectances measured by the EO instruments. Hue is expressed as an angle from 0-360. The hue of the waterbody is an objective way to determine changes in water colour.
Water Temperature	Surface water temperature, a key variable influencing biological activity, stratification, and ecosystem dynamics.
Floating Algal Index (FAI)	FAI is used to detect and monitor floating vegetation and algal blooms, which are often problematic or hazardous. FAI supports early warning and risk management. Areas of algae and surface vegetation must also be mapped before applying algorithms for Chlorophyll-A or turbidity.
Trophic State Index (TSI)	Derived from ChlA, the TSI provides an indicator of lake trophic status and productivity. Values typically range from 0 to 100, with higher values reflecting greater algal biomass and increasing eutrophication. TSI is approximately the log of Chlorophyll-A in mg/m3, converted to classes.

The WQMS is developed to also support African governments in reporting on SDG 6.6.1. By providing continent-specific tools and datasets, the service ensures national control and transparency. Higher resolution data enables flexibility for local adaptations to meet regional monitoring needs.

Currently, the WQMS provides annual mapping that delivers a comprehensive snapshot of all water bodies, including trophic state classification and turbidity levels. This supports year-to-year trend tracking and long-term assessment. The annual products represent typical water quality conditions derived from a multi-spectral geomedian rather than individual satellite observations and therefore do not capture intra-year variability. To enable more detailed analysis, tools and notebook examples are provided to extend the workflow to monthly monitoring, allowing users to undertake seasonal or more frequent monitoring as required.

Specifications

The WQMS covers water bodies across the African continent. The Annual Mapping service covers 2000 to 2025, with yearly updates planned.

Relevant coverage and metadata for each of the WQMS variables can be viewed on the DE Africa Metadata Explorer.

Table 1: WQMS annual product specifications

Specification	Value
Product name	WQMS
Cell size - X (metres)	10
Cell size - Y (metres)	10
Coordinate reference system	EPSG:6933
Temporal resolution	Annual
Temporal range	2000 to 2025
Parent dataset	GeoMAD, WOfS, Waterbodies
Update frequency	Annual

Table 2: WQMS measurements

Band ID	Description	Value range	Data type	No data value
chla	Chlorophyll-A		float32	nan
tsi	Trophic State Index		float32	nan
tsm	Total Suspended Matter		float32	nan
agm_hue	Hue angle	0 - 360	float32	nan
agm_fai	Floating Algal Index	> 0.05	float32	nan
agm_ndvi	NDVI	0.05-1	float32	nan
agm_owt	Optical Water Type	1 - 13	float32	nan
tirs_st_ann_med	Median Surface Temperature		float32	nan
tirs_st_ann_max	Maximum Surface Temperature		float32	nan
tirs_st_ann_min	Minimum Surface Temperature		float32	nan

Figure 1: WQMS geographic extent

The Water Quality Monitoring Service covers the African continent where water has been consistently detected in the Water Observations from Space product. Some data gaps remain due to the availability issue of input surface temperature data; this issue is currently being addressed.

Related Services

• GeoMAD cloud-free composites

• Water Observations from Space

• Waterbodies Monitoring Service

License

CC BY Attribution 4.0 International License

Acknowledgments

The Digital Earth Africa Water Quality product is the result of collaborative contributions from multiple international Earth observation programmes, research organisations, and cloud infrastructure providers. We gratefully acknowledge the following organisations for making this work possible.

Provider	Data / Resource	Role in product
USGS / NASA	Landsat 5 TM, 7 ETM+, 8 OLI/TIRS, 9 OLI-2/TIRS-2	Surface reflectance and thermal imagery underpinning water quality retrievals
ESA / Copernicus	Sentinel-2 MSI (Level-2A)	High-resolution optical imagery for chlorophyll-a, TSM, turbidity and optical water type
Amazon Web Services (AWS)	Cloud compute and storage credits	Scalable cloud infrastructure enabling open, continent-wide data processing and access

---

WQMS Use and Interpretation

The first release of WQMS consists of pre-processed water quality data over all of Africa summarised for each waterbody. It can be used to identify changes in all waterbodies that are identified in the DE Africa waterbodies dataset. This includes over 700,000 individual waterbodies down to dams of 1-2 hectares. Measurements are based on annual geomedian datasets processed through a range of published water quality algorithms. Temperatures are sourced from thermal sensors on the Landsat satellites. Supporting notebooks demonstrate how to load and visualise data. Additional notebooks demonstrate the second phase of the WQMS, monthly monitoring. The material below uses both the WQMS and prototype notebooks that demonstrate the expected approach to monitoring capabilities.

