New 'Artificial Synapses' Could Let Supercomputers Mimic the Human Brain

Charles Q. Choi, Live Science Contributor

Charles Q. Choi is a contributing writer for Live Science and Space.com. He covers all things human origins and astronomy as well as physics, animals and general science topics. Charles has a Master of Arts degree from the University of Missouri-Columbia, School of Journalism and a Bachelor of Arts degree from the University of South Florida. Charles has visited every continent on Earth, drinking rancid yak butter tea in Lhasa, snorkeling with sea lions in the Galapagos and even climbing an iceberg in Antarctica.

In this schematic drawing, the synapse on the left is a biological synapse, and the one on the right represents an artificial synapse.

Credit: Wentao Xu et al / Science Advances

Large-scale brain-like machines with human-like abilities to solve problems could become a reality, now that researchers have invented microscopic gadgets that mimic the connections between neurons in the human brain better than any previous devices.

The new research could lead to better robots, self-driving cars, data mining, medical diagnosis, stock-trading analysis and "other smart human-interactive systems and machines in the future," said Tae-Woo Lee, a materials scientistat the Pohang University of Science and Technology in Korea and senior author of the study.

The human brain's enormous computing power stems from its connections. Previous research suggested that the brain has approximately 100 billion neurons and roughly 1 quadrillion (1 million billion) connections wiring these cells together. At each of these connections, or synapses, a neuron typically fires about 10 times per second.

In principle, the human brain can perform about 10 quadrillion operations per second. In comparison, the world's fastest supercomputer, Tianhe-2 in China, is capable of carrying out up to about 55 quadrillion calculations per second, according to the TOP500 project, which ranks the 500 most powerful computers in the world. However, previous research suggests that the human brain consumes only about 20 watts of power, which is barely enough to run a dim light bulb, whereas Tianhe-2 consumes about 17.8 megawatts of power, which is enough to run about 900,000 such light bulbs, TOP500 notes. [7 Clever Technologies Inspired by Nature]

Scientists would like to build computers that mimic the human brain's power and efficiency. "Development of artificial synapses with comparable behaviors of biological ones will be a critical step," Lee told Live Science.

Until now, artificial synapses consumed much more energy than biological synapses do. Previous research suggested that biological synapses consume about 10 femtojoules every time a neuron fires. Now, Lee and his colleagues have created artificial synapses that require only about 1.23 femtojoules per synaptic event, making them the lowest-energy artificial synapses developed yet, they said. (For comparison, a small apple falling about 3.3 feet (1 meter) to Earth would generate about 1 quadrillion femtojoules of kinetic energy.)

This research suggests that the "energy consumption and memory density of artificial brains will ultimately rival, and even exceed, [those of] biological brains in the future," Lee said.

These new artificial synapses are a kind of transistor, or electronic switch. By flicking on and off, they can mimic how a synapse fires.

The researchers fabricated 144 synaptic transistors on a 4-inch (10-centimeter) wafer. At the heart of these devices are wires that are 200 to 300 nanometers (billionths of a meter) wide. (For comparison, the average human hair is about 100,000 nanometers wide.) The small features of the devices help to lower the amount of energy they consume, the researchers said. [5 Amazing Technologies That Are Revolutionizing Biotech]

The new devices are made out of one kind of organic material wrapped around another. These materials help the artificial synapses trap or release electrically charged ions, mimicking how biological synapses work, and how an electric switch can be flicked on or off, the researchers explained.

The artificial synapses mimic the structure of actual human nerve fibers'long shape and flexibility. In principle, the researchers could also arrange these devices in 3D grids, somewhat imitating the human brain, Lee said. However, advances in 3D printing are needed to create such 3D grids of artificial synapses, he added.

The researchers are now working to develop organic nanowires only a few dozen nanometers wide, Lee said. They also think that they can reduce synaptic transistor energy consumption even further by tinkering with the selection and structure of the materials they use, he added.

The scientists detailed their findings online June 17 in the journal Science Advances.

Original on Live Science.

For further information log on website :
http://www.livescience.com/55114-artificial-synapses-could-improve-supercomputers.html

Gluten Triggers Strange Delusions in Woman with Celiac Disease

Sara G. Miller, Staff Writer

Sara is a staff writer for Live Science, covering health. She grew up outside of Philadelphia and studied biology at Hamilton College in upstate New York. When she's not writing, she can be found at the library, checking out a big stack of books. 

Credit: Viktorfischer | Dreamstime

Gluten has been implicated in a number of symptoms related to celiac disease that go beyond the digestive system, including rashes, anemia and headaches. But according to a recent case report, the wheat protein played a role in one woman's severe psychosis.

The 37-year-old woman, whose case was described in the report, was studying for her Ph.D. when she started having delusions. Her symptoms began with a belief that people were talking about her as part of a conspiracy in which friends, family members and strangers were acting out scenes for her in a "game," the doctors who treated the woman wrote in their report, published May 12 in The New England Journal of Medicine.

After making threats against her family, the patient was admitted to a psychiatric hospital and was diagnosed with a psychotic disorder, the doctors wrote. She was prescribed anti-psychotic medications to help control her symptoms, but they did not work very well, according to the report.  [Here's a Giant List of the Strangest Medical Cases We've Covered]

During the woman's stay at the psychiatric hospital and at follow-up appointments after she was released, doctors noticed that she had several vitamin and mineral deficiencies, had lost a lot of weight and also had thyroid problems, according to the report.

These symptoms led doctors to suspect that the woman had celiac disease, said Dr. Alessio Fasano, director of the Center for Celiac Research and Treatment at Massachusetts General Hospital in Boston and one of the doctors who treated the woman. It was at that point that the doctors who wrote the case report got involved, he said.

Dr. Jekyll and Mr. Hyde

The doctors at Massachusetts General Hospital confirmed that the woman had celiac disease, according to the report. However, her delusions led her to believe that the doctors were being "deceitful," and she refused to follow a gluten-free diet, they wrote.

The woman lost her job, became homeless and attempted suicide, the doctors wrote. Eventually, she was rehospitalized at a psychiatric facility, where she was successfully placed on a gluten-free diet, they wrote.

When the woman was on a gluten-free diet, her symptoms improved, Fasano said. She was once again functional and aware of what gluten was doing to her, he said. She knew that being exposed to gluten caused her to lose control of her life, and she wanted people to understand that the gluten was causing this bizarre behavior, he added.

The differences between how the woman behaved on a gluten-free diet and after being exposed to gluten was like "Dr. Jekyll and Mr. Hyde," Fasano said. "This was a bright young lady on her way to [getting] a Ph.D., and all of sudden," something changed and she would do things that were harmful to herself and people around her, he said.

During the time the doctors were working with the woman, she inadvertently consumed gluten on several occasions, Fasano said. When this would happen, she would become completely lost, he said. But when she was gluten-free, she was well aware that she needed to avoid gluten because "she [didn't] want to go to 'that place,'" Fasano said.  

When Fasano last saw the woman, around January 2016, he reported that she was doing very well. She was completely avoiding gluten, and her symptoms had gone away, he said. In fact, the woman was planning to participate in an experiment with her doctors so that they could study what happened to her when she consumed gluten, he said.

The plan was to do the experiment in a very controlled environment so that the patient would not do anything harmful, he said. The experiment would give the doctors the opportunity to study the inflammatory process that potentially caused these symptoms. They also planned to do some brain scans, he said.

But before the doctors could do the experiment, the woman accidentally ate some gluten, Fasano said. Her delusions returned, and she was put in jail after trying to kill her parents, he said.

Collateral damage

The mechanisms linking celiac disease to problems with the brain and nervous system aren't entirely clear, in part because it's very difficult to study such effects in humans, Fasano told Live Science. [Celiac Disease: Symptoms & Treatment]

But scientists do have a general sense of what's going on, he said.

Fasano likened the effects to a battle waged in the intestines: When someone has celiac disease, the immune system views gluten as the enemy, and deploys weapons to fight it, he said. Inflammation can be thought of as the collateral damage of the fight, he said. When the battle takes place in the intestine, people end up with inflammation there, he added.

But sometimes, the immune cells that wage war against gluten in the gut are able to leave the battlefield and go elsewhere. If those immune cells go to the brain, the same collateral damage — inflammation — can occur there, Fasano said. And depending on factors such as the location of the inflammation and the amount of time it has been there, the consequences can vary, he said.

Indeed, celiac disease can manifest itself in many different ways in the brain and nervous system, Fasano said. Complications can range from mild problems, such as short-term memory loss, to severe consequences, such as seizures, he said. But this woman's case was one-of-a-kind, he added.

The psychosis that the woman experienced was an extreme condition and very unique, Fasano said.

Originally published on Live Science.

For further information log on website :
http://www.livescience.com/55166-celiac-disease-gluten-psychosis.html

Cold-setting starch adhesive

Abstract

A cold-setting starch adhesive comprising the mixture of three different types of corn starch having widely different properties, i.e., high amylose corn starch having an amylose content of at least 50%, ordinary corn starch and waxy corn starch, in specific proportions. The adhesive shows excellent properties due to the effective combination of the outstanding properties of the individual types of starch.

Claims

What is claimed is:

1. A cold-setting starch adhesive comprising hot high solid aqueous paste of gelatinized starch containing as adhesive component a mixture of 20 to 80% by weight of unmodified or modified high amylose corn starch having an amylose content of at least 50%, 10 to 70% by weight of unmodified or modified ordinary corn starch, and 1 to 10% by weight of waxy corn starch.  

