Global custom-tailored machine learning of soil water content for locale specific irrigation modeling with high accuracy
Abstract
A novel approach to irrigation modeling is presented: the locale specific machine learning of soil moisture data. The merits of this new patent pending technique are clear when compared to existing methods, such as the AquaCrop program created by the Food and Agricultural Organization (FAO). From a case study on the comparative performance of AquaCrop and machine learning in the extrapolative modeling of soil moisture, AquaCrop performed with a me an squared error of 0.00165 whereas the machine learning received 0.00013, an order of magnitude lower. In addition, a novel algorithm, the ConserWater™ algorithm, has been created for the purpose of machine learning soil moisture with accuracy and efficie ncy. The performance of the algorithm is very superior when compared to other popular machine learning techniques, as applied to soil moisture. Finally, to allow this technology to reach agriculturalists at the grassroots level, the entire world has been machine learned and the resultant models have been encapsulated into a lightweight easy -to-use smartphone application.
Keywords
Download Options
Introduction
The Earth has 1.4 billion cubic kilometers of water, but at any given time only about 200,000 cubic kilometers is fresh water that is accessible for human use [1]. In addition to the inherent paucity of total water supplies in relation to the escalating human population, anthropogenic traumata have already taken a significant hold. This is especially true in agricultural regions, as agriculture is the largest drain of fresh water [2].
A recent UN report has shown that small farms, preeminent in the developing world, provide the majority of food worldwide, and employ 3 billion people [3]. In addition, most of the people living below the dollar-a-day poverty line are a part of this group [3]. As they cannot afford sprinklers or drip irrigation for their fields, they resort to the age-old inefficient practice of surface irrigation: they flood their fields with significant water and wait for it to drain into their fields. As a result, not surprisingly, nearly 2/3 of global fresh water is used for irrigation [4]. It is by uplifting small farmers in developing countries that we can achieve another agricultural revolution and ameliorate global water consumption while optimizing plant growth.
It is envisaged that with good management and knowledge of the optimum quantity of water needed for irrigation, the efficiency of even the surface irrigation method can be driven up to as high as 90% [5]. To this end, several companies sell field-wide soil moisture sensor arrays to provide real-time data on crops water needs [6]. This approach can be efficient and it is claimed that it can facilitate a 25% decrease in water consumption during drought [6]. However, it can also be prohibitively expensive, except for wealthy farmers. Each moisture sensor can cost about $100 dollars, so implanting an array, along with operational services, can cost several thousand dollars [7]. This is clearly beyond the budgets of small farmers.
Conclusion
A monumental algorithm capable of efficiently machine learning soil moisture data to high accuracy has been presented. Compared to several popular existing techniques, it clearly has the upper hand in terms of accuracy. Additionally, it can provide a much more precise model of localized effects than programs such as AquaCrop, that rely on a simple water balance and evapotranspiration calculation as intermediate steps. This level of preciseness can translate to savings of tens of thousands of gallons of water for just an acre of land. Finally, this technology has been encapsulated into an easy -to-use lightweight smartphone application. These are necessities for the widespread proliferation of irrigation management technology around the world, especially to small farmers who cannot afford soil moisture sensors or other irrigation management technology. More information on using this technology , including download links, can be found on www.conserwater.com .