Detecting Change

How has the water quality here changed through time?

The WQMS indicates how turbidity, trophic state and temperature are tracking through time. Sudden, or gradual and persistent, changes in these variables are likely to be due to actual changes that have occurred or that are occurring. This information might be used to trigger closer investigation of the water quality itself, or further studies to understand how the landscape or events in the weather or climate are affecting water quality (e.g., Zhao et al., 2024, Science of the Total Environment 912 (2024) 169152).

The figure below illustrates output from the annual WQMS, and a potential monthly monitoring service for a small dam in South Africa. The annual WQMS indicates changes in turbidity from 2017 onwards. These are supported by the more detailed monthly monitoring results. Together, these results indicate that, as well as the images appearing to look different:

• the EO instrument readings have changed (across three remote sensing platforms);

• the majority of turbidity algorithms, used across the satellite instruments, take those changes to indicate increased turbidity

To understand what led to these changes would require information on the management of the dam from in the field.

Figure 2: Annual WQMS output and monthly monitoring for a small dam in South Africa

Substantiating apparent change — are the changes that I can ‘see’ in the image backed up by the numbers?

When examining a specific waterbody, we can often see changes in the appearance of remote sensing images, but a visual appraisal will often not be enough to tell if the changes in appearance are real or significant. WQMS converts the images to multiple measures of water quality. If the numbers are changing, there is a good chance that the water quality is changing too, as seen in the examples below.

Figure 3: Statistical confirmation of change in a waterbody

Figure 4: This information can also be accessed directly from the DE Africa Map

Substantiating apparent change - are the changes that I can ‘see’ in the image backed up by the numbers?

When examining a specific waterbody, we can often see changes in the appearance of remote sensing images, but a visual appraisal will often not be enough to tell if the changes in appearance are real or significant. WQMS converts the images to multiple measures of water quality. If the numbers are changing, there is a good chance that the water quality is changing too, as seen in Figure 5 below.

Figure 5: Lake Manyara water regime change

Along with statistical information the WQMS captures spatial patterns which can be useful for visualisation, as in the following figure.

Figure 6: Spatial patterns in a waterbody

Comparing waterbodies?

WQMS can be used to compare water bodies. As well as direct comparisons, the database captures the characteristics of all 700,000 water features in the Waterbodies service. Since WQMS applies water quality algorithms across all water areas, the ‘heavy lifting’ has already been done, allowing stakeholders to move more quickly to the important questions and focus effort in areas of need.

Figure 7: A comparison of Lakes Abiata, Shalla and Langano

Limitations

The WQMS annual mapping provides a robust approach to detecting change. For the annual values to change, the typical condition of the waterbody must have changed between the two years. Furthermore, since the overall results for turbidity and chlorophyll are based on multiple measurements from all available satellites, the ensemble of measurements will need to ‘agree’, in order for a change to be indicated. In addition, the geomedian datasets provide a statistically robust, cloud-free dataset leading to more reliable estimates.

However these benefits have accompanying limitations. The annual products are based on a geomedian, which represents the typical state of the waterbody, rather than individual satellite observations. By definition, these geomedian datasets do not capture intra-year variability, and mask short-term changes and seasonal dynamics. For example, high levels of turbidity, chlorophyll or floating algae that are present for only a small part of the year or do not persist in a single location, will not be detected in the geomedian, which records the prevailing state of the water. Such conditions will only be detected in the annual products when they do become ‘typical’.

It follows that the annual values are not, in general, average values for the year, since average values will be influenced by short-term events which the geomedian removes. For example, the annual values can be expected to underestimate the proportion of a waterbody affected by floating vegetation, unless that vegetation becomes more permanent than not.