2. A water-resistant, cold-setting starch adhesive comprising hot high solid aqueous paste of gelatinized starch containing as adhesive component a mixture of 20 to 80% by weight of unmodified or midified high amylose corn starch having an amylose content of at least 50%, 10 to 79% by weight of unmodified or modified ordinary corn starch, and 1 to 10% by weight of waxy corn starch, and formaldehyde resin solution.  

3. An adhesive as set forth in claim 1 or 2, wherein the temperature of said paste is 80° C. to 100° C.  

4. An adhesive as set forth in any one of claims 1 to 3, wherein the solid content of said paste is 20 to 50% by weight.  

5. An adhesive as set forth in claim 1, wherein said modified high amylose corn starch is a product produced by oxidation, etherification, esterification, or like treatment of high amylose corn starch to facilitate its gelatinization.  

6. An adhesive as set forth in claim 1, wherein said processed product of regular corn starch is the product of crosslinking, etherification, or like treatment which alters the viscosity or water retaining capacity of said paste.  

7. An adhesive as set forth in claim 1, which further contains one or more fillers selected from the group consisting of bentonite, clay, calcium carbonate, wood meal, walnut shell flour, and coconut shell flour.  

8. An adhesive as set forth in claim 1, for use in the manufacture of corrugated fibreboard.  

9. An adhesive as set forth in claim 2, wherein said formaldehyde resin solution is selected from the group consisting the solutions of urea-formaldehyde, melamineformaldehyde, resorcinol-formaldehyde and ketone-formaldehyde.  

10. An adhesive as set forth in claim 9, wherein said formaldehyde resin solution comprises a mixture of at least two of said formaldehyde resins. 

11. An adhesive as set forth in claim 9, wherein said formaldehyde resin solution comprises a co-condensation product of at least two formaldehyde resins.  

12. An adhesive as set forth in claim 9, 10 or 11, wherein said formaldehyde resins are modified with a substance selected from the group consisting of acetoguanamine, thiourea, ethylene urea, phenol, cresol, ethylenediamine, and diethylenetriamine.  

13. An adhesive as set forth in any one of claims 2, 9, 10, or 11, wherein the quantity of said formaldehyde resin solution is from 5 to 50% by weight based on the total weight of said starches.  

14. An adhesive as set forth in claim 13, wherein said quantity of said formaldehyde resin solution is from 10 to 25% by weight based on the total weight of said starches. 

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention 

This invention relates to a cold-setting starch adhesive which is suitable for use in the manufacture of corrugated fibreboard. 

2. Description of the Prior Art 

Starch adhesive has been commonly used in the manufacture of corrugated fibreboard. Known starch adhesives generally consist mainly of ungelatinized starch paste. Thus, they are used by Stein-Hall process. That is, after the starch paste is firstly gelatinized to sticky paste by steam heating and treatment with caustic soda solution, it is successively applied to the tips of corrugated medium to be bonded to the liner to form corrugated fibreboard, and subsequent heating of the corrugated fibreboard vaporize the water from the paste to dryness to complete the setting. 

The heat energy consumed by the corrugating operation of Stein-Hall process occupies the major part of the total energy consumption in corrugated fibreboard manufacturing plants. If the heating procedure can be eliminated from the corrugating operation, it will contribute to energy saving in corrugated fibreboard manufacturing plants. It has, therefore, been of great interest to the corrugated fibreboard industry to develop a cold-corrugating process which does not require any heating procedure at the corrugating step of corrugated fibreboard. 

Some hot-melt adhesives obtained as petrochemical product is known as being cold-setting. They, however, have the following disadvantages. The raw materials of the hot-melt adhesives are becoming less available because of the high price of petroleum. Used corrugated fibreboard made by employing a hot-melt adhesive is difficult to recover and reuse. Accordingly, it has been strongly desired to develop a cold-setting corn starch adhesive for use in the manufacture of corrugated fibreboard. 

Certain properties are required for a cold-setting corn starch adhesive which is to be used for bonding in the manufacture of corrugated fibreboard without heating procedure. That is, the paste solution must have high solid content, so that it may contain only a small quantity of water to be vaporized. It must be able to gel rapidly, as it is required to set immediately after the application to the fibreboard surfaces to be bonded together. 

Other properties are also required for an adhesive for use in the manufacture of corrugated fibreboard, namely, during the manufacture of corrugated fibreboard, production speed of as high as possible is desired on a corrugator in normal operation, while it must be slowed down when a small lot of products is manufactured, or at the beginning or the end of the operation. It is, therefore, necessary to employ an adhesive having a constant gluing ability irrespective of the production speed, whether it may be high or low. This requirement is nowadays of increasing importance, since the operation speed on a corrugator is as high as 250 meters per minute.

However, ordinary corn starch adhesives, when they are used as cold-corrugating adhesives, tend to gel so slowly that unsatisfactory setting appears if the operating speed of the corrugator is increased, and even a slight external force can cause the separation of the glued surfaces when, for example, the corrugated fibreboard is cut to a prescribed size. Therefore, when they are used as cold-corrugating adhesives in the process of manufacturing corrugated fibreboard, the efficiency of the process is far inferior as compared when they are used in Stein-Hall process. 

Japanese Laid-Open Patent Specification No. 32570/1981 discloses a cold-corrugating adhesive for corrugated fibreboard having improved glueability at high production speed. This adhesive consists mainly of high amylose corn starch instead of ordinary corn starch which is the main component of all the known starch adhesives. High amylose corn starch is a starch prepared from specialty corn created by selective breeding. It contains at least 50% of amylose, while ordinary corn starch contains only about 25% of amylose. 

Because of its high amylose content, high amylose corn starch has a variety of outstanding properties, for example: 

(a) It is hard to be gelatinized under normal conditions;

(b) it can form high solid paste; 

(c) it is highly susceptible to retrogradation;

(d) it forms strong film; and 

(e) it has high glueability.

Some of these properties of high amylose corn starch make it a very suitable substance which can impart rapid glueability to cold-corrugating adhesive for use in the manufacture of corrugated fibreboard. As already pointed out, paste prepared from cold-corrugating starch adhesive must have high solid concentration when it is used at higher operating speed so that it may contain only a small amount of water to be vaporized, and must be capable of gelling rapidly after the application to the fibreboard surfaces to be bonded together. These requirements are fully satisfied by the properties (b) to (e) of high amylose corn starch. 

The high amylose corn starch adhesive disclosed in Japanese Laid-Open Patent Specification No. 32570/1981 is, thus, partly satisfactory for application at higher operating speed. It, however, retrogrades and dries too rapdily, and can retain only a small amount of water. Consequently, if the operating speed is low, it is likely to solidify on the applicator roll, or on the fluit tip of the corrugating medium before bond it with the liner. Thus the adhesive from high amylose corn starch cannot show satisfactory glueability outside of only a narrow range of operating speeds. The use thereof for application at lower operating speed is likely to result in an increased percentage of defective bond. 

Another problem resides in the quality of the material which is available for making corrugated fibreboard. Recent shortage of pulp resources makes it difficult to obtain good pulp, and necessitates the use of fibreboard material of lower strength for making corrugated fibreboard. Synthetic resins, or other reinforcing agents are often added to the material in order to compensate the deficient strength. Synthetic resins are sometimes added purely for the purpose of producing reinforced liner or medium having improved strength irrespective of the quality of the pulp. The adhesive prepared from high amylose or ordinary corn starch is, however, very unsatisfactory for gluing any such resin-reinforced liner or medium. 

Under these circumstances, it has hitherto been considered impossible to produce from starch a cold-corrugating adhesvie for corrugated fiberboard which shows high glueability both at higher and lower operating speeds. 

SUMMARY OF THE INVENTION

It is an object of this invention to provide a cold-corrugating starch adhesive which is sutiable for corrugated fibreboard. 

It is another object of this invention to provide a cold-corrugating starch adhesive which is satisfactorily applicable both at higher and lower operating speeds, and which does not require heating during application, and can, therefore, contribute to energy saving in the manufacture of corrugated fibreboard. 

It is still another object of this invention to provide a cold-corrugating starch adhesive which shows satisfactory glueability for any type of material employed in the manufacture of corrugated fibreboard. 

It is a further object of this invention to provide a water-resistant, energy-saving adhesive which is satisfactorily applicable both at higher and lower operating speeds, and which does not require heating during application.

DETAILED DESCRIPTION OF THE INVENTION

This invention provides a cold-corrugating starch adhesive which is satisfactorily applicable both at higher and lower operating speeds for making corrugated fibreboard by a corrugator. 

The inventors of this invention have discovered that the combined use of high amylose corn starch, ordinary corn starch and waxy corn starchsurprisingly produces a cold-corrugating starch adhesive suitable for corrugated fibreboard, which is satisfactorily applicable at lower operating speed, while retaining the rapid glueability which is characteristic of high amylose corn starch. 

Thus, this invention provides a cold-corrugating starch adhesive in the form of a hot aqueous high solid paste of gelatinized starch which comprises 20 to 80% by weight of high amylose corn starch having an amylose content of at least 50%, or any modified form thereof, 10 to 79% by weight of ordinary corn starch or any modified form thereof, and 1 to 10% by weight of waxy corn starch (total is 100% by weight). 

