223: New Decision Support System for Irrigation Efficiency

If irrigation efficiency is a goal of yours, a new predictive model may make scheduling easier in the future. José Manuel Mirás Avalos, Tenured Scientist at Misión Biológica de Galicia in the Spanish Nation Research Council (CSIC) (MBG-CSIC) in Santiago de Compostela (Spain) is working on a Decision Support System (DSS) prototype for irrigation and fertilization of winegrapes. This computer model accounts for multiple variables including weather, soil moisture, evapotranspiration, soil type, plant spacing, bud break, variety, and wine quality goals to help farmers make more informed irrigation decisions throughout the growing season.

Resources:

Vineyard Team Programs:

Get More

Subscribe wherever you listen so you never miss an episode on the latest science and research with the Sustainable Winegrowing Podcast. Since 1994, Vineyard Team has been your resource for workshops and field demonstrations, research, and events dedicated to the stewardship of our natural resources.

Learn more at www.vineyardteam.org.

Transcript

Craig Macmillan 0:00

Our guest today is José Manuel Mirás Avalos. He is tenured scientists at the Misión Biológica de Galicia and the Center for Spanish Research Council. Thanks for being on the podcast.

José Manuel Mirás Avalos 0:10

Thank you very much for inviting me. It's a great pleasure for me.

Craig Macmillan 0:14

We were interested in talking to you because we saw that you've been working on a pretty interesting type of technology with it with a whole group of folks around the idea of decision support systems, particularly around irrigation, fertilization for grapes, possibly even root stock selection, when I read, first of all, for our audience, what exactly is a decision support system?

José Manuel Mirás Avalos 0:34

The idea behind that decision support system is to provide a within one package in this case is a computer platform in which we use different kinds of information coming from real data coming from models that that are implemented within this platform to provide the users the end users with information to make certain practices easier, or more rational. In the vineyard. In this case, we were centered in this particular case in irrigation and fertilization. And there was another it's not exactly a decision support system is more like decisions help decision making for the rootstock which is a independent from the, from the irrigation fertilization system.

Craig Macmillan 1:27

How does the grower use this kind of tool so I'm trying to make decisions about irrigating my vineyard and how did the tool play into it?

José Manuel Mirás Avalos 1:36

At the moment is just a prototype, the computer program or the DSS for being short? The DSS Decision Support System can give some information very easy to obtain such as the geographical coordinates the plant spacings location about the nearest weather station for instance, and that information and the algorithm which is inside this platform in the user will receive an information okay for this conditions over this growing season, you will have to use that much amount of irrigation to obtain given in this case, we use an indicator of grape vine water status the user can modulate within a wide array of values. So, you can decide okay, I want that, on average, my grapevines in this particular danger go between these and these values of of water potential. And then from this decision support system says okay, in that case, you must follow these instructions that is to get that much irrigation for obtaining full genes. Or you can use less irrigation in order to obtain a given quality parameter in that case was soluble sugars in the grapes.

Craig Macmillan 2:59

How's that algorithm developed? Your modeling is a predictive model, basically, you're saying. The vine is going to respond a certain way over time. How is how is that done?

José Manuel Mirás Avalos 3:10

We've capitalized on previous works from other research groups. And I said that that this work is not my work is I work for in collaboration with several research institutions in Spain with people that have a strong expertise, viticulture, in grapevine physiology, we capitalize on that, on that knowledge in order to model soil water balance, and adapted to vineyards in this case by using a proxy of the grapevine architecture in order to model vegetative growth over the growing season. So with that, we modulate the evapotranspiration of the vineyard. And from that we calculate soil moisture, according to weather that and using data from experiments carried out here in Spain, with seven grapevine varieties located in different regions of the country, let's say correlate soil water content with a measure of grapevine water status in this case was a stem water potential, which is a measure which is considered here in Europe. Well, in the States, there are some kinds of schools that refer other other types of measuring grape vine water status but in this case here in Spain, A, we proved in a previous work that stem water potential is a modality of great buying water status indicator, which works best for irrigation purposes.

The work that you're basing this on the research that's been done, was it done specifically for developing this model? Are you able to take work that's been done for other purpose And then put it all together to get the algorithm that you want?

