How Nature and Climate Will Benefit from Advances in Artificial Intelligence and Remote Sensing
Monitoring, reporting and verification (MRV) of carbon captured by trees, crops, soils, hedgerows and other ‘nature assets’ requires effective technology. This is usually by earth observation from satellites where algorithms are used to convert the images obtained to leaf area index and biomass enabling authoritative information on the state of the nature asset and estimates on the amount of carbon captured.
Although traditional ground measurements are expensive in labour, they are currently the standard MRV method and are always needed to correlate satellite data. For satellite observation a variety of areas can be used, but the larger the area monitored, the cheaper the cost. The use of artificial intelligence (AI) and machine learning (ML) is facilitating processing that in near-real time, accurately, and cost effectively to a consistent standard.
As a standard, satellites with resolutions of about 22 metres were used but now a variety of high-resolution satellites can be used down to 0.3 metre, which gives greater accuracy in assessment and can identify individual trees. It also increases cost significantly and has less frequent data collections and limited historical archives. Conversely the Sentinel constellation with a resolution of 10m provided by the European Space Agency is free and has data back to 2016 so there is the ability to back test data. The fact that you can go back 5-6 years allows for historic activity to be measured and monitored to understand the impact of decisions and investments made to date. Planet Labs Planetscope data at c 3-5M is a good in- between option. Data choice and their costs is an important factor to consider as MRV projects will take decades to monitor the carbon capture life cycle as it takes about 30 years for a tree to grow and mature.
In the example below using Planetscope at a resolution of 3-5 metres supplied by Earth-i

The real cost however includes processing of data to give information, and it is usually more cost effective to buy this as ‘analysis ready data’ rather than imagery alone. Generally, the cost of such information has been decreasing while quality and frequency increases.
This is illustrated by the image below, again by Planetscope at 3-5 metres resolution supplied by Earth-i, showing the power of identifying landscape and vegetation change using satellite imagery.

For carbon and nature accounting, information needs to be provided in such a way that national and international agencies find acceptable. In the UK, Forestry Commission sets the guidelines and the Woodland Carbon Code, part of the International Carbon Reduction Offset Alliance, is recognised by them as a certification standard, but they do not usually use satellite data. On the ground inspections will not be sustainable at the scale required of a country or even county level. The UK is a relatively small country at 250000 square km and the 78th largest in the world. Internationally the UN, through FAO, is a key agency and countries usually align to those, sometimes supported by the World Bank. They will need to show leadership if the recommendations of COP26 are to be implemented.
Generally, agencies are keen that MRV is fitted into the context of Land Use and Land Use Change and Forestry (LULUCF) as part of assessment of Global Landscapes. One of the challenges is that trees are often planted to achieve maximum biomass production and therefore achieve maximum carbon credits, but this may not be the best outcome for the environment. For example, the extensive planting of non-native Sitka spruce in Scotland in the post-World War II period resulted in land degradation and habitat destruction. Clear felling also generated concerns of the visual landscape quality. There is a balance that needs to be found between promoting biodiversity and maximizing carbon sequestration
Continuous cover forestry with broadleaves and evergreen trees makes sense for both nature and climate. MRV should therefore include assessment of environmental impacts and one of the favoured procedures for this is Life Cycle Assessment (LCA) in which all the aspects of land use for agriculture and forestry can be estimated to demonstrate the value to nature and climate of the land management being used. Assessment of biodiversity change of key indicator species is often a good guide to environmental quality. Measurement of soil carbon, which can be assessed from satellites, is important in determining soil health.
In terms of carbon trading, it is likely that assessment agencies will demand an increased vigilance in MRV of land management and it is important that companies involved in trading have access to the technological capability to satisfy such demands as the carbon market develops while balancing the demands of supporting biodiversity. The challenge is to realise the opportunities while keeping unit costs as low as possible – in general, this is best achieved through ground measurement being replaced by remote sensing, enabled by technology and scale. We say that 80% accuracy now via remote sensing is better than 100% accuracy in two years and too late for optimal decision making. Also, accuracy will improve over time with more data points, targeted ground truth validation, and increasing accuracy and variety of future remote sensing satellites and AI methodologies.