If the proprotion of high amylose corn starch is less than 20% by weight, no satisfactory glueability is obtained at higher operating speed, while no satisfactory glueability is obtained at lower speed if less than 10% by weight of ordinary corn starch is employed. 

The most important feature of this invention resides in the fact that the ashesive is endowed with improved glueability at lower operating speed, while retaining high glueability at higher operating speed by the incorporation of waxy corn starch into a mixture of high amylose corn starch and ordinary corn starch. 

Waxy corn starch does not contain amylose, and consists solely of amylopectin. Therefore, it has a number of properties which are different from these of high amylose corn starch. For example, it is (a) easy to be gelatinized, (b) higher sticky, and (c) capable of holding large amount of water. The high glueability of the adhesive of this invention at lower operating speed is due to the presence of waxy corn starch. It provides a starch paste with an improved water-holding capacity during gluing at lower speed and it prevents any undersirable solidification of the paste on the applicator roll, or drying or solidification of the paste before the completion of the gluing speration. 

The adhesive of this invention containing waxy corn starch is satisfactorily applicable also to the aforementioned resin-reinforced liner or medium as opposed to any known adhesive not containing waxy starch. 

If the proportion of waxy corn starch is less than 1% by weight, however, it is difficult to expect the improved glueability of the adhesive at lower gluing speed, while a paste containing more than 10% by weight of waxy corn starch is too stringy to be applied uniformly by rolls. 

It is surprising that the advantages of the three different types of starch in the adhesive of this invention exhibit themselves to the maximum extent without cancelling one another. This is only possible as a result of the combination of the three different types of starch. Such adhesive having high glueability both at higher and lower operating speeds as that of the present invention would not be obtained from the mixture of only high amylose and waxy corn starches, nor from the mixture of high amylose and ordinary corn starches in any proportion. 

A method for preparing the cold-corrugating starch adhesive may comprise adding water to a mixture of high amylose, ordinary and waxy corn starches, and if required, an oxidizing agent such as persulfate or perborate; gelation promotor such as boric acid, borax or a sulfite; pH adjustor such as sodium hydroxide, or the like, mixing them together, and heating the resulting mixture to the temperature of between 80° C. and 100° C. by conventional method, e.g. by using a cooking device such as jet cooker, autoclave or Onlator (indirect heating cooker) to give a hot high solid paste of gelatinized starch with a viscosity of between 500 and 1500 cps and having a solid content of 20-50% by weight. This paste is applied to the fibreboard surfaces to be glued together, and them immediately allowed or forced to cool, whereby the paste sets rapidly to produce strong adhesion. 

The adhesive of this invention may contain, as high amylose corn starch having an amylose content of at least 50%, any modified form thereof obtained by oxidizing, etherifying, esterifying, or otherwise treating the starch in a known manner. Such modification of high amylose corn starchby oxidation, etherification, esterification, or the like makes it easier to form a gelatinized paste, and to prepare a more homogeneous and stable paste. 

The modification of ordinary corn starch by a known method such as crosslinking, etherification or the like is also useful for obtaining an adhesive of improved glueability in the form of a short stringy paste having an improved water-holding capacity. Waxy corn starch may likewise be modified, if required. 

The adhesive of this invention may shrink when it sets. In order to prevent such shrinkage, it is possible to add any known inorganic or organic filler, such as bentonite, clay, calcium carbonate, ground wood powder, walnut shell flour or coconut shell flour, so that a stronger adhesive layer may be obtained. The filler may be added in the amount of 1 to 100%, preferably 2 to 20%, by weight based on the starch, depending on the filler which is employed. The smaller amount less than 1% is not effective, while the addition of greater than 100% is likely to bring about disadvantages, such as lower glueability. 

If water resistance is required for the adhesive of this invention formaldehyde resin solution can be incorporated. It is possible to use urea-formaldehyde resin solution, melamine-formaldehyde resin solution, resorcinol-formaldehyde resin solution, or ketone-formaldehyde resin solution, or a mixture or co-condensation resin solution thereof, or a solution of any such resin modified with acetoguanamine, thiourea, ethylene urea, phenol, cresol, ethylenediamine, diethylenetriamine, or the like. 

Such formaldehyde resin solution is usually employed in the quantity of 5 to 50% by weight based on the total starch amount, considering the effect and the expense of the resin solution. More preferably, it is employed in 10 to 25% by weight. The formaldehyde resin solution, when employed in the prescribed range of quantity, imparts outstanding water resistance to the adhesive of this invention without impairing its glueability at both higher and lower operating speeds. It may be added into the high solid gelatinized paste prepared as hereinabove described, after the pH of the paste has been adjusted to a range of 4 to 7. 

The water-resistant adhesive thus prepared in the form of a hot high solid paste of gelatinized starch is particularly suitable for use in the manufacture of water-resistant corrugated fibreboard. The adhesive is applied to the fibreboard surfaces to be glued together, and after the surfaces have been glued together, the adhesive is immediately allowed or forced to cool, whereupon it sets rapidly, and corrugated fibreboard with strong, water-resistant adhesion is obtained. 

The cold-setting starch adhesive of this invention shows outstanding glueability at both higher and lower operating speeds. As it can accomplish strong adhesion without heating during the corrugating operation, it contributes greatly to reducing the consumption of heat energy in corrugated fibreboard manufacturing plants. It does not cause any appreciable inconvenience in the recovery and reuse of used corrugated fibreboard, as compared to petrochemical hot-melt adhesives. 

Although the adhesive of this invention is particularly useful in the manufacture of corrugated fibreboard, it is, of course, equally applicable to ordinary paper, cloth, wood, plastics, inorganic material, or any other material in general. 

The invention will now be described with reference to examples thereof.

EXAMPLE 1

Sixty parts by weight of high amylose corn starch having an amylose content of 70%, 35 parts by weight of ordinary corn starch, and 5 parts by weight of waxy corn starch were suspended in 186 parts by weight of water. Added into the suspension were 2 parts by weight of sodium persulfate, 0.5 part by weight of sodium sulfite, 2 parts by weight of boric acid and 1 part by weight of sodium hydroxide. Gelatinized paste was formed from the resulting mixture by a continuous paste forming apparatus (Onlator of Sakura Seisakusho, Japan). 50% by weight aqueous solution of sodium hydroxide was added into the paste to adjust the pH value to 9.0, whereby there was obtained a cold-setting starch adhesive in the form of a gelatinized paste having a temperature of about 90° C. and a solid content of about 33% by weight. 

The glueability of the adhesive was tested by applying it to the single-face of the fibreboard by a testing corrugator both at a low operating speed of 4 m/min. and at a high operating speed of 20 m/min. For this test, the hot high solid gelatinized paste was maintained at a temperature of about 90° C., and applied to the flute tip of the corrugating medium at an application rate of about 5 g/m2 (dry weight). After a kraft liner had been placed on the corrugating medium, cold air was blown against the paste to cool and set it, whereby glued corrugated fibreboard was obtained. K liner-280 of Honshu Paper Co., Japan, SCP 125 of Honshu Paper Co. and MM 180 of Honshu Paper Co. were used as the liner, the corrugating medium and the reinforced corrugating medium respectively. 

From this fibreboard, test pieces of 8×5 cm were cut out and tested for adhesion strength by a compression testing machine in accordance with JIS (Japanese Industrial Standards) Z-0402. The results, as well as those of the other examples, are shown in TABLE 1 below.

EXAMPLE 2

An aqueous suspension was prepared from 350 g of high amylose corn starch having an amylose content of 70% in 400 ml of water at 40° C. 3% aqueous solution of sodium hydroxide was added into the suspension to adjust the pH value to 11.5. Into this starch suspension was further added 28 ml of 50% aqueous solution of 3-chloro-2-hydroxypropyltrimethylammonium salt. The resulting solution was stirred for four hours at 40° C. After hydrochloric acid had been added into the solution to adjust the pH value to 6.5, the solid content of the solution was separated and washed with water, and dried, whereby cationic high amylose corn starch was obtained. 

The procedures of EXAMPLE 1 were repeated, except for using 60 parts by weight of the cationic high amylose corn starch, 35 parts by weight of commercially available etherified ordinary corn starch (hydroxyethyl corn starch 5-B of Nippon Shokubai Kagaku Kogyo, Japan), and 5 parts by weight of waxy corn starch, and there was obtained a cold-setting starch adhesive in the form of a gelatinized paste of about 90° C., and a concentration of about 33% by weight. 

The adhesive was used to make corrugated fibreboard in accordance with the procedures described in EXAMPLE 1, and the adhesion strength of the fibreboard was tested as described in EXAMPLE 1. 

COMPARATIVE EXAMPLE 1

The procedures of EXAMPLE 1 were repeated to prepare a hot high solid gelatinized paste, except for using only ordinary corn starch having an amylose content of 24%, and corrugated fibreboard was made from this paste to test its adhesion strength as in EXAMPLE 1. 

COMPARATIVE EXAMPLE 2A

The procedures of EXAMPLE 1 were repeated to prepare a hot high solid gelatinized paste, except for using 100 parts by weight of high amylose corn starch of an amylose content of 70%, and corrugated fibreboard was made from this paste to test its adhesion strength as in EXAMPLE 1. 

COMPARATIVE EXAMPLE 2B

The procedures of COMPARATIVE EXAMPLE 2A were repeated, except for using 60 parts by weight of high amylose corn starch of an amylose content of 70% and 40 parts by weight of ordinary corn starch, while using no waxy corn starch. 