It was basically the second option that that you have mentioned, we have extensive experience working on irrigation in vineyard. So we have several, we participated in several experiments concerning different irrigation protocols within the vine. So we let's say capitalize those data from those experiments. And also perform it a couple of experiences during the course of the project, which led to the development of this decision support system.

Craig Macmillan 5:35

Obviously, there's a lot of variation placed to place region to region. And so how do you account for that?

José Manuel Mirás Avalos 5:41

This is a nice question, because it was the most difficult part.

Craig Macmillan 5:45

Yeah, I would imagine Yeah,

José Manuel Mirás Avalos 5:47

For developing this. For accounting this of course, we have weather records from different stations located in different regions of the country, which are close to the vineyards, we use for validating this this model. But we also took into account soil properties such as texture, organic matter, which also vary a lot from region to region, we added two different equations depending on if the soil is calcareous or nor calcareouse, because the hydraulic properties of the soil would be right in case of high calcium carbonate content. But these are the main the main aspects, but there is a parameter within the model that also arise depending on the on the variety, we can imagine this parameter as the threshold of soil water content, until which the given variety of grapevine begins to show signs of water stress. Unfortunately, we could not make a lot of measurements to obtain a wider range of of values for this parameter. According to the data that we have for five varieties, it was very similar, independently of the of the region in which the vineyard is located. So it's more dependent on the variety itself than on the on the location. These are the three main aspects that allow us for four plus eight capturing the variability within within the regions.

Craig Macmillan 7:20

If I was using such a tool, I would give my location that will tell me a lot about what region I'm in, it'll tell me some things. Then you mentioned we put in like the density, the spacing, because that's going to have to what to do with the total leaf area per hectare? Basically, variety, you'd put in varieties as well.

José Manuel Mirás Avalos 7:38

Yes, but for accounting for a variety, we also asked, they use it too, to provide their date of bud break. Because it's different depending on the region and also on the variety.

Craig Macmillan 7:51

Okay, so they'll tell me about the season, we're talking about, like season long recommendations. So at the beginning of the season, I would say okay, this is what I'm shooting for. And I would actually put in this is the the vine water status that I would like this is the lead potential I'd like to see. And then it'll say, okay, based on the historical weather, and based on all these other factors, we believe that applying this amount of water would really have that result eventually.

José Manuel Mirás Avalos 8:18

That is how it works right now planning to adopt it. But we need help from some company or some people who is expert on computer science, we are planning to develop a tool with the aid of some computer science guys or programmers that allow us to divide the growing season in, let's say, to flowering or flower into venison. Verasion to harvest for instance, on the other hand, using four stages, for instance, ranges of water potentials were tabulated more or less. And the objectives are for the wine producing end user, let's say we'll have the curve that is produced by the model, but in this case, divided into four stages and with the theoretical curve that we should have in order to produce a certain type of wine. So we will be able to say ahead or behind the limit. During this specific stage of the growing season. Suppose you have to apply more water or you should not apply water.

Craig Macmillan 9:35

There's a lot a lot of different ways of trying to achieve this. And that's why this one seems to be kind of a new approach. Even if you're in the development stage. It's still a very intriguing concept and how it might be how it might be applied. And tell me more about how do you actually fit these different pieces together because you got work in a vineyard in one region and work in another you got weather information. You found some way correlating these things with the outcome that you're interested in, which is the water potential.

José Manuel Mirás Avalos 10:04

It was kind of difficult but not not so much as as at a first glance, it would be within a soil water balance you have your result is the soil water content at any specific date. But in order to get to that, you need to know the soil water content, let's say the previous day, but you also have already rainfall, you calculate transpiration of the values, the evaporation from the soil is just putting those pieces together that make the thing work. It's It's not so difficult, it's kind of intriguing too, because all these these parts have their equations in the middle to get to them. But in the end is just fit a soil water balance model with with the data from the different, let's say inputs that you have.

