Soil salinity mapping in the Lajas Valley - Puerto Rico, using Watson Studio AutoAi
Updated: May 2, 2021
BEVERLY ÁLVAREZ TORRES1, JOSÉ PABLO CASTRO CHACÓN1, DAVID SOTOMAYOR RAMÍREZ1, LUIS PÉREZ-ALEGRÍA1, GUSTAVO MARTÍNEZ RODRIGUEZ1, & TOM DESUTTER2
1 University of Puerto Rico - Mayagüez Campus (UPRM)
2 North Dakota State University (NDSU)
Keywords: artificial intelligence, soil, salinity, sodicity, mapping, agriculture, soil saturated paste, soil:water solution, agricultural reserve, webapp
Hello!👋🏼 My name is Beverly Alvarez and I am Biologist and my name is Pablo Castro, I am Geographer. We are both Soil Science master candidates from the University of Puerto Rico- Mayaguez Campus, under the mentoring of Dr. David Sotomayor Ramirez, Dr. Luis Perez Alegria and Dr. Gustavo Martinez Rodriguez from University of Puerto Rico and Dr. Tom DeSutter from North Dakota State University. Today we will talk about soil salinity mapping in Lajas Valley using Watson Studio AutoAI.
Saline and sodic soils are soils with a high content of soluble salts. Soil solutions with high salt concentrations decrease the water availability for plants, minimizing the adsorption of nutrients. The result is a reduction in the germination, growth and crop yield. Some of the visual signs in plants are physiological drought, leaf curling, chlorosis, burning, and necrosis.
Figure 1. Visual signs and effect of soil salinity and sodicity in soils of the Lajas Valley, Puerto Rico
Figure 2. Anions and cations released when the soil salts solubilize
Soil salinity and sodicity is a big agricultural problem. For 2013, it had an economic impact of $ 27.3 billion worldwide, due to decreases in crop yields (Qadir et al., 2014). It is estimated that over 100 countries have saline and/or sodic soils, with Australia, northern and central Asia and South America being the most impacted regions. Global warming is expected to contribute to the development of saline and sodium conditions in soil in other parts of the world. Furthermore, with the population increase expected by 2050, the need for fertile soils for agricultural production will be vital.
Figure 3. Some of the countries with saline and/or sodic soils
💡An example of how salts are harmful to agriculture was evident during the Third Punic War between Carthage and the Romans. The Romans devastated the city of Carthage and plowed the agricultural lands with salt, damaging permanently the soils and condemning the population to hunger.
Puerto Rico is one of the countries with these types of soils. At the southwest of this island, located in the Caribbean, is the Lajas Valley, an area classified as the first agricultural reserve. In 1958, soil scientists Bonnet and Brenes estimated that 48% of the valley soils at a depth of 60 cm presented saline and/or sodic conditions. In 2020, the Project H-483: An updated assessment of soil salinity of the Lajas Valley Agricultural Reserve, Puerto Rico determined that the distribution was reduced to 31%. Both results demonstrated the reduction of soil salinity and sodicity, an expected result after 60 years of the building of a regional irrigation and drainage system known as the "Proyecto del Suroeste".
Figure 4. Zoom to location at the Lajas Valley, southern Puerto Rico
Projects to assess soil salinity at a regional scale are expensive and require highly trained technical personnel, which is why they are usually funded by government agencies such as USDA-NRCS or global entities such as FAO. At local scale, it is extremely complex for farmers to perform salinity analysis on farm soils due to its high cost and the need for technical assistance. In general, the recommended analysis to evaluate soil salinity is the measurement of three parameters in saturated paste: electrical conductivity (ECe), pHe and sodium adsorption ratio (SARe) (USSL, 1954):
Figure 5. Soil parameters analyzed to classify soil salinity
The lab analyst requires physical effort to process a saturated paste soil sample and a minimum of 48 hours before obtaining a result. Other alternative methods have been developed, such as soil:water solutions, a method that reduces the cost of analysis, the amount of soil sample used and the processing time, becoming a fast and economical alternative to monitor soil salinity.