The results of the foregoing examples are shown in TABLES 1 and 2 below. TABLE 1 shows the results of the adhesion strength tests conducted on corrugated fibreboard prepared by employing SCP 125 of Honshu Paper Co. as the corrugating medium, while TABLE 2 shows the test results on corrugated fibreboard prepared by employing MM 180 of Honshu Paper Co. as the reinforced corrugating medium. 

TABLE 1

______________________________________

Adhesion strength (kg) Corrugated fibreboard Corrugated fibreboard obtained at an operating obtained at an operating speed of 4 m/min. speed of 20 m/min.

______________________________________

Example 1 19.3 19.8 (invention) Example 2 19.6 20.4 (invention) Comparative 14.3 Gluing impossible Example 1 Comparative 17.1 19.2 Example 2A Comparative 16.7 18.8 Example 2B

______________________________________

NOTE: The testing machine had a variable speed of 0 to 20 m/min. 

TABLE 2

______________________________________

Adhesion strength (kg) Corrugated fibreboard Corrugated fibreboard obtained at an operating obtained at an operating speed of 4 m/min. speed of 20 m/min.

______________________________________

Example 1 18.1 17.9 (invention) Example 2 19.8 18.4 (invention) Comparative 10.4 Gluing impossible Example 1 Comparative 12.6 14.1 Example 2A Comparative 12.1 13.4 Example 2B

______________________________________

As is obvious from the test results shown in TABLES 1 and 2, the adhesives of this invention showed excellent glueability at both high and low operating speeds. 

On the other hand, the conventional adhesive of COMPARATIVE EXAMPLE 1 comprising only ordinary corn starch was found inapplicable for operating at a high speed. The high amylose corn starch adhesive of COMPARATIVE EXAMPLE 2A was very poor in glueability at a low speed. 

The adhesive of COMPARATIVE EXAMPLE 2B comprising high amylose corn starch and ordinary corn starch was inferior in glueability at both high and low speeds to the adhesive of this invention. The adhesive of COMPARATIVE EXAMPLE 2B was also found to show an undesirable variation in glueability depending on the material from which corrugated fibreboard was formed. 

EXAMPLE 3

A gelatinized paste was prepared in accordance with the procedures in EXAMPLE 1. After 50% by weight aqueous solution of sodium hydroxide had been added to this paste to adjust the pH value to 6.0, 20 parts by weight of a urea resin solution (UW-062 of Hohnen Oil Co., Japan) were added into the paste, and mixed to obtain a cold-setting, water-resistant starch adhesive in the form of a gelatinized paste of a temperature of about 90° C. and a solid content of about 35% by weight. 

The glueability of the adhesive was tested by applying it to the single-face of the fibreboard by a testing corrugator both at a low operating speed of 4 m/min. and at a high operating speed of 20 m/min. For this test, the hot high solid gelatinized paste was maintained at a temperature of about 90° C., and applied to the flute tip of the water-resistant corrugating medium (SSCP 125 of Honshu Paper Co.) at an application rate of about 10 g/cm2 (dry weight). After a water-resistant liner (SK 280 of Honshu Paper Co.) had been placed on the corrugating medium, cold air was blown against the paste to cool and set it, whereby corrugated fibreboard was obtained. 

From this fibreboard, test pieces of 8×5 cm were cut out and after the specimens had been immersed for one hour in water at 20° C., they were tested for water-resistant adhesion strength by a compression testing machine. The results, as well as those of other examples, are shown in TABLE 3 below. 

EXAMPLE 4

Cationic high amylose corn starch was prepared in accordance with the procedures of EXAMPLE 2. A gelatinized paste having a pH value of 7.0 was prepared in accordance with the procedures of EXAMPLE 3 from 60 parts by weight of the cationic high amylose corn starch, 35 parts by weight of commercially available etherified ordinary corn starch (hydroxyethyl corn starch 5-B of Nippon Shokubai Kagaku Kogyo, Japan), and 5 parts by weight of waxy corn starch. To this paste was added 15 parts by weight of a melamine-formaldehyde resin solution ML-044 of Hohnen Oil Co.), and there was obtained a cold-setting, water-resistant starch adhesive in the form of a gelatinized paste of a temperature of about 90° C. and a concentration of about 35% by weight. 

The procedures of EXAMPLE 3 were repeated to made water-resistant corrugated fibreboard from the adhesive thus obtained, and to test its water-resistant adhesion strength.

COMPARATIVE EXAMPLE 3

The procedures of EXAMPLE 3 were repeated for preparing a hot high solid gelatinized paste except for using only ordinary corn starch having an amylose content of 24%, and water-resistant corrugated fibreboard was made from this paste to test its water-resistant adhesion strength. 

COMPARATIVE EXAMPLE 4

The procedures of EXAMPLE 3 were repeated for preparing a hot high solid gelatinized paste except for using only high amylose corn starchhaving an amylose content of 70%, and water-resistant corrugated fibreboard was made from this paste to test its water-resistant adhesion strength. 

COMPARATIVE EXAMPLE 5

The procedures of EXAMPLE 3 were repeated for preparing a hot high solid gelatinized paste except that no urea-formaldehyde resin solution was employed, and water-resistant corrugated fibreboard was made from this paste to test its water-resistant adhesion strength. 

The test results obtained in the foregoing examples are shown in TABLE 3 below.

TABLE 3

______________________________________

Water-resistant adhesion strength (kg) Corrugated fibreboard Corrugated fibreboard obtained at an operating obtained at an operating speed of 4 m/min. speed of 20 m/min.

______________________________________

Example 3 3.85 3.72 (invention) Example 4 4.29 4.18 (invention) Comparative 1.70 Gluing impossible Example 3 Comparative 1.28 3.67 Example 4 Comparative Separated Separated Example 5

______________________________________

NOTE: The testing machine had a variable speed of 0 to 20 m/min. 

As is obvious from the results shown in TABLE 3, the water-resistant, cold-setting starch adhesive of this invention showed excellent water-resistant adhesion both when applied at a high operating speed, and at a low operating speed.

For further details log on website :
http://www.freepatentsonline.com/4374217.html

Estimation of biomass in wheat using random forest regression algorithm and remote sensing data

Published Date

June 2016, Vol.4(3):212–219, doi:10.1016/j.cj.2016.01.008

Open Access, Creative Commons license, Funding information

Title 

Estimation of biomass in wheat using random forest regression algorithm and remote sensing data

• Author 

• Li'ai Wang ^a,
• Xudong Zhou ^b
• Xinkai Zhu ^a
• Zhaodi Dong ^a
• Wenshan Guo ^a,,

• aKey Laboratory of Crop Genetics and Physiology of Jiangsu Province, Yangzhou University, Yangzhou 225009, China
• bInformation Engineering College of Yangzhou University, Yangzhou 225009, China

Received 15 October 2015. Revised 29 January 2016. Accepted 15 March 2016. Available online 30 March 2016.

Abstract

Wheat biomass can be estimated using appropriate spectral vegetation indices. However, the accuracy of estimation should be further improved for on-farm crop management. Previous studies focused on developing vegetation indices, however limited research exists on modeling algorithms. The emerging Random Forest (RF) machine-learning algorithm is regarded as one of the most precise prediction methods for regression modeling. The objectives of this study were to (1) investigate the applicability of the RF regression algorithm for remotely estimating wheat biomass, (2) test the performance of the RF regression model, and (3) compare the performance of the RF algorithm with support vector regression (SVR) and artificial neural network (ANN) machine-learning algorithms for wheat biomass estimation. Single HJ-CCD images of wheat from test sites in Jiangsu province were obtained during the jointing, booting, and anthesis stages of growth. Fifteen vegetation indices were calculated based on these images. In-situ wheat above-ground dry biomass was measured during the HJ-CCD data acquisition. The results showed that the RF model produced more accurate estimates of wheat biomass than the SVR and ANN models at each stage, and its robustness is as good as SVR but better than ANN. The RF algorithm provides a useful exploratory and predictive tool for estimating wheat biomass on a large scale in Southern China.

Keywords

Above-ground dry biomass

Triticum aestivum

Vegetation indices

Wheat

---

1 Introduction

Biomass is one of the most useful indicators of crops vegetation development and health. Measuring biomass directly is a destructive and expensive procedure. More recent estimates are based on remotely sensed data, such as vegetation indices (VIs) [1], [2], [3] and [4]. Kross et al. [1] established relationships between corn biomass and VIs such as the NDVI (Normalized Difference Vegetation Index), Green-NDVI, RVI (Ratio Vegetation Index), and MTVI2 (Modified Triangular Vegetation Index 2) computed from the SPOT and Landsat images. Gnyp et al. [3] found that SAVI (Soil-Adjusted Vegetation Index), OSAVI (Optimized Soil-Adjusted Vegetation Index), and MTVI2 had stronger relationships with rice biomass at the jointing stage than that at booting. Gao et al. [4] proposed that maize biomass could be estimated by VIs calculated using Chinese environmental satellite (HJ) images [e.g. NDVI, RVI, and the enhanced vegetation index (EVI)]. Jin et al. [5] reported that the estimation accuracy of wheat biomass was better using a combination of VIs and radar polarimetric parameters (RPPs) than using VIs or RPPs alone.

