PORTFOLIO

Space4Good has collaborated with partners all over the world to work on challenging and impactful projects. Explore a selection of our projects below.

Tip: Click on a marker to go to a project or simply scroll down

 

Bark Beetles

Analysis of past bark beetle monitoring and ways forward

Over the year 2020, Space4Good has monitored forests in Switzerland for bark beetle risks. Bark beetles often target weakened trees by, as the name suggests, breeding between the bark and the wood of the tree. In some cases, even healthy trees suffer from bark beetles, due to the sheer amount of bark beetles overwhelming the tree's defences. Outbreaks seriously impact the lumber industry, water quality, fish and wildlife and property values.

Bark beetles breed between bark and wood

Beetles affect lumber industry, water quality, wildlife and property value

Space4Good used two risk models that classify the area in 7 risk levels; 1 being very low risk and 7 being high risk. Both models use a combination of Sentinel-2 imagery, soil information, weather data and forest conditions. 

The first model, which was conducted during the season, takes all parameters into account, with a heavy emphasis on soil characteristics. The model resulted in a classification where the largest area is classified as risk level 1 and the largest number of outbreaks are classified in risk level 7, as one would expect.

The second model focused more on parameters as described in scientific literature. The results are less clustered than the first model, but get pixelated and require aggregation based on pixels with the same value. This makes it easier to see high-risk level areas and visit them.

In the future, the models will be improved to prevent overlooked outbreaks by adding Sentinel-1 imagery and shifting to a more data-based approach.

 

Biomass Estimation

Remote sensing based biomass estimation

Biomass estimations using remote sensing provided unlimited opportunities. Carbon sequestration monitoring, carbon credit markets, predictions. Space4Good has applied two distinct approaches to ascertaining biomass in selected projects in Southeast Asia, Africa, and South America. But how do we do it? Using two approaches for biomass estimation in three different plots in Indonesia, Kenya and Brazil, we were able to make the decision for the most accurate. 

The first approach utilizes drone imagery, RGB photos, and Google Earth imagery on which we apply an object detection algorithm to identify individual trees. Based on this tree and the various size clusters determined by the algorithm,  biomass for the entire plot could be calculated. This first approach turned out to be too inaccurate to use in different plots, due to differences in tree species, age, and structure.

First approach

The second approach used a generic biomass algorithm, which requires specific tree data inputs Coupled with a deep-dive investigation of species metrics, we were able to then apply the algorithm to ascertain an above-ground biomass estimation. The final estimate uses a weighted average of the biomass depending on the area covered by each species.

Although both approaches have their inaccuracies to start, these will decrease as reference plots and training data are input into the system, improving the AI-based algorithm.

Second approach

 

Carter Center

Using satellite imagery to verify, assess and monitor the consequences of military conflict

In 2020, Space4Good met with the Carter Center at the Data for Peace & Security Conference in The Hague. We connected over remote sensing and geospatial analysis to monitor and support conflict resolution in war-torn regions. It was then that we saw major potential and proposed potential solutions in how to support assessing the consequences of military conflict by providing remote sensing-based insights and increasing transparency. Carter Center agreed and we started our project focused on the Al Bab District and Harim District in Syria.

Using change detection analyses to assess damage to buildings and neighborhoods as a proxy for remaining explosive weapons, Space4Good. Optical change detection and multi-spectral change mapping using Sentinel 1 and Sentinel 2 data were used in conjunction with optical artificial intelligence algorithms to identify buildings. Then, combined with filtering out seasonal fluctuations and noise, we were able to detect small-scale changes during and after bombardment events. Further combining these with land use classification mapping, more accurate results could be had. 

The project enabled a better understanding of the consequences of conflict, urban damage, and growth for good with the added benefit of protecting lives on the ground and supporting peacebuilding activities.

 

Fire Detection

Near real-time feedback on fires in mixed tropical systems

Space4Good builds on fire detection insights offered by NASA and Global Forest Watch by offering algorithmic analyses of Sentinel datasets overlaid with existing datasets. These insights are then fed through pipelines directly to local field workers to better mitigate the destruction of fire activities in mixed tropical forests in good time. The pipeline as of now is as follows:

A server will check the NASA fire detection hourly. Once a fire is detected in the project area in Kalimantan, Indonesia, Space4Good’s algorithms then determine a fire alert. This alert is then disseminated to in-field operatives via a navigation application on the phone. This is also able to be used offline to ensure nothing is missed. Additionally, messages are pinged to WhatsApp groups to alert fieldworkers. This pipeline occurs in a matter of minutes. Field operatives then navigate their way to the identified locations and validate if the event occurred (or not). Our partner will then inspect the area and send us their feedback with the location of the event. 

Fire alerts detected in 2020. Click on a marker for more information

 

Flooding Detection

Near real-time feedback on fires in mixed tropical systems

In January 2021, the World Bank requested a quick high-resolution flood assessment, mainly around the Casablanca area in Morocco. This result could serve as a second reference to the Global Precipitation Measurement data that was being used by several World Bank clients and insurers.

The assessment started by selecting 2 before and after Sentinel-1 images for the flood extent mapping. Next, a thresholding method was used on the Sentinel-1 difference images. The method was further refined by masking the flood pixels from areas of permanent water bodies (i.e. where there is water for over 10 months per year) and the areas with more than 5% slope. The flood area was estimated in ha with a threshold measurement to boot.

