(Yicai Global) May 26 -- Agriculture has always been a weather-dependent industry, with dozens of decisions based on the seasons and climate. Beijing Gago Group is a big data company that hopes to help farmers with large land areas better manage production by using satellite data.
Da Buxi, an Inner Mongolia-based farmer, grows 1,000 hectares of alfalfa, a high-protein flowered grass that demands special attention to grow. He plans to sell the grass, and in order to maintain its protein content and color, needs to dry it out in the sun.
During the harvest season, Da needs a whole rain-free week to dry the grass effectively. If he mistimes it, he'll need to dry the alfalfa by machine, which significantly reduces its value.
Da regularly has to watch weather forecasts on TV or through mobile apps, but these cover vast areas of land and aren't specific to his farm. Over the pastures of Inner Mongolia, it's not uncommon for the east to be sunny while rain pours in the west, so the forecasts are of little use. Last year, Gago, an agricultural Big Data company, approached Da and recommended Yunjing, a farm management platform which provides weather predictions, including temperature, humidity, wind speed and rainfall, for smaller areas. Da decided to give it a try, and splashed out tens of thousands of yuan on the service.
Initially, he wasn't happy. He would receive notifications through WeChat, a major Chinese messaging platform, which warned him of when the rain would come. However, it was often inaccurate. "It told me it would rain at 06.00 a.m., but it wouldn't rain until 09.00 p.m.," he said. "When people were out harvesting, I looked a fool just standing watching, expecting rain."
A month later, Da received warning of a big storm. Given the severity of the notice, he harvested his alfalfa earlier than planned. Three days later, a hailstorm destroyed the surrounding pastures, with only Da avoiding the loss of crops. This put one of Gago's founders, Zhang Gong, at ease, as he feared Da Buxi was losing trust in the platform.
More and more people have begun to farm on larger areas of land since China pushed the circulation of land use rights. Agriculture is an asset-heavy industry, and requires large payments up front for land. Even if a bad decision only results in a 1-percent loss, the overall impact on efficiency would be huge.
Gago sought to help those farmers more effectively manage their land. "If a farmer only owns a small area of land, he probably knows that land better than we do," said Gu Zhu, one of the platform's founders. "But if they own thousands of hectares, they need our services."
The Gago team are not farmers, they don't work the fields. They are the 'farmers' eyes,' watching the land with satellites far from earth. The platform gets its data through purchases and cooperative agreements. Image and data analysis algorithms are used to extract spatial and geographical information from this information and combine past meteorological and vegetation records to create a database. This allows Gago to offer services covering plot measurement, growth planning and monitoring, output predictions and disease and insect control.
In September 2015, Gago secured tens of millions of yuan in funding from Matrix Partners China and Grains Valley Capital. In April this year, it finished a CNY60-million (USD8.7-million) financing round led by DCM Ventures and with further injections from Matrix Partners and Grains Valley.
Zhang Gong served as a data scientist at NASA's Ames Research Center for eight years. There, he was responsible for monitoring the growth of surface crops by analyzing satellite data. He turned his eye to agricultural conditions in the US, where a few family farms were able to successfully manage huge farms through 'precision farming.' Since the beginning of the 1990s, America has used satellite technology in agriculture to analyze soil fertility and plan their crops according to growth conditions. This precise management method can raise output by as much as 30 percent.
"America has acquired a bonus from scale production, but failed to reap rewards from precision management," said Zhang Wenpeng, vice president of Gago. "Refined management is needed to increase earnings."
In 2013, sustainable agricultural firm Monsanto Co. Inc. [NYSE:MON] bought Climate Corp., a digital agriculture company, for USD930 million. Climate Corp. uses historical climate data to provide accurate weather forecasts to farmers and guide crop production.