Remote estimation of biomass requires application of diverse methods and techniques. In recent years machine-learning algorithms were trialed for ability to perform flexible input–output nonlinear mappings between remotely sensed data and biomass [6], [7] and [8]. Typically, artificial neural networks (ANNs) and support vector regressions (SVRs) were employed to couple with VIs to build monitoring models with improved prediction accuracy of remote estimation of biomass in crops. For instance, Wang et al. [9] provided an effective model for assessing the biomass of wheat with ANNs and VIs (i.e. RVI, NDVI, GNDVI, SAVI, OSAVI, RDVI) calculated based on ASD FieldSpec data. Clevers et al. [10] estimated grassland biomass using SVRs and VIs such as the RVI, NDVI, WDVI, SAVI, GEMI (Global Environmental Monitoring Index), and EVI (Enhanced Vegetation Index) calculated based on ASD FieldSpec data.

Among various machine-learning algorithms, the emerging Random Forest (RF) algorithm proposed by Leo Breiman and Cutler Adele in 2001 has been regarded as one of the most precise prediction methods for classification and regression, as it can model complex interactions among input variables and is relatively robust in regard to outliers. The RF algorithm presents several advantages; it runs efficiently on large datasets, it is not sensitive to noise or over-fitting [11], it can handle thousands of input variables without variable deletion, and it has fewer parameters compared with that of other machine-learning algorithms (e.g. ANN or SVR). The RF classification algorithm has been applied to many remote sensing domains such as land cover classification [12], [13] and [14] and other fields related to the environment and water resources [15] and [16]. To our knowledge, only a few studies have reported the use of the RF regression algorithm in remote sensing applications, including monitoring of forest growth, wetland vegetation, and water resources [6], [17] and [18]. Furthermore, few studies have employed the RF regression algorithm based on VIs for estimating the biomass of winter wheat.

The major objectives of this study were to: (i) investigate the applicability of the RF regression algorithm in combination with VIs to remotely estimate wheat biomass, (ii) test the performance of RF regression for estimating biomass, and (iii) compare the performance of RF with that of other machine-learning algorithms for the estimation of wheat biomass. Specifically, based on VIs calculated from China's environmental satellite (HJ) charge-coupled device (CCD) images, we employed the RF algorithm to construct models to estimate wheat biomass, and then, the RF algorithm was compared with the SVR and ANN machine-learning algorithms in terms of accuracy, goodness of fit, and robustness for estimating wheat biomass.

2 Data source

2.1 Experimental design and data acquisition

Experiments were carried out in four counties (YiZheng, JiangYan, GaoYou and TaiXing) of Jiangsu province during the winter wheat growing seasons of 2010, 2011, 2012 and 2014. The local wheat cultivars were Yangmai 13, Yangmai 15, Yangmai 16, and Yangfumai 2. For each year's experiment, fifteen sample sites were established in each county and a plot of 30 × 30 m was randomly demarcated at each site. Within each plot, five subplots of 0.4 m × 0.4 m were established at least 10 m from each other. During three growth stages (jointing, booting and anthesis) wheat plants from each subplot (positions determined with a Global Positioning System GPS, Trimble GeoExplorer 2008 Series GeoXH, Trimble Navigation Limited, USA) at each site were collected, sealed in plastic bags, and sent to a laboratory for analysis. In the laboratory, the wheat plants from each subplot were dried in an oven at 80 °C for 48 h, after which the dry weight was determined. The dry weight was divided by the surface area of the subplot, and then the weight was converted to kg ha^{− 1}. The biomass values of plants from the five subplots within each plot were averaged to represent the biomass of the entire plot.

For each stage, the pooled data from 2010, 2011, 2012 and 2014 were randomly divided into a training dataset and an independent test dataset (75% and 25% of the pooled data, respectively). For the training dataset, the number of samples was 174 at jointing, 174 at booting, and 147 at anthesis. For the test dataset, the number of samples was 58 at jointing, 58 at booting, and 49 at anthesis. The training dataset was used to establish models to predict biomass during each growth stage, and the test dataset was used to test the quality and reliability of each prediction model.

2.2 Remote sensing data and preprocessing

Remotely sensed data (HJ satellite charge-coupled device) of wheat from the three stages were retrieved online from the China Centre for Resources Satellite Data and Application (CRESDA). The HJ satellite charge-coupled device (HJ-CCD) satellite system is China's environmental disaster and environmental monitoring satellite system. It includes two optical satellites, HJ-1A and HJ-1B, which are symmetrically equipped with two CCD cameras. They comprise four multispectral bands with a 30-m resolution and a 720-km swath. The spectral ranges of the four bands are 430–520 nm (B₁-blue), 520–600 nm (B₂-green), 630–690 nm (B₃-red) and 760–900 nm (B₄-near infrared).

All HJ-CCD image data used in this study were completely corrected using ENVI4.7 remote sensing image processing software. Ground control points were located with a differential GPS unit during the field experiments. The map projection used a geographic coordinate system (Lat/Lon) as the projection type (WGS84) and a pixel size of 30 m × 30 m. A radiometric calibration was conducted using the HJ satellite calibration coefficients (e.g. gains and offsets). Atmospheric corrections were conducted using the MOTRAN 4 model embedded in the ENVI/FLAASH module of ENVI 4.7 software, and the input parameters were set based on the location, sensor type and ground weather conditions observed on the day each image was acquired. To improve the accuracy of pixel registration to within one pixel, coarse geometric corrections were made based on the 1:10,000 digitized raster map, after which, precise geometric corrections were made based on the GPS ground control points.

2.3 Vegetation indices

Vegetation indices (VIs) are usually used to quantify crop biomass. This study examined 15 VIs (Table 1) reported in literature to be well correlated with biomass. These VIs were calculated based on the four HJ-CCD bands.

Table 1. Formulas of remote sensing vegetation indices.

Acronym	Index	Formula	Reference
NDVI	Normalized Difference Vegetation Index	(RNIR–RR)/(RNIR + RR)	[19]
SAVI	Soil-Adjusted Vegetation Index	(RNIR–RR) / (RNIR + RR + 0.5) × 1.5	[20]
OSAVI	Optimized Soil-Adjusted Vegetation Index	(RNIR–RR) / (RNIR + RR + 1.6) × 1.16	[21]
NRI	Nitrogen Reflectance Index	(RG–RR) / (RG + RR)	[22]
GNDVI	Green-NDVI	(RNIR–RG) / (RNIR + RG)	[23]
SIPI	Structure Insensitive Pigment Index	(RNIR–RB) / (RNIR + RB)	[24]
PSRI	Plant Senescence Reflectance Index	(RR–RB) / RNIR	[25]
RVI	Ratio Vegetation Index	RNIR / RR	[26]
CRI	Carotenoid Reflectance Index	1/RG + 1/RNIR	[27]
EVI	Enhanced Vegetation Index	2.5 × (RNIR – RR) / (1 + RNIR + 6 × RR – 7.5 × RB)	[28]
MSR	Modified Simple Ratio Index	((RNIR / RR) – 1) /	[29]
NLI	Nonlinear Vegetation Index	(RNIR × RNIR – RR) / (RNIR × RNIR + RR)	[30]
RDVI	Re-normalized Difference Vegetation Index	(RNIR – RR) /	[31]
TVI	Transformational Vegetation Index		[32]
MTVI2	Modified Triangular Vegetation Index 2	1.5 × [1.2 × (RNIR – RG) – 2.5 × (RR – RG)] /	[33]

R_i denotes reflectance at band i (nanometer); R_B represents reflectance of the blue band of HJ-CCD; R_G represents reflectance of the green band of HJ-CCD; R_R represents reflectance of the red band of HJ-CCD; R_NIR represents reflectance of near infrared band of HJ-CCD.

3 Models and statistics

Based on the vegetation indices (VIs) in Table 1, RF, SVR and ANN were respectively used to remotely estimate wheat biomass during each growth stage. In each model, the vegetation indices were considered to be independent variables and biomass was the dependent variable.

3.1 Random forest regression algorithm (RF)

The RF regression algorithm is an ensemble-learning algorithm that combines a large set of regression trees. A regression tree represents a set of conditions or restrictions that are hierarchically organized and successively applied from a root to a leaf of the tree [34], [35] and [36]. The RF begins with many bootstrap samples that are drawn randomly with replacement from the original training dataset. A regression tree is fitted to each of the bootstrap samples. For each node per tree, a small set of input variables selected from the total set is randomly considered for binary partitioning. The regression tree splitting criterion is based on choosing the input variable with the lowest Gini Index, i.e.  , where f(t_X(xi), j) is the proportion of samples with the value x_i belonging to leave j as node t [36]. The predicted value of an observation is calculated by averaging over all the trees. Two parameters need to be optimized in the RF: the number of regression trees (ntree; default value is 500 trees) and the number of input variables per node (mtry; default value is 1/3 of the total number of variables).

To model the relationship between VIs and wheat biomass in this study, given the set of training input–output (i.e. VIs–biomass) pairs, the RF regression model was conducted as follows:

• 1)
  ntree bootstrap sample sets, i.e. X_i (i = bootstrap iteration, and its value was limited to the range of [1, ntree]), were randomly drawn with replacement from the original training dataset. The elements not included in X_i are referred to as out-of-bag data (OOB) for that bootstrap sample set.
• 2)
  At each node per tree, mtry vegetation indices were randomly selected from all 15 vegetation indices and the best split from among those indices was chosen according the lowest Gini Index.
• 3)
  For each tree, the data splitting process in each internal node of a rule was repeated from the root node until a previously specified stop condition was reached.