Craig Macmillan 10:54

And essentially you as your career actually, I've looked at some of the other things you've done, you do find good fits, you can take multiple variables, this very complicated world. And when you kind of put it all together, you can start to get a picture, you can actually get some fixes to some other variable

José Manuel Mirás Avalos 11:15

Sometimes is easier than some others. So we tried also to model genes, that was very more, much more difficult. We got two nice results. But there was a lot of variation, depending on on climate and in also on the irrigation management. When we validated this other model for dealing with data from from Spain, if I remember correctly, there were poor regions within Spain. It worked well for some regions, but it didn't work for for some others, we didn't get to the solution to get a unique model for all the regions for for GLD. Also, because we combine a deal with dry matter partitioning within the plant.

Craig Macmillan 12:06

Oh, interesting.

José Manuel Mirás Avalos 12:07

Yes. And we did that in collaboration with with American professional.

Craig Macmillan 12:14

Who's that?

José Manuel Mirás Avalos 12:15

Alan Luxo from the from Cornell University.

Craig Macmillan 12:18

Oh, fantastic. Cool.

José Manuel Mirás Avalos 12:20

Yes, because at that moment in time, my supervisor have made it previously researcher stay in at Cornell University with this professor, he began with the with modeling. In that case, it was apple trees. And we adapted that model to vineyards.

Craig Macmillan 12:40

The cross crop work, it's fascinating because you know, grape vines are a very unusual, kind of unique plant cropping world. But they do have a lot in common with other, you know, Woody perennial crops, other orchard crops. And if we can take the research that's done across multiple areas and use it that's really exciting, increases the efficiency and increases the depth of what we can do, which I really, really like, how you validate this model, where you have people try it, and then you'll come out and you'll take measurements or have them take measurements.

José Manuel Mirás Avalos 13:14

For developing the model, we employ that it's a restricted set of that of data, in order to the few parameters that are inside the construction of the model. But we have met so many experiments within the consortium, that we were working on this decision support systems. So we have finally had a set of more than 100 scenarios to validate the model with data collected from the field. In some cases, we have both soil water content and stem water potential. In some others, we only have a stem water potential. So we tested the model against those data. In many cases it worked well. In a few instances, it didn't work well, because of we detected several particularities within the vineyards we use to test the modeling in a given region. For instance, if is you want me to give any an example a specific example, in a region in western Spain southwest, Southwest or Spain, we have vineyard that the field data said that during the summer, it was not as water stress, as the model is saying that it was a fact is a that was occurring is that the vineyard was very close to a river. So it is it is likely that the water table rose within those periods, or that the vine roots were able to reach that water table and the moel wasn't able to capture that. That feature.

Craig Macmillan 14:54

No, but I would imagine if I'm using a tool like that, and I know my sight like can take that kind of thing into account could say, why would a little bit of experience you can say I always know that this recommendation is a little bit higher than what I actually need. But by using that I can say, well, then I'm going to use this number stead. That's kind of the idea. Now, do you see this technology leading to increases in efficiency, reduced water use or just more efficient water use?

José Manuel Mirás Avalos 15:22

I like to seem that, that this is a step forward to auto efficient water use in vineyards. But maybe at this moment is, too let's say it's a scope is too broad. And we have to work on in order to be more specific, or I don't know the word in English.

Craig Macmillan 15:43

Particular.

José Manuel Mirás Avalos 15:44

The idea is that to be able from this from this, let's say decision to port that we have to build more detailed decision support system that allows allows end users to manage irrigation on a daily or we'll be on a weekly basis, but it's still some work to do.

Craig Macmillan 16:07

Yeah, exactly. I mean, this is this is early days, and this tool isn't isn't out yet. So we wanted to talk about the the concept, which is fascinating, which always reminds me of something I noticed when I was doing the research, you mentioned a consortium, when you look up these topics, you'll hit pages that have many, many different organizations listed at the bottom. And I believe you just moved around between a couple that are part of the same group or consortium is what it looks like, how does this work you've got you've got different agencies, you have different educational institutions, you have different departments have different parts of government that are collaborating, they're working together, they're coordinating what they do, is that how this works.

José Manuel Mirás Avalos 16:44

In his particular case, this project came from a network of collaboration, which was funded by the the Spanish government. And that involved, I don't remember how many but maybe 12 institutions that work on different aspects of viticulture, in order to increase the impact of their research that is really done. Because sometimes, I don't know if this happens in the States. But in Spain, we have the problem that many times we work isolated once from the others, and then our research doesn't reach the level of impact that the funding agencies desire. So in order to, to overcome this weakness, the main funding agency for research here in Spain, asks for creating networks of specific topics, between several research institutes, maybe research institutes, but also universities, and in some cases, private companies, but this later is less frequent.