In order to validate which of the 2 maps was more accurate, news articles and real-time information from the local residents of Casablanca who reported the flooded streets were used. Comparing field data with the flood extent maps validated the outcomes well. 

Flooding detection in Casablanca

 

Illegal Logging

Illegal logging detection and prediction for mixed tropical systems 

Deforestation will release as much CO2 in the atmosphere in 24 hours, as 8 million people flying from London to New York. It is estimated that in 100 years' time, there will be no rainforests left if deforestation continues at the same present rate. The key to combating these pressing problems is predicting and detecting illegal deforestation practices to ensure a quick response.

There are currently platforms, such as the Global Forest Watch, that offer these services. The problem, however, is that the imagery from the MODIS and Landsat satellites is not reliable, due to a low temporal resolution and cloud cover.

Space4Good has started a collaboration with Arsari Enviro Industri to monitor forested areas for deforestation events using imagery from the revolutionary Sentinel-1 satellite. This imagery comes with a high temporal resolution and can see through clouds. The goal of the platform is to deliver a reliable and near-real-time monitoring platform that can detect deforestation events at a high spatial resolution.

In combining  Sentinel-1 data with artificial intelligence, we are able to remotely monitor deforestation in an area covering thousands of hectares and alert local authorities in near-real-time to the occurrence of an event. We can even go so far as to ascertain risk areas and predict deforestation events by taking into consideration pattern recognition using AI, land-use change, and a multitude of other factors to determine drivers and therefore with a level of certainty, key areas where deforestation is predicted. The platform is an innovative evolution of other platforms and improves on them by offering a faster detection of both the extent and time of the deforestation.

There are several improvements coming up that will eventually refine the platform even further. Namely, the algorithm will be expanded upon by adding slope correction and seasonal changes to the process. Soon, deforestation alerts will also be automatically sent on WhatsApp and not just email to local authorities for a speedy and applicable insight for mitigation.

Detected deforestation in 2019 and 2020. Switch layers on and off using the toggle and get more information by clicking on the polygons

 

NO2 Platform

Monitoring economic COVID-19 response and recovery from space

NO2 emissions are generally caused by anthropogenic activities, such as fossil fuel combustion, which makes it a great indicator of industrial activity and motorized mobility. To measure air quality, NO2 used to be measured using ground stations. Nowadays, satellites are a better alternative to this, as it allows for measuring on a larger scale and are independent of geographical obstacles. A new satellite mission, Sentinel 5P, is capable of detecting tropospheric NO2 using the TROPOMI instrument. The Sentinel 5P resolution has a higher spatial resolution when compared to other satellite missions with the same purpose. This high resolution enables analysis on a regional, national and sub-national scale.

NO2 emissions are generally caused by anthropogenic activities, such as fossil fuel combustion, which makes it a great indicator of industrial activity and motorized mobility. To measure air quality, NO2 used to be measured using ground stations. Nowadays, satellites are a better alternative to this, as it allows for measuring on a larger scale and are independent of geographical obstacles. A new satellite mission, Sentinel 5P, is capable of detecting tropospheric NO2 using the TROPOMI instrument. The Sentinel 5P resolution has a higher spatial resolution when compared to other satellite missions with the same purpose. This high resolution enables analysis on a regional, national and sub-national scale.

NO2 emissions in Vietnam, before and during lockown. Toggle layers on and off to see changes

The Space4Good platform has several advantages over similar platforms, such as elaborate information in pop-ups and it is more detailed on a local level. In a second phase, more information will be added to the pop-ups and there will be more capabilities to compare locations.

Did we spark your interest?

Get a demonstration of our portal by sending us a message!

 

PARSEC

Pest and disease early warning and prediction platform

Farmers all over the world are having trouble in their daily decision-making when it comes to the health of their crops. Inefficient horticulture and the following plagues and diseases can cause a decrease in yield of up to 40%. Per day where no action is taken, the yield can decrease by 1-3% and the profit by 3-5%. It is also difficult to predict plagues and diseases and detect them in an early stage, which often makes the approach to get rid of them too large and is damaging to healthy crops as well.

Space4Good has in cooperation with Fauna Smart Technologies founded Space4Fauna to help farmers effectively combat plagues and diseases and increase their yield.

Space4Fauna has created a mobile application that is capable of predicting and detecting plagues and diseases. The moisture, greenness, humidity, and temperature of crops are monitored using satellite imagery and are used to determine and map the risk of a plague or disease outbreak. Farmers can monitor the state of their crops and take action accordingly.

Effects of plagues and diseases

The field is divided into healthy crops, mild pest attack and high level of pest attack. This information is connected to the Knowledge DataBase, which determines the best environment-friendly and economically viable solutions to handle an outbreak.

 

Rotterdam Rooftop Analysis

Developing a Roof Potential Map forRotterdam

With the Netherlands being one of the most densely populated countries in the world, efficient use of space is crucial in urban planning. Space4Good has collaborated with zoarchitecten and the Municipality of Rotterdam to map existing uses of roofs in the city center of Rotterdam. This resulted in a Roof Potential Map that urban planners and architects can use to find opportunities for practical and environmentally friendly solutions.

The final product consists of a map that visualizes the existing uses of the roofs, classified into 16 categories, such as blue or green infrastructures or even solar energy.

How do we do this? Together with the University of Leiden, we developed an algorithm that uses machine learning and deep learning methods to automatically detect existing roof types for potential clients, generating usable maps for planners and decision-makers alike.

Rooftop classification - Rotterdam City Center