The objective of this project was to create Soil Salinity AiCalculator, a tool capable of predicting ECe, SARe and soil classification using the database generated in the H-483 project. This project generated a database with 346 soil samples analyzed by both methods, soil saturated paste and soil:water solutions at a 1:5 ratio. The structure of the database is compose of:
Depth: soil sampling depth (1- 0-30 cm, 2- 30-60 cm, 3- 60-90 cm, 4- 90-120cm)
EC_15: electrical conductivity of soil:water solution at 1:5 ratio
pH_15: pH of soil:water solution at 1:5 ratio
ECe: electrical conductivity of the soil saturated paste
SARe: sodium adsorption ratio of the soil saturated paste
Soil Salinity Class: based on the classification guide by United State Soil Salinity Laboratory (USSL, 1954)
Figure 6. Soil classification for saline and sodic conditions
The training and test files, along with the python notebooks, are located in the following folder on google drive.
The database was uploaded to the Watson Studio platform and three different prediction models were generate: a regression model for ECe, another regression model for SARe, and a classification model for soil type, all using the new Watson Studio AutoAI tool.
Watson Studio AutoAI is designed to analyze a data set and generate custom AI models capable of predicting a decision or variable by optimizing multiple models to the best possible result, a task that could take weeks or years if performed by humans. The presentation of the results provided by the platform is rich from the visual point of view, with useful and easy-to-understand graphs and tables, such as the selection of the best models, the transformations that are carried out and the weight of each of the variables in the model.
Figure 7. To access the platform Watson Studio AutoAI, press here
The best prediction models were stored using IBM's Cloud Object Storage, deployed and put online to be tested by users through three APIs, one for each variable to be predicted. The results from Watson Studio AutoAI were mixed with a free map from the Leaflet application and the results were stored in a Google FireBase database. Proving that in addition to its usefulness for data processing, we can also connect Watson with multiple other applications.
💡 It is a fast, easy and accessible tool for agronomists, scientists and farmers to constantly monitor soil salinity
💡 It allows keeping a record of all soil samples, feeding a database to continue doing AI and improve predictions
💡 It is a cheap tool for the end user, it accelerates the delivery of results and decision making related to crops that need fast results
📲 Aitech tool accessible, cheaply, quickly and reliably for junior users
📲 Allows to generate applications from the beggining to the end user impacting positively the socialization of science
📲 Free versions for simple projects
📲 Available for use in multiple science research
📲 Huge data processing power without the need of heavy hardware,
📲 Eliminates barriers to access knowledge
📲 Opens new possibilities for online agrobusiness such as applications, consulting, training, and workshops
🌱Give prediction models, generating statistical transformations and detailed results automatically, reducing the heavy work of statistics
🌱 Reduces the amount of soil sample, minimizing the environmental impact of the samplings
🌱 Reduces laboratory work related wit soil sample process and analysis, reducing the fatigue of the human beings who have to carry out the soil tests since they are quite laborious.
🌱 Allows to take faster agricultural management decisions
🌱 Offers a tool to improve the long term management and development of saline and sodic soils worldwide
Now Soil Scientists have more time to make discoveries!✌🏼✨
Graduate school is a space to meet wonderful people and do things that exceed our limits like a Soil Salinity AiCalculator with IBM and its Watson Studio AutoAI😂. If you have questions, comments or just want to tell us something, contact us! 😎 You just have to send us a message when visiting our social networks on:
💕 Learn more about what it's like to be a Soil Scientist👩🏻🔬 by following us on:
🌱 Support other UPRM students and researchers of the Soil and Water Quality Research Laboratory on:
And if you want more posts like this one, tell me! Let's create an ever-growing community that promotes agricultural science, agrocareers, and the importance of self-esteem and professional development. . . because by growing a plant, you cultivate your heart 💚.