The WQMS will not replace in-situ measurements. It is impossible for remote observations to replicate physical samples taken directly from a waterbody (often at multiple depths) and this is not the purpose or objective of the WQMS. Instead, the WQMS brings a complementary and unique point of view, capturing entire areas of water systematically through time to provide new information, evidence and insights that are impossible with in-situ samples alone. The monthly monitoring capability, demonstrated in notebooks and anticipated as the next development of the WQMS, will capture more detailed variations in water quality and allow particular algorithms or combinations to be chosen. Where in-situ data are available these may be used to develop a local calibration for the WQMS monthly monitoring products. Combining in-situ and remote capabilities in this way is the ideal approach.

The reliability of WQMS depends somewhat on the quality and amount of data available. Before 2013 only Landsat-5 and Landsat-7 were available, and EO data capture was not continuous due to limitations of satellites, producing much less data. The quantity and quality of data improved markedly after 2016. WQMS results can be inconsistent because of these changes and differences in the DE Africa geomedians, although we have taken steps to minimise these effects.

Water quality measurements may also be affected by residual cloud contamination or artefacts in the input data.

---

Data Access

The Digital Earth Africa WOfS data can be accessed from the associated S3 bucket.

Table 3: AWS data access details

AWS S3 details
Bucket ARN	arn:aws:s3:::deafrica-services
Product name	wq_annual

The bucket is located in the region af-south-1 (Cape Town).

OGC Web Services (OWS)

This product is available through DE Africa’s OWS.

Open Data Cube (ODC)

The WQMS collection can be accessed through the Digital Earth Africa ODC API, which is available through the Digital Earth Africa Sandbox.

Waterbodies API

Summary statistics for each waterbody by accessing the Waterbody API in two ways:

1. Clicking on a waterbody after loading the waterbodies product in DE Africa Maps.

2. Accessing the get_water_quality_summary() and get_water_quality_rankings() functions in Python. This is demonstrated in the ‘Water Quality Summary’ section in the dataset introduction notebook.

---

Technical Information

Processing

Overview

The WQMS annual products use DE Africa Water Observations from Space (WOfS) to determine areas of water, DE Africa Geomedians to provide annual data, and DE Africa Waterbodies to provide polygon areas for each water body. Using these services as building blocks helps to ensure an efficient approach which can be maintained into the future.

Using reflectance data from the Geomedians, published algorithms are applied to estimate water properties - floating algae, water colour, optical water type, turbidity and Chlorophyll-A. As the Geomedians do not include temperature measurements, these are instead processed directly from the EO data (thermal bands / instruments on the Landsat satellites).

Data are first processed for all pixels considered to be water, and then summarised for each waterbody. The gridded data and the polygon summaries are then available through the DE Africa Open Data Cube and associated services.

Geomedians

Annual Geomedian images are statistically robust annual summary datasets. Unlike simple multivariate medians, every pixel in a Geomedian is a spectrally representative reflectance value derived from observed data, preserving inter-band relationships and supporting the valid application of EO algorithms.The Geomedians capture the prevailing state for the year, and changes in the geomedians from year to year are therefore likely to be due to actual change rather than noise. Conversely, when deriving water quality information from the Geomedian data we cannot capture variability within the year. DE Africa Geomedians do not include all bands (in particular Band 1 in the MSI and OLI sensors) limiting some of the algorithms that can be applied. Geomedians from TM /ETM data are not produced after 2012.

Identifying Areas of Water

All products are restricted to areas expected to be water during each year of data. DE Africa Water Observations from Space (WOfS) annual summary data is used to identify areas that are expected to be water during the year of observations that make up the Geomedian. Since floating vegetation can dominate the water surface for long periods of time, WOfS values are considered over a rolling 5-year period to identify water areas.

Water Temperature, T

Temperature (K) is estimated from Landsat data: the TIRS instrument on Landsat 8/9, and the thermal bands on the Landsat 5 and 7 TM and ETM instruments. These data include uncertainty and emissivity bands. Pixels are accepted where T > 273 K, temperature uncertainty is less than 5 K, and emissivity E > 0.95 (water has an emissivity of above 0.96 up to 14,000 nm; values below 0.95 are counter-indicative of water). Valid Landsat surface temperature observations acquired within each calendar year are summarised to derive per-pixel 10th, 50th (median), and 90th percentile values, providing a robust representation of the annual temperature range and typical conditions, while reducing sensitivity to extreme values.

