# Detecting Clay and Sand with SBI

## Business Case

Soil texture influences water holding capacity, nutrient availability, and yield stability. 

Traditional maps are costly and becoming outdated quite quickly. 

With GeoPard, you can:

* Flag **clay (darker)** vs. **sand (brighter)** areas fast
* Focus soil sampling where it matters
* Create **variable-rate** seeding, irrigation, and fertilizer zones
* Refresh insights each year with minimal effort

## Dynamic Tutorial

Results match **soil scans from German fields** measuring sand and clay content. 

A few targeted ground samples help fine-tune thresholds.

{% hint style="warning" %}
The tutorial below supports multiple languages. Choose the one that best fits you from the language icon <img src="https://3272281156-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYICBELdyAXXebKAzfLOR%2Fuploads%2FBI5BLx6lztxQTYIHqvsr%2Fimage.png?alt=media&#x26;token=0d402011-668e-4a9c-8853-0fb4f56ace99" alt="" data-size="line">.&#x20;
{% endhint %}

{% @arcade/embed flowId="wF4zb4RmoZmbHgjJaldA" url="<https://app.arcade.software/share/wF4zb4RmoZmbHgjJaldA>" %}

## Solution

How it works:

* **SBI (Soil Brightness Index)** rises with surface reflectance (red + NIR).
* **Clay**: finer particles, more moisture/oxides → **lower SBI** (darker).
* **Sand**: coarser, drier, higher albedo → **higher SBI** (brighter).

Workflow in GeoPard:

1. **Collect SBI layers** from bare-soil dates (several per year).
2. **Normalize each scene** (removes illumination/acquisition bias).
3. **Average across time** (cuts random noise).
4. **Cluster spatially** into **0.25-0.5 ha (0.5-1 ac)** zones (reduces pixel speckle).
5. **Classify** the distribution:
   * Lower SBI → **clay-leaning**
   * Higher SBI → **sand-leaning**
6. **Export** zones for scouting and VR prescriptions.

Normalization + multi-date averaging + small clusters make SBI a **stable proxy** for texture and improve separability between darker (clay) and brighter (sand) areas.