This acquisition spurned on Zhang Gong, who pondered how people were commercializing agricultural technology. He noticed that many people specializing in agricultural Big Data in the US had once worked at Ames, so decided it was time to move back to China and start a business. Early on, he brought in Wang Yungang, who worked at the US Department of Energy, to develop an in-depth learning algorithm model for satellite image recognition. He realized that this would be a key advantage and could form the core of the company.
The entire algorithm system took about two years to come to fruition. They chose to use satellite data at first, with the choice between WorldView, which was high-definition but slow, or China's public satellite, which was low quality but a faster service. However, the data these services provided needed to be translated to imagery, which took manpower. Zhang Gong aimed to auto-analyze the data using algorithms to cut out the middle man. Later, Gu Zhu, who once led machine learning at NASA, joined the team to take charge of plot recognition.
Before resigning from their posts in America, Zhang Gong and Wang Yungang visited more than 10 provinces in southern and northern China to assess agricultural productivity. The amount of machine-assisted labor was far beyond Zhang's expectations. The pair saw that in the countryside, agricultural machinery was commonplace, with only those aged around 55 or over still relying on their hands.
"Large bodies of farming are outsourced to agricultural machinery owners," said Zhang. "They would spend months reaping summer wheat from northern Hubei to Inner Mongolia. China's use of machinery was even higher than the US. These advancements were all things in which data could assume a role." In the middle of 2015, Zhang, Wang and Gu returned to China and founded Gago.
Gago takes aim at "big field farmers" ever since its inception, namely, large cooperatives, agricultural groups or individuals with more than 2,000 mu (1.333 square kilometers) land. Their first client is a cooperative in Qipan village of Jilin city in northeastern China's Jilin province, with cooperation started from the most easily recognized parcel management. Unfortunately, the 6,000 mu (4 square kilometers) land controlled by the cooperative is divided into several hundred blocks, which are distributed in several villages. So, even circling the farm by car would take quite a long time. Apart from separated distribution, these lands are also characterized by diverse kinds of crops. Potatoes are planted in the middle of some lands, with corns raised all around. Under such circumstances, it is hard to imagine the working load and the cost if the cooperative is to find out the condition of the parcels and crops by totally relying on traditional manpower. Moreover, farming work requires meticulous attitude, but who would be truly willing to go to the field for check in the sweltering sun?" Zhang Wenpeng said.
That was the moment when Gago's deep learning algorithm model came into play. Via image identification and deep learning, the satellite imagery system is able to judge farmlands in terms of land texture, field ridge lines and distance between plants. It allows to mark different land locations of the client by multiple colors, followed by proper division of land boundaries and calculation of total land area.
This "virtual intensification" means the system interface enables the customer to remotely monitor land, 'as plain as daylight.' "Leaving aside refinement, we initially target basic management, aiming to tell the client the land size, the current crop cultivated, as well as the definite location of the land," said Zhang Gong.
Real-time monitoring of the growth conditions of tens of thousands mu of crops also acts as a significant part in the basic module of Yunjing System. Zhang Gong has specialized in analyzing this type of data thanks to his earlier work experience at NASA. For example, plant leaves in the process of growing will absorb red light from the sun and reflect near-infrared light. In such a case, the Yunjing System is able to screen data related to the red light and near-infrared light. Then, the system can judge the photosynthesis of the plants in a certain land by means of the algorithm model, which, coupled with marks in different colors, will make the growth trend of plants crystal clear. At present, Gago boasts an accuracy of 0.5 meter in terms of its growth trend forecast model.
So far, Gago has realized real-time supervision on each block, with a view to providing quantitative basis for the customer's farming decision-making. A good case in point is the plague of insects in a given parcel. In the old days, farmers had no choice but to spray pesticides covering all the range. But now, precise water replenishing, disinfection or fertilization for a tiny parcel will become a reality once the growth condition of each land is understood.