Craig Macmillan 17:58

Interesting. Yeah, it's interesting there's there's more kind of a multi organizational collaboration here in the United States all the time, we've noticed for a particular topic, and some folks are working on this and some folks are kind of working on this and and coming back with things from different regions or different aspects, but they can all be brought into kind of a coordinated outcome for growers is very, very, very, very practical and very exciting. Is there one thing that you would say to a grower regarding this idea of decision support systems, especially around things like irrigation or fertilizers, or one piece of advice or something that they should be excited about or one reason why they might want to consider using such a tool when it becomes available.

José Manuel Mirás Avalos 18:39

Nowadays, the number of decision support system is increasing. There are many companies which are developing tools or recycling other tools coming from other let's say, organizations must be aware that the decision support system generally, which is those that I know that are available on the market are general not specific for a given crop. In order to obtain the best results is is better to have a specific decision support system. So that's for one for the one part. And the second part is that in the end, these are tools to help making decisions, but one cannot disregard the experience of the grower. Of course, in the end these these kind of decision support systems might must be used as a tool. If you allow me to give a recent experience that I have working with a private company, not in the case of vineyards, it was developing a general platform for aiding in irrigation decisions. The final aim that they have is to automate the process of irrigation this can be a little bit dangerous, because if you if you let a model perform the whole process of ollecting that data, make a decision and then execute that decision in the whole process, they can be accumulation of errors that may give a final response, which is not the desired one.

Craig Macmillan 20:15

What you're getting out and you've touched on and what makes sense is it's a decision support. It's not decision, it's not making a decision for you. It's saying, This is what the model says. And you say, Yeah, okay, I hadn't thought of that, or, okay, that works. Or, okay, let's try that. It's not just executing it. I mean, you know, I can imagine I can imagine a world where you would have a decision system that would take all this into account, and then it would open the irrigation valves automatically. And that may not be where we really want to be headed. The human is always going to be the arbiter, that the human is going to be the decision maker, this is about providing the best information to help make a good decision. That's it. And I think that that's really, really crucial, because I am familiar with another variety of other systems. So we look at all this information. And the readings might say, Okay, this is the direction to go, or this is this is what you should do or that, but the grower will say, That's fine for this variety on this root stock on this soil. Absolutely. That works. For me, that makes sense. But we know for a fact that this variety on this root stock on this soil is not going to work because of the experience with all the details. And I've had some very interesting conversations with folks where I'm looking at the database stuff and saying, Hey, these vines look fine, there's plenty of water, the water potential looks great. There's water in the soil, everything seems to be fine. And the grower says, Well, we're going to irrigate. And I'm like, that seems surprising to me. And they say, Well, under this condition, this variety will collapse out of nowhere, when it hits a certain threshold, and I want to make sure we don't get anywhere near that threshold. So that that information was useful for making decisions in one scenario, they make a slightly different decision in another scenario, and literally those two spots are across the road from each other. A lot of similarities between the two but the grower has that has that experience to say yeah, but under certain conditions, this is what's going to happen. And so again, it's about having the best information to make the best choice, but the human is the one that's going to make the call the human is never gonna go away. And I would be really fascinated once you have once this stuff becomes available. I would love to see some research on how people use it, how people use the technology. Where can people find out more about you?

José Manuel Mirás Avalos 22:37

I have profiling on research gate which is a social network for researchers there you can find my All my publications in a no the top is a working over my ground and also on LinkedIn.

Craig Macmillan 22:52

Fantastic. Yep, I found you very easily and you have a lot on there and a whole variety of other topics that we will have don't have time to get to today, but it's really cool work. So our guest today has been José Manuel Mirás Avalos. He is a scientist at Misión Biológica de Galicia in the Spanish Nation Research Council. Spain with the center Spanish Research Council was the one well, thanks for being here. This is really interesting stuff.

José Manuel Mirás Avalos 23:13

Thank you very much, Craig. It was a pleasure for me to talk to you

Nearly perfect transcription by https://otter.ai