Floating Algae Index (FAI)

FAI is calculated using the method described in Hu (2009). The index detects unexpectedly high reflectance in the near-infrared bands by comparison of the observed reflectance with the reflectance that would be expected based on the shortwave infrared and red reflectances.

Reference: Hu, C. (2009). A novel ocean color index to detect floating algae in the global oceans. Remote Sensing of Environment, 113, 2118–2129.

Values of FAI below a threshold are discarded. Values above the threshold are indicative of the presence of algal blooms or floating vegetation. The threshold applied for the WQMS geomedian datasets is 0.05.

FAI values from different sensors (ETM, OLI, MSI) differ slightly. To compensate for observed biases between values of FAI derived from different sensors, the following factors were applied:

Sensor	Factor
MSI	1.0
OLI	0.95566
ETM	1.00832

For years where geomedians are available from more than one of the TM, OLI and MSI sensors, FAI values from each geomedian are combined to a single FAI value using a mean, weighted according to the number of observations contributing to each geomedian.

Figure 8: Steps in the production of the turbidity (tss) and tropic state (chla) water quality indicators

Turbidity and Chlorophyll Water Quality Indicators

Water quality measures indicative of turbidity and trophic state are calculated after identifying water areas (see the diagram above) and after removal of pixels with high FAI values as previously described. Data from the visible wavelength bands are corrected for Rayleigh scattering using SWIR bands.

The following table lists the algorithms used, which sensors they are used with, and the name of the variable used in the internal workflow in each case. To give continuous coverage from 2000 to the present day, and to provide as many estimates of water properties as possible, we apply a variety of published algorithms to all available Geomedian data.