Dabuxi's clover field is composed of various round grass circles, with irrigation equipment installed in the middle of each circle. The mechanical arm stretching out from the equipment is akin to the pen point of compasses, with a diameter of 120 meters. "We found a contrasting growth situation between the middle parts and outside of grass circles through satellite image," which was surely caused by watering problems. However, the client claimed he did constant watering and fertilization on time. A field inspection showed that something was wrong with the sprinkling irrigation of that grass circle, "In the case of a grass circle with a diameter of more than 100 meters, it would be extremely difficult to identify such a problem just by manpower, and the client would not know it," Zhang Wenpeng said.
Alfalfa, planted by Dabuxi, is a perennial plant. In case of a low temperature in a cold spell in late spring during the period of seedling establishment, the seedlings could be frozen to death. Thus, it takes time and effort to replant them. However, with the weather forecast of Gago, if irrigation is cut down ahead of the expected cold spell in later spring, which would slow down their growth, the seedlings would not be affected by the cold, because they are short and close to the soil. Hence, there is no need to replant them.
However, the reality is not so smooth. It takes a long time to confirm data and match them with users' needs. At the very beginning, when he provided weather monitoring service to Dabuxi, Zhang Gong used the commonly used weather forecasting methods in the city. The information on the weather conditions of each period of a given day is pushed aside. Dabuxi was not interested in the exact rainfall amount precisely. What he was concerned about was whether the rain would be very heavy on those days to prevent agricultural machinery use.
Later, Zhang Gong took the rainfall data along with him to meet Dabuxi. He asked Dabuxi to select the days with heavy rain so that he could update them accordingly. Eventually, they developed a weather forecast solution for Dabuxi. The new solution only forecast possible rainy days. The necessary info has been entered into the system. For instance, the first day may indicate thunderstorm, the next day moderate rain.
Gago gives advice on agriculture, including whether a given day is suitable for mechanical operation or not. "Agriculture only needs the change process in weather conditions rather than too much accuracy," Zhang Gong emphasizes.
Dabuxi's is not exceptional case. Though they are accurate, satellite data and analysis may deviate from the actual situation. Zhang Gong often finds that the fields with satellite images indicating poor growth actually have no problem. The reason is simple: users have different operating habits. Some seedlings are planted at bigger intervals. Whereas, they appear sparse on satellite maps. Sometimes, the figures derived from photosynthesis may look normal, but, because smaller sized products are used, the final yield may deviate from the model predicted by Zhang. Therefore, during the adjustment of yield and forecast of algorithm model at the later period, Zhang Gong will ask users about their past yields, planting history, and operating habits so as to combine "the sky and the earth circumstances."
"Although we can monitor the health of leaves through photosynthesis, we are not clear how the final yield will turn out to be. Good health of leaves does not necessarily mean good ultimate harvest. The key agricultural nodes need to be reviewed with ground data," said Zhang.
As precision agriculture is in the bud in China, Gago can only try step by step. "But, at first, we did not know what to do, so we did whatever we could," Zhang Gong said. At the beginning, Gago was eager to provide a whole set of solutions to its customers. Even "the cost of water transmission and the loss of crop failure would be evaluated in certain cases," Zhang added. He also considers the actual execution and price tolerance of customers. "Agriculture cannot be done by just pressing one button." Gago actively helps its customers choose practical functions and attempts to match data with customer business logic. "We promise to help solve their core problems rather than all the problems," he says.
Gago has so far covered two million mu of arable land in China, including analyzing the origins of grain in the north and the fruit plantations of pitaya in the south. Among Its customers is the digital agricultural project of Qipan Village, Jilin Province. Zhang sets prices based on the quantity of modules required by customers according to their needs. In 2016, Gago's revenue amounted to CNY10 million.
Zhang Gong "has an ambitious plan" in the field of precision agriculture. Besides farming management, Gago hopes to provide rural financial services, including subject management for insurance companies and land risk assessment for credit companies. But, in the face of the current state of the domestic scale farming, there is still a long way to go to open up the whole process.