For the three stages, the parameter values (ntree and mtry) were optimized using the training dataset and RMSE to find the values that could best predict the wheat biomass. For each stage, ntree values from 1000 to 9000 with intervals of length 1000 were tested [37], [38], [39], [40] and [41], and mtry was tested from 3 to 10 (Fig. 1). The ntree and mtry values that yielded the lowest RMSE were selected. According to Fig. 1, the values of ntree and mtry were 1000 and 3 at jointing and booting, respectively, and 3000 and 9 at anthesis.

Fig. 1. Optimization of random forest parameters (ntree and mtry) using RMSE.

3.2 Support vector regression (SVR)

The Support Vector Machine (SVM) was originally used for classification problems, i.e. support vector classification (SVC) and was then extended for use with regression problems, i.e. namely support vector regression (SVR) [42]. The quality of the SVR models depends on a proper setting of the SVR meta-parameters, the loss function ε and the error penalty factor C. In addition, selection of the kernel function has an important impact on the final models. The commonly used radial basis kernel function (RBF), i.e. K(x, x^′) = exp (−| | x − x^′2/σ²) was applied in this study. Finally, we employed a cross-validation procedure to optimize these parameters including C, ε, and the RBF kernel parameter σ, yielding values of 30, 470 and 2.5 at jointing, 5, 400 and 1.1 at booting, and 5, 850 and 8 at anthesis, respectively.

3.3 Artificial neural network (ANN)

Among various machine-learning algorithms, artificial neural networks (ANNs) are the most common approaches to develop nonlinear regression [43]. Training an ANN needs selections including the network structure (i.e. the number of hidden layers and nodes per layer), proper initialization of the weights, learning rate, and training algorithm. In this work, the input layer was vegetation indices, and the output layer was wheat biomass. We optimized a two-layer back propagation neural network (BPNN) with tan-sigmoid (i.e.  ) hidden neurons and log-sigmoid (i.e. ) output neurons using the Levenberg–Marquardt algorithm. The ANN weights were initialized randomly according to the Nguyen–Widrow method [44]. Meanwhile, a cross-validation procedure was employed to set the number nodes per layer (i.e. 67 at jointing and booting, and 49 at anthesis, respectively).

3.4 Statistical analysis

Regarding model performances in this study, we used the coefficient of determination (R²) to account for goodness-of-fit, and the root mean square error (RMSE) and relative RMSE (%) to assess accuracy. The relative RMSE was used to compare performances across different machine-learning algorithms [44]. Generally, the performance of the model was estimated by comparing the differences in R² and RMSE of the estimated-versus-measured value plots. Higher R² and lower RMSE values, respectively, corresponded to higher precision and accuracy of a model for predicting wheat biomass.

4 Results

4.1 Evaluating model accuracy

Using the R² and RMSE values as metrics the performances of all models at each stage were evaluated with the test data from the corresponding stages and compared to identify the best model. For each stage, the R² and RMSE values between estimated (using the RF, SVR or ANN model) and measured biomass values were compared by means of scatter plots (Fig. 2). The performance of the RF model shows an overall improvement compared to that of the SVR and ANN models. Compared with SVR, the RMSE of the RF model decreased to 32.3 kg ha^{− 1} at jointing, 296.1 kg ha^{− 1} at booting, and 366.0 kg ha^{− 1} at anthesis, and the corresponding R² values increased to 0.067, 0.210 and 0.173; compared with ANN, the RMSE decreased to 293.6, 1490.9 and 1215.1 kg ha^{− 1} at each stage, and R² increased by 0.233, 0.287 and 0.297, respectively.

Fig. 2. One-to-one relationships between predicted and observed biomass values.

4.2 Evaluating model robustness

Relative RMSE results of the three regression methods for biomass at different growth stages are presented in Fig. 3. The error bars provide an idea of model robustness with respect to the input data. Different stages hardly impact the RF and SVR models performance in training or testing datasets. For each stage, the relative RMSE for the RF and SVR models respectively stabilize around 8% in the training and 20% in testing datasets. Regarding ANN, the performance of the training dataset was also robust at all three stages with the relative RMSE about 4%, but it performed unstably when applied to testing dataset. Specifically, the relative RMSE is about 35% at jointing, about 45% at booting, and about 30% at anthesis. For each model at each stage, the performance in testing is poorer than in the training dataset. ANN, in particular, showed a much better performance in training than in testing. Hence, in further analysis it will be important to determine how accurately a trained model performs when tested against ground reference measurements rather than the training data [44].

Fig. 3. Relative RMSE (%) results for biomass estimation using RF, SVR, and ANN at different growth stages.

5 Discussion

The objective of this study was to employ accurate and robust random forest (RF) machine-learning algorithms to accurately estimate wheat biomass. Previous studies already used machine-learning algorithms such as SVR or ANN for remote estimations of biomass [6], [7], [8], [9] and [10]. It remains however to be questioned whether these are the most adequate algorithms to fulfill the requirement. This study compared RF with SVR and ANN for accuracy and robustness.

By analyzing the estimated-versus-measured values (Fig. 2) the RF model had higher R² and lower RMSE values than the SVR and ANN models for biomass estimates at each growth stage, indicating that RF models can provide accurate biomass estimations. Each node of the standard regression tree is created using the best split among all variables. Unlike this strategy, RF splits each node using the best among a subset of variables chosen randomly at the node. The specific size of the subset is the parameter mtry. Although this method seems to be contradictory, it performs relatively well compared to SVR and ANN (Fig. 2).

RF rendered similar robustness with SVR at different growth stages in both the training and testing datasets (Fig. 3), and shows better robustness than ANN at each stage. Meanwhile, the RF model for each stage has a little better generalization capability than the ANN model, which behaves relatively unpredictable when used with independent input data that deviate from what was presented during the training stage [44] and [45]. Compared with the RF and SVR results for each stage, ANN shows much poorer performance in testing than in training. This is due to the fact that ANN is often applied to large amounts of sampling data, but SVR and RF are suitable for small amounts of sampling data. Another reason for this is possibly that the learning ability is too strong during the ANN process training, and thus the model obtained cannot reflect the hidden rules of samples that ultimately weaken prediction ability.

Most of the 15 vegetation indices in this study are correlated. However, as demonstrated by Cutler et al. [46] RF is not sensitive to collinearity. This is very valuable in modeling, especially for a complex, nonlinear system because it is commonly difficult to decide which variable to remove when two (or more) variables correlate with each other [47].

For estimation models of vegetation biochemical and biophysical variables to be useful in guiding on-farm crop management, they must perform well in farmers' fields. Therefore, data that fully represents real farm conditions should be included in model training and testing. Data in many previous studies were based on designated experimental sites rather than farmers' fields [48], [49] and [50]. In the present study, we pooled data from farmers' fields in 2010, 2011, 2012 and 2014, and then randomly divided it into a training dataset and an independent testing dataset (75% and 25%, respectively).

A single vegetation index was usually selected in previous studies, to remotely estimate biomass in crops [51] and [52]. However, a single vegetation index is influenced by different degrees of saturability or soil background, and is consequently affected by regional specificity and timeliness [53]. This study shows that use of a combination of 15 vegetation indices and the RF regression algorithm improved the accuracy of prediction of wheat biomass. We propose for the first time use of RF regressions for remote monitoring of biomass, but the prediction accuracy of the method should be further investigated by optimizing the modeling algorithms.

Previous studies of crop growth monitoring based on remotely sensed data have often used a single algorithm to monitor different growing parameters at different growth stages [54] and [55]. In this work, we used RF to estimate wheat biomass on a much larger scale, assuming that it would help to improve wheat growth monitoring in the study areas. It would be interesting to apply the method to monitor other crop growth parameters with different features to verify reproducibility. This research contributes to the establishment of management strategies for non-destructive monitoring and precise modeling methods.

6 Conclusion

Biomass is an important indicator of crop growth. To estimate biomass in wheat rapidly and non-destructively, an improved method that combines vegetation indices based on HJ-CCD and random forest (RF) regression method is proposed. Estimation accuracy and robustness of the RF model were verified for each stage (i.e. jointing, booting, and anthesis). Furthermore, the RF model results were compared with support vector regression (SVR) and artificial neural network (ANN) models. The estimation accuracy of RF outperformed that of SVR and ANN at each stage. For RF models, the R² values for the estimated-versus-measured biomass regression for the three stages were 0.533, 0.721 and 0.79, respectively, and the corresponding RMSE values were 477, 1126.2 and 1808.2 kg ha^{− 1}. The RF model was as robust as SVR and more robust than ANN. The relative RMSE values obtained from the RF and SVR models were about 8% in training and 20% in testing for each stage, respectively. The relative RMSE of ANN was about 4% in training at each stage, whereas the RMSE values in testing were about 35% at jointing, 45% at booting, and 30% at anthesis.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (No. 31271642), the Natural Science Foundation of Education Department of Jiangsu Province (No. 09KJB20013, No. 12KJB520018), the Six Talent Summit Project of Jiangsu Province (No. 2011-NY039), and the Science and Technology Innovation Development Foundation of Yangzhou University (No. 2015CXJ022).