Algorithms applied and sensors used to estimate Turbidity and Chlorophyll levels

Algorithm	ETM/TM (2000–2012)	OLI (2013–present)	MSI (2017–present)
Chlorophyll-A algorithms
NDCI — Normalised Difference Chlorophyll Index, NIR-Red. Mishra and Mishra, Normalized difference chlorophyll index: A novel model for remote estimation of chlorophyll-a concentration in turbid productive waters. Remote Sensing of Environment, Volume 117, 15 February 2012, Pages 394-406 https://doi.org/10.1016/j.rse.2011.10.016	ndci_tm43	ndci_oli54	ndci_msi54, ndci_msi64, ndci_msi74
Toming — Toming et al., First Experiences in Mapping Lake Water Quality Parameters with Sentinel-2 MSI Imagery. Remote Sensing. 2016, 8, 640; doi:10.3390/rs8080640			chla_toming_msi
Gurlin 2B — Gurlin et al., Remote estimation of chl-a concentration in turbid productive waters — Return to a simple two-band NIR-red model? Remote Sensing of Environment 115 (2011) 3479–3490			gurlin2b_msi
3BDA — Three band DA. Buma and Lee, Il Evaluation of Sentinel-2 and Landsat 8 Images for Estimating Chlorophyll-a Concentrations in Lake Chad, Africa. Remote Sens. 2020, 12, 2437.	3bda_tm	3bda_oli	3bda_tsm
Tebbs — Tebbs, Remedios and Harper. Remote Sensing of Chlorophyll-a as a Measure of Cyanobacterial Biomass in Lake Bogoria, a Hypertrophic, Saline—Alkaline, Flamingo Lake, Using Landsat ETM+. Remote Sensing of Environment 2013, 135, 92–106.	chla_tebbs_tm	chla_tebbs_oli	chla_tebbs_msi
MERIS 2B — Gurlin et al., Remote estimation of chl-a concentration in turbid productive waters — Return to a simple two-band NIR-red model? Remote Sensing of Environment 115 (2011) 3479–3490			chla_meris2b_msi
MODIS 2B — Gurlin et al., Remote estimation of chl-a concentration in turbid productive waters — Return to a simple two-band NIR-red model? Remote Sensing of Environment 115 (2011) 3479–3490	chla_modis2b_tm		chla_modis2b_msi
TC-2 — Lui et al 2020. An OLCI-based algorithm for semi-empirically partitioning absorption coefficient and estimating chlorophyll-a concentration in various turbid case-2 waters. Remote Sensing of Environment 239 (2020) 111648, And Corrigendum with corrections, https://doi.org/10.1016/j.rse.2020.111726. Zhao et al., 2024. Dynamic monitoring and analysis of chlorophyll-a concentrations in global lakes using Sentinel-2 images in Google Earth Engine. Science of the Total Environment 912 (2024) 169152. TC-2 is a semi-analytical algorithm that uses multiple wavelengths of light to model absorption and scattering, before deriving Chl-A concentration. It is a more advanced method than alternatives. TC-2 is not implemented in the WQMS annual service due to the lack of MSI band 1 in the DE Africa Geomedian datasets, but will be implemented in monitoring notebooks and future monthly monitoring service developments.			tc2_chla_msi
Turbidity algorithms
ndssi_rg and ndssi_bnir — Normalised Difference Suspended Sediment Index, red:green, and blue:nir. Hossain et al., Development of Remote Sensing Based Index for Estimating/Mapping Suspended Sediment Concentration in River and Lake Environments. 2010. 8th International Symposium on ECOHYDRAULICS (ISE 2010) Paper No. 0435, pp. 578-585.	ndssi_rg_tm	ndssi_rg_oli, ndssi_bnir	ndssi_rg_msi
ti_yu — Yu et al. An empirical algorithm to seamlessly retrieve the concentration of suspended particulate matter from water color across ocean to turbid river mouths. 2019. Remote Sensing of Environment 235, 111491	ti_yu_tm	ti_yu_oli
tsm_lym — Lymburner et al., Landsat 8: Providing continuity and increased precision for measuring multi-decadal time series of total suspended matter. Remote Sensing of Environment Volume 185, November 2016, Pages 108-118	tsm_lym_tm	tsm_lym_oli	tsm_lym_msi
spm_qiu — Qiu. A simple optical model to estimate suspended particulate matter in Yellow River Estuary. Optics Express 2013 21(23):27891-27904 DOI: 10.1364/OE.21.027891	spm_qiu_tm		spm_qui_msi
tss green-red — Zhang et al 2021 Remote sensing: Landsat Image-Based Retrieval and Analysis of Spatiotemporal Variation of Total Suspended Solid Concentration in Jiaozhou Bay, China. To avoid numerical overflows with Zhang’s method as published, we implement this as an index: (G + R) / (G/R)	tss_gr_tm	tss_gr_oli	tss_gr_msi
tss green-red-blue — Zhang 2023 Remote sensing monitoring of total suspended solids concentration in Jiaozhou Bay based on multi-source data. Ecological Indicators 154 (2023) 110513. To avoid numerical overflows with Zhang’s method, this is implemented as an index: (G + R) / (B/R)	tss_grb_tm	tss_grb_oli	tss_grb_msi

Harmonisation Across Instruments

Where an algorithm is used with more than one sensor the resultant variables are harmonised before further processing. For example the normalised difference Chlorophyll Index can be calculated equally well using TM/ETM, OLI and MSI sensor data. In general, the values produced from the OLI and TM instruments are modified to be statistically consistent with the results from the MSI instrument. The process is based on a set of 30 waterbodies/sites with a geographic range of 15.75 W to 39.72 E, and from 34.1 S to 37 N, and a time interval of 1-1-2019 to 31-12-2020 (which ensures overlap of instruments). All available Landsat and Sentinel-2 images were used in the process.

Figure 9: Cross-sensor harmonisation process for water quality variables

Ensemble Values

Results from all turbidity algorithms, and from all ChlA algorithms, are considered as an ensemble. The median value of all measurements is taken as the best estimate of the water quality parameter for a given time. This approach recognises that no single algorithm will give perfect results and places value on having a range of estimates, from a range of algorithms that will perform differently in different conditions.

Re-scaling

Whilst some algorithms produce direct estimates of properties (e.g., mg/m³), others are simply numbers. To be able to compare ‘like to like’ we re-scale using values from all the algorithms for over 60 lakes across Africa. To re-scale the measures we assume that the median values returned by each algorithm are equivalent (i.e., that overall they are ‘seeing’ the same phenomenon), and we assume that the lowest values returned by each algorithm correspond to near-zero (since turbidity and ChlA measured as concentrations cannot go below zero). Re-scaled values from different algorithms are then comparable.

