Satellite Analytics: The Technological Key for Agriculture in Arid Environments

Arid regions present significant challenges for regenerative agriculture but also immense opportunities. With the global population continually growing and agricultural yields declining by 5% to 17% for each degree Celsius of warming (source: IPCC ), humanity must find new cultivable land. Every year, 13 million hectares of land are lost due to desertification (source: UN Convention to Combat Desertification ), intensifying the pressure on existing arable land. According to the IPCC, an additional 593 million hectares of land will be required by 2050 to feed the global population. Arid regions, often seen as inhospitable and underexplored, offer untapped potential that urgently needs to be explored.

For Sand to Green, assessing the land for our clients is the first step in unlocking the value of these complex environments. The areas to be analyzed are often vast, covering thousands of hectares, which complicates the process. For example, for one of our clients, we analyzed three plots of several hundred thousand hectares to identify the best site for an agroforestry and regenerative agriculture project.

Land analysis focuses on several key areas:

1. Topography and Physical Constraints Certain slopes make the land unsuitable for regenerative agriculture without modification. A plot with a slope of less than 5% is ideal, while slopes between 5% and 10% require adjustments, and beyond 10%, the land becomes difficult to exploit.
2. Water Cycles and Resource Management In arid regions, understanding the cycles of rainfall and evapotranspiration is crucial. We often work with a fixed volume of desalinated water per month, requiring optimal management. It is essential to store water during surplus periods to use it during times of deficit.
3. Choice of Plant Species Each plot has its own specificities (pH, salinity, wind, etc.). To maximize the success of plantings, the selection of species must be adapted to these local constraints.

Data Collection for Arid Lands

Arid zones cover about one-third of the Earth's land surface. The lands to be analyzed are geographically diverse, which requires data covering vast areas. However, the availability of agronomic data is uneven across countries: in the United States, land use maps are accessible, while in many African countries, this data is scarce. To overcome this obstacle, Sand to Green relies on satellite data, which provides a fast, precise solution covering a wide geographic range.

These analyses use different spectral bands of light, such as near-infrared (NIR), mid-infrared (MIR), and thermal bands. These wavelengths allow us to detect elements such as soil moisture, biomass, or salinity. Each material reflects light in a unique way depending on its composition, which enables us to extract precise information about soil structure, vegetation, and the overall health of ecosystems.

With data resolutions ranging from 30 to 250 meters, we can create highly detailed maps that highlight the best areas of a territory and their ecological dynamics.

Furthermore, billions of data points are available for free via Google Earth Engine, totaling more than 90 petabytes of information. Thanks to Google's computing power, these analyses are performed directly on their servers, allowing for rapid results despite the complexity of the calculations.

For instance, using a vegetation index (NDVI), it is possible to clearly distinguish areas with healthy vegetation (dark green) from those that are suffering (orange). Each point on the map can be explored to assess its evolution over time. In the example provided, vegetation health has steadily improved since 2014 but has started to plateau since 2022.

These data can also be aggregated to gain a better understanding of the studied environment, which can be particularly useful for assessing pH or salinity.

Satellite analyses can even provide data on soil composition, which helps us better understand the land's amendment needs before considering planting crops!

Using Data to Maximize Impact

These terrain data, in addition to being informative, have allowed us to develop advanced features:

1. Feasibility Validation: We have integrated a feature that quickly checks whether a plot is suitable for agroforestry and regenerative agriculture planting. This preliminary analysis is based on topographical, climatic, and political criteria. For example, if a selected plot has steep slopes or is located in a protected area, the user is immediately alerted that this plot may not be suitable for agricultural activity.

2. Selection of Cultivable Plants: By using criteria such as pH, altitude, soil salinity, prior presence of the plant in the country, and Köppen climate classification, our software can determine which species are most likely to thrive on a given plot. The selection process is carried out from a database of more than 2500 species, which serves as a starting point for designing a specific agronomic plan. This enables the selection of the most relevant species based on the client's objectives, whether for fruit production, forage, or environmental rehabilitation.

3. Land Comparison: By accumulating data from different plots, we are able to compare their agricultural potentials. This comparison helps our clients select the land with the best conditions to achieve their goals, whether it’s agricultural yield, ecological resilience, or sustainable production.

In Summary

In summary, the integration of satellite analyses into the study of environments represents a true revolution. These data, free of charge, globally accessible, and historically archived, offer enormous potential. Google Earth Engine, by making them available to its users (with a simple Google account), opens exciting prospects for many startups!

Of course, like any external-facing computer system, this creates complexity for developers (especially in error management), but the value creation for the product is undeniable.

It should be noted, however, that the application of these analyses is particularly relevant for large plots, as the resolution of free images is often insufficient for detailed analysis of smaller surfaces. For specific precision needs (down to 50 cm), paid images will be required, allowing for fine-grained analysis at a smaller scale.

Article written by: Louis-Marie Bonnefont, CTO at Sand to Green