References

1. • [1]
   • A. Kross, H. McNairn, D. Lapen, M. Sunohara, C. Champagne
   • Assessment of RapidEye vegetation indices for estimation of leaf area index and biomass in corn and soybean crops
   • Int. J. Appl. Earth Obs. Geoinf., Volume 34, 2015, pp. 235–248
   • Article
      | 
      PDF (4537 K)
      | 
     View Record in Scopus
     Citing articles (14)
2. • [2]
   • Y.Y. Fu, G.J. Yang, J.H. Wang, X.Y. Song, H.K. Feng
   • Winter wheat biomass estimation based on spectral indices, band depth analysis and partial least squares regression using hyperspectral measurements
   • Comput. Electron. Argic., Volume 100, 2014, pp. 51–59
   • Article
      | 
      PDF (1050 K)
      | 
     View Record in Scopus
     Citing articles (11)
3. • [3]
   • M.L. Gnyp, Y.X. Miao, F. Yuan, S.L. Ustin, K. Yu, Y.K. Yao, S.Y. Huang, G. Bareth
   • Hyperspectral canopy sensing of paddy rice aboveground biomass at different growth stages
   • Field Crops Res., Volume 155, 2014, pp. 42–55
   • Article
      | 
      PDF (3413 K)
      | 
     View Record in Scopus
     Citing articles (24)
4. • [4]
   • S. Gao, Z. Niu, N. Huang, X.H. Hou
   • Estimating the leaf area index, height and biomass of maize using HJ-1 and RADARSAT-2
   • Int. J. Appl. Earth Obs. Geoinf., Volume 24, 2013, pp. 1–8
   • Article
      | 
      PDF (861 K)
      | 
     View Record in Scopus
      | 
     Full Text via CrossRef
     Citing articles (23)
5. • [5]
   • X.L. Jin, G.J. Yang, X.G. Xu, H. Yang, H.K. Feng, Z.H. Li, J.X. Shen, C.J. Zhao, Y.B. Lan
   • Combined multi-temporal optical and radar parameters for estimating LAI and biomass in winter wheat using HJ and RADARSAR-2 data
   • Remote Sens., Volume 7, 2015, pp. 13251–13272
   • View Record in Scopus
      | 
     Full Text via CrossRef
6. • [6]
   • C.J. Gleason, J. Im
   • Forest biomass estimation from airborne LiDAR data using machine learning approaches
   • Remote Sens. Environ., Volume 125, 2012, pp. 80–91
   • Article
      | 
      PDF (1294 K)
      | 
     View Record in Scopus
     Citing articles (45)
7. • [7]
   • J.M. Montes, F. Technow, B.S. Dhillon, F. Mauch, A.E. Melchinger
   • High-throughput non-destructive biomass determination during early plant development in maize under field conditions
   • Field Crops Res., Volume 121, 2011, pp. 268–273
   • Article
      | 
      PDF (178 K)
      | 
     View Record in Scopus
     Citing articles (41)
8. • [8]
   • R. Prasad, A. Pandey, K.P. Singh, V.P. Singh, R.K. Mishra, D. Singh
   • Retrieval of spinach crop parameters by microwave remote sensing with back propagation artificial neural networks: a comparison of different transfer functions
   • Adv. Space Res., Volume 50, 2012, pp. 363–370
   • Article
      | 
      PDF (1059 K)
      | 
     View Record in Scopus
     Citing articles (9)
9. • [9]
   • D.C. Wang, J.H. Wang, N. Jin, Q. Wang, C.J. Li, J.F. Huang, Y. Wang, F. Huang
   • ANN-based wheat biomass estimation using canopy hyperspectral vegetation indices
   • Trans. CSAE, Volume 24, 2008, pp. 196–201 (in Chinese with English abstract)
   • View Record in Scopus
     Citing articles (10)
10. • [10]
    • J.G.P.W. Clevers, G.W.A.M. van der Heijden, S. Verzakov, M.E. Schaepman
    • Estimating grassland biomass using SVM band shaving of hyperspectral data
    • Photogramm. Eng. Remote. Sens., Volume 73, 2007, pp. 1141–1148
    • View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (27)
11. • [11]
    • X.L. Jin, W.Y. Diao, C.H. Xiao, F.Y. Wang, B. Chen, K.R. Wang, S.K. Li
    • Estimation of wheat agronomic parameters using new spectral indices
    • PLoS One, Volume 8, 2013, e72736
    • Full Text via CrossRef
12. • [12]
    • R. Jhonnerie, V.P. Siregar, B. Nababan, L.B. Prasetyo, S. Wouthuyzen
    • Random forest classification for mangrove land cover mapping using Landsat5 TM and ALOS PALSAR imageries
    • Procedia Environ. Sci., Volume 24, 2015, pp. 215–221
    • Article
       | 
       PDF (1291 K)
       | 
      View Record in Scopus
      Citing articles (1)
13. • [13]
    • I. Nitze, B. Barrett, F. Cawkwell
    • Temporal optimization of image acquisition for land cover classification with random forest and MODIS time-series
    • Int. J. Appl. Earth Obs. Geoinf., Volume 34, 2015, pp. 136–146
    • Article
       | 
       PDF (5209 K)
       | 
      View Record in Scopus
      Citing articles (8)
14. • [14]
    • P.O. Gislason, J.A. Benediktsson, J.R. Sveinsson
    • Random forests for land cover classification
    • Pattern Recogn. Lett., Volume 27, 2006, pp. 294–300
    • Article
       | 
       PDF (202 K)
       | 
      View Record in Scopus
      Citing articles (349)
15. • [15]
    • A. Puissant, S. Rougier, A. Stumpf
    • Object-oriented mapping of urban trees using random forest classifiers
    • Int. J. Appl. Earth Obs. Geoinf., Volume 26, 2014, pp. 235–245
    • Article
       | 
       PDF (4869 K)
       | 
      View Record in Scopus
      Citing articles (12)
16. • [16]
    • D.J. Booker, T.H. Snelder
    • Comparing methods for estimating flow duration curves at ungauged sites
    • J. Hydrol., Volume 434, 2012, pp. 78–84
17. • [17]
    • M.L. Liu, X.N. Liu, D. Liu, C. Ding, J.L. Jiang
    • Multivariable integration method for estimating sea surface salinity in coastal waters from in situ data and remotely sensed data using random forest algorithm
    • Comput. Geosci., Volume 75, 2015, pp. 44–56
    • Article
       | 
       PDF (8254 K)
       | 
      View Record in Scopus
      Citing articles (4)
18. • [18]
    • O. Mutanga, E. Adam, M. Azong Cho
    • High density biomass estimation for wetland vegetation using WorldView-2 imagery and random forest regression algorithm
    • Int. J. Appl. Earth Obs. Geoinf., Volume 18, 2014, pp. 399–406
19. • [19]
    • J.W. Rouse, R.H. Haas, J.A. Schell
    • Monitoring the Vernal Advancement of Retrogradation (Green Wave Effect) of Natural Vegetation
    • 1974, Remote Sensing Center, Texas A&M University College Station, USA
20. • [20]
    • A.R. Huete
    • A soil vegetation adjusted index (SAVI)
    • Remote Sens. Environ., Volume 25, 1998, pp. 295–309
21. • [21]
    • G. Rondeaux, M. Steven, F. Baret
    • Optimization of soil-adjusted vegetation indices
    • Remote Sens. Environ., Volume 55, 1996, pp. 95–107
    • Article
       | 
       PDF (1077 K)
       | 
      View Record in Scopus
      Citing articles (532)
22. • [22]
    • T.D. Schleicher, W.C. Bausch, J.A. Delgado, P.D. Ayers
    • Evaluation and refinement of the nitrogen reflectance index (NRI) for site-specific fertilizer management, in: 2001 ASABE Annual International Meeting, St-Joseph, MI, USA. ASABE Paper No. 01–11151
    • 2001
23. • [23]
    • A.A. Gitelson, Y. Kaufman, M.N. Merzlyak
    • Use of a green channel in remote sensing of global vegetation from EOS-MODIS
    • Remote Sens. Environ., Volume 58, 1996, pp. 289–298
    • Article
       | 
       PDF (994 K)
       | 
      View Record in Scopus
      Citing articles (508)
24. • [24]
    • J. Penuelas, I. Filella, J.A. Gamon
    • Assessment of photosynthetic radiation-use efficiency with spectral reflectance
    • New Phytol., Volume 131, 1995, pp. 291–296
    • View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (291)
25. • [25]
    • M.N. Merzlyak, A.A. Gitelson, O.B. Chivkunova, Y.R. Rakitin
    • Non-destructive optical detection of pigment changes during leaf senescence and fruit ripening
    • Physiol. Plant., Volume 106, 1999, pp. 135–141
    • View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (262)
26. • [26]
    • F. Baret, G. Guyot
    • Potentials and limits of vegetation indices for LAI and APAR assessment
    • Remote Sens. Environ., Volume 35, 1991, pp. 161–173
    • Article
       | 
       PDF (1053 K)
       | 
      View Record in Scopus
      Citing articles (895)
27. • [27]
    • A.A. Gitelson, Y. Zur, O.B. Chivkunova, M.N. Merzlyak
    • Assessing carotenoid content in plant leaves with reflectance spectroscopy
    • Photochem. Photobiol., Volume 75, 2002, pp. 272–281
    • View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (172)
28. • [28]
    • H.Q. Liu, A.R. Huete
    • A feedback based modification of the NDVI to minimize canopy background and atmospheric noise
    • IEEE Trans. Geosci. Remote Sens., Volume 33, 1995, pp. 457–465
    • View Record in Scopus
      Citing articles (246)
29. • [29]
    • J.M. Chen
    • Evaluation of vegetation indices and modified simple ratio for boreal applications
    • Can. J. Remote. Sens., Volume 22, 1996, pp. 229–242