We scale the data to a preferred algorithm. For ChlA the Meris2b algorithm (which is applied using corresponding bands from the Sentinel-2 MSI sensor) shows a median value, over 30 diverse lakes, of 29. This is within the range of 21 ± 9 published by Zhao et al., 2024. For TSS we scale to the algorithm of Lymburner et al., 2016. This algorithm produces reasonable values over a wide range of lakes.

TM, OLI and MSI ‘Eras’

Because different instruments are available at different times, the combination of algorithms that can be applied varies through time. To ensure that WQMS results are consistent through time, values are adjusted slightly to ensure that the median values for water quality are consistent across the three eras. The availability of DE Africa Geomeidan data and median values for each era are summarised below:

	2000		2012	2013	2017	2025
TM era	median TSS  : 45.58 median ChlA : 29.05
OLI era			median TSS  : 35.25 median ChlA : 29.05
MSI era				median TSS  : 36.28 median ChlA : 27.80

Waterbody Attributes

DE Africa Waterbodies produces polygons of each water body across Africa. To describe each water body, gridded data are summarised to the DE Africa polygons (details to be added).

Validation

The practical objective of validation is to establish the fitness for purpose of the WQMS data, and to identify limitations of the service. Several approaches can help to establish this fitness for purpose.

Quantitative Validation Against In-Situ Measurements

A lack of accessible in-situ data on water quality is widely recognised as a problem for the development and quantitative validation of EO based water quality measures, and this is particularly the case in Africa. Where in-situ measurements are available, there are several practical constraints on their value. The underlying assumption is that the samples are representative of the area of water viewed from space. Samples that are taken near to shore, at depth, or at different times to the satellite observations will be less correlated with the EO observations making it difficult to establish a relationship.

To establish a connection between EO data values and in-situ observations we compared our ensemble algorithm results to in-situ data from the Global Environment Monitoring System for Freshwater (GEMS/Water) of the United Nations Environment Program (UNEP). GEMS stations from Morocco, Tanzania and South Africa were available between March 2000 and January 2016. Unfortunately these dates pre-date the Sentinel-2 satellite and the substantial improvements in EO data quality and quantity since 2017.

We generated remote sensing time series for each of the GEMS stations, using a 20m pixel size, for an area of 100m by 100m at each station site. EO measures of turbidity and chlorophyll-a were averaged over this area. In-situ data were included where they were close to the water surface (less than 0.5 metres for chlorophyll, 2 metres for turbidity) and within 7 and 10 days of the satellite observations. These filters were informed by examination of the relationship between depth and observed values at the GEMS stations. In regard to time, we assume that the relationship between in-situ and observed values will decline with time, but that some coherence will be present over a period of 7-10 days.

After filtering the in-situ data the number of data points for comparison was reduced from 2414 observations to 149 for chlorophyll, and from 754 to 36 for turbidity.

Results of the comparison reveal that:

1. The median measures from the EO data were comparable to the in-situ sites for chlorophyll; however,

2. The chlorophyll measures from the EO data used in the annual product do not capture low values well, and were not tightly related to the in-situ observations. Whilst some of this effect is due to the limited number of in-situ observations it also reflects the limitations of using an ensemble approach in which all measures are combined;

3. The median measures from the EO data for turbidity were higher than the in-situ data; however,

4. Turbidity from the EO data was well correlated with the in-situ data.

The findings indicate that:

• for the annual WQMS products, a simple linear adjustment will lead to realistic values, although these products will not capture variations especially at the low end of the scale;

• monthly monitoring products (which are prototyped in the initial release as notebooks) should use individual algorithms (as well as the ensemble approach) to support detection of low values and enable calibration using local data.

Validation results are summarised in the figures below.

Figure 10: Summary of validation results for the WQS monitoring service

Comparison with other studies

Validation of floating algal index results was achieved by comparison of results with previously published work as described in the figure below

Figure 11: Validation of floating algal index (FAI) results against previously published studies.

Qualitative Assessments

Qualitative assessments can support quantitative assessments of the utility of the information. In the figure below we compare the information provided by the WQS monitoring service for a large lake in Mauritania with our qualitative interpretation of EO imagery for specific dates.

Figure 12: Qualitative validation of WQS monitoring service data for a large lake in Mauritania