Crop Yield Forecasting in Sugarcane Using Remote Sensing
Brazil leads the world in sugarcane production and sugarcane is one of Brazil’s major crops. With a total annual production value of $23 bn, Brazil produces more than 50% of the world’s sugarcane. Since 1532 sugarcane is one of the first most important crops in Brazil and its contribution to the Brazilian GDP is almost the same as for soybeans. Sugarcane contributed to 2% of the Brazilian GDP in 2014. Brazil’s sugarcane crushing capacity has grown rapidly over the last decade and increased by 61% from 2012. However, difficult pricing conditions and the financial crisis in 2008 forced sugarcane mills to invest less. This has negatively impacted profit margins and has led to a significant consolidation in the industry—over 60 mills have closed during last 9 years. Eventually, this consolidation has had a positive impact on the production efficiency and crushing capacity of the individual sugarcane mills. Currently, approximately 200 industrial players share the sugarcane industry in Brazil.
Sugarcane is a unique multi-year crop, which can be harvested annually up to 6-7 years without replanting. After an annual harvest, the roots and lower parts of the plant remaining in the field, the ratoons, grow new stems that are cut the following year. On most farms, sugarcane is harvested continuously over 9 months. Importantly, this allows crushing to be continuous. Therefore, managing the harvesting and crushing occurs almost daily. Such high throughput production creates a tremendous need to provide regular crop yield estimates. Adding complexity, the daily harvest of cut sugarcane must be efficiently transported to a sugar mill for processing. To ensure that the cut sugarcane is processed as quickly as possible and that the mills have a continuous supply, most sugar mills surround themselves with sugarcane farms for which they coordinate operations. One industrial sugar mill can easily have more than 100 machines, such as trucks, harvesters, and tractors, for harvesting operations. On average, sugarcane mill can crush about 1’000-5’000 tons of raw sugarcane a day. Each sugar mill critically needs information about sugarcane production to manage and allocate their resources.
Existing methods to reliably estimate crop yield in sugarcane
Sugarcane crop yield forecasting is a multi-billion dollar (US) business in Brazil. In addition to sugar mills, there are other stakeholders across the agricultural value chain, who are interested in understanding and forecasting the supply side of the sugarcane market. These include commodities traders, banks and insurance companies, ethanol producers, sugarcane refineries, and large industrial sugarcane growers themselves. The efficient management of the harvesting and crushing processes help to increase the profitability in the Brazilian sugarcane sector, where price is controlled by the government. Therefore, crop yield forecasting has a direct impact on the bottom line of a sugar mill.
How do sugar mills accurately and reliably estimate sugar and biomass yield? Sugar mills have been using different methodologies to estimate sugarcane crop yield since the beginning of commercial cultivation. The most traditional ways to crop yield forecasting are manual scouting and crop sampling. One industrial farm might employ 100-200 field workers to scout and sample sugarcane fields every day. Scouting and sampling are inefficient, time-consuming, expensive, and highly dependent on the experience of field workers and agronomists. Additionally, these methods produce incomplete observations—mature sugarcane can be 5-6 meters tall and densely planted, impeding workers from walking through the fields. Unfortunately, these constraints lead to inaccurate assessments of crop biomass and yield.
Another way to make a sugarcane crop yield prediction is to leverage (Normalized Difference Vegetation Index) NDVI-based methods. NDVI-based methods are relatively new and whereas they address some of the concerns of scouting and sampling, suffer from reduced reliability, accuracy, and scalability. Changing climatic conditions, soil types, varieties and farming practices make it extremely difficult to estimate, accurately, yield for sugarcane over very large areas. Some research suggests that the relationship between NDVI-based metrics and sugarcane yield has R squared values of 0.48-0.53, concluding that “due to limitations associated with NDVI, additional plant indices evaluation would be beneficial”. Most NDVI-based methods use multispectral imaging technology, which has a number of limitations, such as not being able to account for the diversity of the crop, soil and climatic conditions.
Gamaya solution to address crop yield forecasting
At Gamaya, we’ve been working on how to provide accurate, reliable, and scalable biomass and yield estimates for the sugarcane market in Brazil. Yield prediction is a function of biomass, growth stage, nitrogen uptake, precipitation, variety potential, soil type, etc. The only way to address and account for the diversity of sugarcane varieties, soil, climate and farming practices is to build comprehensive, dynamic crop models, that simulate the growth of sugarcane plants throughout the crop season based on multiple variables. The growth development cycle is divided into four phases, based on the sugarcane physiology. Data fed into these crop models include remote sensing, meteorology, agronomy, among other supplementary data. Remote sensing, and particularly hyperspectral imaging, offers a unique and efficient way to obtain spectral information in relation to the biophysical characteristics of the crop. This relation between spectral information and crop biophysical characteristics helps to assess crop growth, chlorophyll content, biomass and yield.
Gamaya’s crop yield prediction maps are available through the Gamaya web platform or via a sugarcane producer’s existing farm management software. The maps are delivered at different stages of sugarcane growth and pair with detailed statistics about different levels of biomass (kg/ha) in the field. These data can be aggregated from fields to domains and fazendas. All fields can be compared, analysed and prioritized for specific agro operations. Resulting maps can be downloaded in Esri Shapefile or Google Earth KML formats and statistics are available in Microsoft Excel or comma-separated values (csv) formats.
Gamaya CaneFit solution offers a unique and efficient way to accurately and reliably forecast biomass and yield throughout the full growing season. It rapidly equips sugar mills with essential information to make key decisions about sugarcane harvesting and crushing. Gamaya’s solution helps mill operators manage their logistics (opening dates, inputs, distribution of harvesting machines, etc.) and allocate resources (labor, machinery, etc.) according to the yield forecasts. Specific use cases of Gamaya’s solution include optimization of harvesting based on maturity; growth monitoring throughout the season to better plan on-field operations; and biomass prediction to maximise yield output. By focusing on Brazilian sugarcane, Gamaya endeavours to impact the largest producers of a globally important crop.