    • View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (243)
30. • [30]
    • N.S. Goel, W.H. Qin
    • Influences of canopy architecture on relationships between various vegetation indexes and LAI and FPAR: a computer simulation
    • Remote Sens. Environ., Volume 10, 1994, pp. 309–347
    • View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (77)
31. • [31]
    • K. Wang, Z.Q. Shen, R.C. Wang
    • Effects of nitrogen nutrition on the spectral reflectance characteristics of rice leaf and canopy
    • J. Zhejiang Agric. Univ., Volume 24, 1998, pp. 93–97
    • View Record in Scopus
      Citing articles (12)
32. • [32]
    • N.H. Broge, E. Leblanc
    • Comparing prediction power and stability of broad band and hyperspectral vegetation indices for estimation of green leaf area index and canopy chlorophyll density
    • Remote Sens. Environ., Volume 76, 2000, pp. 156–172
    • View Record in Scopus
      Citing articles (9)
33. • [33]
    • D. Haboudane, J.R. Miller, E. Pattey, P.J. Zarco-Tejada, I.B. Strachan
    • Hyperspectral vegetation indices and novel algorithms for predicting green LAI of crop canopies: modeling and validation in the context of precision agriculture
    • Remote Sens. Environ., Volume 90, 2004, pp. 337–352
    • Article
       | 
       PDF (1140 K)
       | 
      View Record in Scopus
      Citing articles (604)
34. • [34]
    • L. Breiman, J. Friedman, R.A. Olshen, C.J. Stone
    • Classification and Regression Trees, The Wadsworth Statistics/Probability Series
    • 1984, Wadsworth, Belmont, CA
35. • [35]
    • J.R. Quinlan
    • C4.5 Programs for Machine Learning
    • 1993, Morgan Kaurmann, San Mateo, CA
36. • [36]
    • V. Rodriguez-Galiano, M.P. Mendes, M.J. Garcia-Soldado, M. Chica-Olmo, L. Ribeiro
    • Predictive modeling of groundwater nitrate pollution using random forest and multisource variables related to intrinsic and specific vulnerability: a case study in an agricultural setting (southern Spain)
    • Sci. Total Environ., Volume 476–477, 2014, pp. 189–206
    • Article
       | 
       PDF (2460 K)
       | 
      View Record in Scopus
      Citing articles (16)
37. • [37]
    • A. Prasad, L. Iverson, A. Liaw
    • Newer classification and regression tree techniques bagging and random forests for ecological prediction
    • Ecosystems, Volume 9, 2006, pp. 181–199
    • View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (543)
38. • [38]
    • A. Huete, K. Didan, T. Miura, E.P. Rodriguez, X. Gao, L.G. Ferreira
    • Overview of the radiometric and biophysical performance of the MODIS vegetation indices
    • Remote Sens. Environ., Volume 83, 2002, pp. 195–213
    • Article
       | 
       PDF (1343 K)
       | 
      View Record in Scopus
      Citing articles (2155)
39. • [39]
    • F. Li, Y.X. Miao, S.D. Hennig, M.L. Gnyp, X.P. Chen, L.L. Jia, G. Bareth
    • Evaluating hyperspectral vegetation indices for estimating nitrogen concentration of winter wheat at different growth stages
    • Precis. Agric., Volume 11, 2010, pp. 335–357
    • View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (47)
40. • [40]
    • F. Li, B. Mistele, Y.C. Hu, X.L. Yue, Y.X. Miao, X.P. Chen, Z.L. Cui, Q.F. Meng, U. Schmidthalter
    • Remotely estimating aerial N status of phenologically differing winter wheat cultivars grown in contrasting climatic and geographic zones in China and Germany
    • Field Crops Res., Volume 138, 2012, pp. 21–32
    • Article
       | 
       PDF (1900 K)
       | 
      View Record in Scopus
      Citing articles (17)
41. • [41]
    • J.L. Hatfield, J.H. Prueger
    • Value of using different vegetative indices to quantify agricultural crop characteristics at different growth stages under varying management practices
    • Remote Sens., Volume 2, 2010, pp. 562–578
    • View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (51)
42. • [42]
    • V.N. Vapnik
    • The Nature of Statistical Learning Theory
    • 1995, Springer-Verlag, New York
43. • [43]
    • S.O. Haykin
    • Neural Networks: A Comprehensive Foundation
    • second ed., 1999, Prentice Hall
44. • [44]
    • J. Verrelst, J. Muñoz, L. Alonso, J. Delegido, J.P. Rivera, G. Camps-Valls, J. Moreno
    • Machine learning regression algorithms for biophysical parameter retrieval: opportunities for Sentinel-2 and -3
    • Remote Sens. Environ., Volume 118, 2012, pp. 127–139
    • Article
       | 
       PDF (2241 K)
       | 
      View Record in Scopus
      Citing articles (51)
45. • [45]
    • F. Baret, S. Buis
    • Estimating canopy characteristics from remote sensing observations: review of methods and associated problems
    • Advances in Land Remote Sensing: System, Modeling, Inversion and Application, S. Liang, 2008, Springer, The Netherlands, pp. 173–201
    • View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (51)
46. • [46]
    • R.D. Cutler, T.C. Edwards, K.H. Beard, K.T. Cutler, H.J. Gibson, J.J. Lawler
    • Random forests for classification in ecology
    • Ecology, Volume 88, 2007, pp. 2783–2792
47. • [47]
    • S. Fukuda, E. Yasunaga, M. Nagle, K. Yuge, V. Sardsud, W. Spreer, J. Müller
    • Modelling the relationship between peel colour and the quality of fresh mango fruit using random forests
    • J. Food Eng., Volume 131, 2014, pp. 7–17
    • Article
       | 
       PDF (1153 K)
       | 
      View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (6)
48. • [48]
    • J. Bendig, K. Yu, H. Aasen, A. Bolten, S. Bennertz, J. Broscheit, M.L. Gnyp, G. Bareth
    • Combining UAV-based plant height from crop surface models, visible and near infrared vegetation indices for biomass monitoring in barley
    • Int. J. Appl. Earth Obs. Geoinf., Volume 39, 2015, pp. 79–87
    • Article
       | 
       PDF (2399 K)
       | 
      View Record in Scopus
      Citing articles (6)
49. • [49]
    • K. Prabhakara, W.D. Hively, G.W. McCarty
    • Evaluating the relationship between biomass, percent groundcover and remote sensing indices across six winter cover crop fields in maryland, United States
    • Int. J. Appl. Earth Obs. Geoinf., Volume 39, 2015, pp. 88–102
    • Article
       | 
       PDF (5726 K)
       | 
      View Record in Scopus
      Citing articles (1)
50. • [50]
    • X.L. Jin, K.R. Wang, C.H. Xiao, W.Y. Diao, F.Y. Wang, B. Chen, S.K. Li
    • Comparison of two methods for estimation of leaf total chlorophyll content using remote sensing in wheat
    • Field Crops Res., Volume 135, 2012, pp. 24–29
    • Article
       | 
       PDF (376 K)
       | 
      View Record in Scopus
      Citing articles (8)
51. • [51]
    • P.F.C. Monteiro, R.A. Filho, A.C. Xavier, R.O.C. Monteiro
    • Assessing biophysical variable parameters of bean crops with hyperspectral measurements
    • Sci. Agric., Volume 69, 2012, pp. 87–94
    • View Record in Scopus
      Citing articles (4)
52. • [52]
    • P.M. Hansen, J.K. Schjoerring
    • Reflectance measurement of canopy biomass and nitrogen status in wheat crops using normalized difference vegetation indices and partial least squares regression
    • Remote Sens. Environ., Volume 86, 2003, pp. 542–553
    • Article
       | 
       PDF (681 K)
       | 
      View Record in Scopus
      Citing articles (449)
53. • [53]
    • X.C. Li, X.G. Xu, Y.S. Bao, W.J. Huang, J.H. Luo, Y.Y. Dong, X.Y. Song, J.H. Wang
    • Retrieving LAI of winter wheat based on sensitive vegetation index by the segmentation method
    • Sci. Agric. Sin., Volume 45, 2012, pp. 3486–3496 (in Chinese with English abstract)
    • View Record in Scopus
      Citing articles (1)
54. • [54]
    • E. Boegh, R. Houborg, J. Bienkowski, C.F. Braban, T. Dalgaard, N. van Dijk, U. Dragosits, E. Holmes
    • Remote sensing of LAI, chlorophyll and leaf nitrogen pools of crop and grasslands in five European landscapes
    • Biogeosciences, Volume 10, 2013, pp. 6279–6307
    • View Record in Scopus
       | 
      Full Text via CrossRef
      Citing articles (5)
55. • [55]
    • J.U.H. Eitel, T.S. Magney, L.A. Vierling, T.T. Brown, D.R. Huggins
    • LiDAR based biomass and crop nitrogen estimates for rapid, non-destructive assessment of wheat nitrogen status
    • Field Crops Res., Volume 159, 2014, pp. 21–32
    • Article
       | 
       PDF (2712 K)
       | 
      View Record in Scopus
      Citing articles (12)

• Peer review under responsibility of Crop Science Society of China and Institute of Crop Science, CAAS.

• ⁎ 
  Corresponding author.

Open access funded by Institute of Crop Sciences

For further details log on website :