Fu, Y., P. Taneja, S. Lin, W. Ji, V. Adamchuk, P. Daggupati, Biswas, A. 2019. Predicting soil organic matter from cellular phone images under varying soil moisture. Geoderma: 114020. doi:10.1016/j.geoderma.2019.114020
Research summary by Aidan O’Brien
Key Messages
- This study relates soil colour to soil water content and soil organic matter (SOM).
- Cell phone images accurately estimated SOM when the soil water content is less than 10%.
- This research brings us closer to using smart phones for fast, accurate and non-destructive estimations of SOM.
Soil organic matter (SOM) is the backbone of soil health and is known to change soil color. But measuring SOM using traditional soil sampling and laboratory analysis is costly, time consuming and labour intensive. Digital images are increasingly being used to predict soil properties including SOM based on color. Cell phone images are becoming feasible and accessible ways to analyze SOM.
As soil images become better at predicting SOM, there is potential for cell phone images to be used as a proximal soil sensor for fast, accurate and non-destructive estimations of SOM. Recent advances in image-based SOM prediction from soil colour remain challenged by variables influencing soil colour, such as vegetation, soil moisture and illumination. To use cell phone images of soil to accurately predict SOM and support management decisions in the future, it will be important to quantify the impact of soil water content on the relationship between SOM and soil colour.
What they found
This study observed that the impact of SOM on soil moisture content and soil colour varies with initial soil water content. In fact, it is possible to use cell phone images to accurately quantify SOM on soil samples that has water content less than 10%. However, if the soil was wetter than this, we needed to first measure soil water content using other instruments (soil moisture sensors) and include that information in combination with the cell phone images in order to predict SOM. A critical moisture content of 10% is to be used in further studies for quantifying SOM from digital images.
Why it matters
Cell phone (and particularly smart phone) use in the general public has become a norm. Developing a cell phone app that can be used by anyone with a smart phone to estimate SOM offers the potential for more affordable, quick, and easy-to-use soil analysis, helpful for supporting management decisions. However, challenges still remain in estimating SOM from soil colour when the soil is moist or wet. Our research continues to simplify the process of accurately predicting SOM from soil images taken with cell phones.
How they did it
Images were captured using a 10-MP cell phone camera for 25 topsoil samples at 6 varying moisture levels (total of 146 images of soil samples – 4 corrupted during data transfer). The images of the soil samples were pre-processed to account for differences in illumination, and other non-soil objects. Simple univariate regression was used to examine the effect of soil moisture content on SOM prediction using colour parameters. Separate regression models were developed for each moisture group and the optimal colour parameters were determined by comparing the model performances.
The relationship between SOM, soil moisture content and colour parameters was examined using scatter plots. The critical soil moisture content was determined based on the change in distribution of colour parameters contributed to the significant variations in correlation. So, critical soil moisture content represents the soil moisture content at which the soil moisture begins to significantly influence colour parameters. After critical soil moisture content was determined (10%), the images were split into two groups 1) SMC > 10%, and 2) SMC < 10%, in order to calibrate and validate the predictive model.
