MODERATOR: Well, good morning, everyone, and welcome to the Washington Foreign Press Center’s briefing on Russian damage and destruction to Ukraine’s food storage supplies. My name is Doris Robinson and I am the briefing moderator. As a reminder, this briefing is on the record. We will post the transcript and video of the briefing later today on our website at fpc.state.gov. Please make sure that your Zoom profile has your name and media outlet listed so that the briefers know who is on the call today.
And now I will introduce our briefers. Dr. Kaveh Khoshnood is an associate professor at Yale School of Public Health and faculty director for Humanitarian Research Lab. Nathaniel Raymond is executive director of Yale’s humanitarian lab and lecturer at the Yale School of Public Health. Caitlin Howarth serves as the director of operations for the Conflict Observatory team based at the Yale Humanitarian Research Lab. A link to their full bios are in the briefing announcement along with a link to the report.
The briefers will each make opening remarks and then we will open for your questions. Dr. Khoshnood will start us off.
MR KHOSHNOOD: Doris, thank you so much. I’d like to begin with a brief overview of our Humanitarian Research Lab at the Yale University School of Public Health. We refer to it as HRL. My colleagues and I created the Humanitarian Research Lab with the mission to strengthen scientific, evidence-based humanitarian decision making and response.
We can put the work of our HRL into two large buckets. One is to support UN agencies, ministries of health, humanitarian organizations that care about the health of populations that are affected by armed conflict. Examples of this work include establishing disease surveillance as well as mortality surveillance in conflict zones. For example, we are supporting our colleague in northwest Syria in a conflict zone to establish mortality surveillance to make sure every death counts and we understand the cause of death. We also work on projects dealing with both infectious disease and non-infectious disease in humanitarian settings, and we want to do our best to make sure we take care of those individuals. We also have projects related to mental health of refugees and those who are affected by conflict.
The second bucket of our work is to produce evidence that will make governments accountable for their war crimes. This includes analyzing open source data, including satellite imagery, to document and validate attacks on hospitals and other civilian facilities. The Conflict Observatory project in Ukraine is an example of our work. Our team has done similar work in other conflict-affected countries, including Sudan and South Sudan.
I want to emphasize that we are scientists, and we are independent and impartial. The Ukraine project is led by my colleague, Nathaniel Raymond, and a fantastic team that he put together, and I’m just going to turn it over to Nathaniel to tell you about our project in Ukraine.
MR RAYMOND: Good morning. It’s great to see you all here today. The report we released yesterday – a damage assessment of crop storage facilities in Ukraine – is notable for two reasons. One is its findings, which I will run through in brief, and the second is its methodology, which my colleague, Caitlin Howarth, will go into detail with you in just a moment to discuss.
The headline findings of this report, which was primarily a what we call multi-temporal change detection analysis, which I’ll put that in English: a comparison of archival and recently collected satellite imagery of over 340 grain storage facilities in Ukraine. The change detection showed us that at least 75 grain storage facilities have been damaged or destroyed. And what does that 75 number mean? Well, what it means is that 5.36 percent at least of Ukraine’s total grain storage capacity has been damaged or destroyed. That equals 21 percent of those in our sample size.
It wasn’t just damaged and destroyed facilities that we documented, though. There was a massive amount of facilities documented as being in areas under the control of Russia and its proxy forces. And in total, that means that 15.73 percent of – and I apologize for the puppy in the background – that 15 percent of Ukraine’s total storage capacity has been damaged, destroyed, or fallen under the control of Russia’s forces.
The overall headline here is that to date, we have been looking at a crisis of affordability in terms of grain caused by Russia’s invasion and its impact on Ukraine’s agricultural sector. The impact that we documented through our report on its ability to store grain could lead to a second crisis, which is a crisis of availability. If Ukrainian farmers cannot have enough storage to be able to plant their winter wheat, we could look at food shortages that have global implications.
Caitlin, do you want to walk us through the methodology and then turn it back to me?
MS HOWARTH: Will do. Good morning, everyone – or afternoon or evening depending on where you are. The – there was a combination of tools that were utilized for this report. This was one of the more advanced ones that we’ve generated so far through the fortunate partnership with Department of Energy’s Oak Ridge National Laboratory.
In order to conduct this assessment, there were a variety of factors that we had to take into account. One was that, obviously, we were utilizing the very high-resolution satellite imagery. This is resolution that’s usually greater than 50 – greater than or equal to about 50 centimeters per inch. That means we’re able to see a higher level of detail in terms of assessing levels of damage and the severity of that damage. It is important to note that there is a very wide variety of sites that we had to assess for this. Our analysts were looking at not just one particular type of grain silo, but looking at silos, they were looking at different types of crop storage facilities, and all of these have a variety of shapes, different metals, different materials that are used. And all of that creates a very complex picture for effective damage analysis.
In order to do all of that, we collected a – we basically set an AOI that was over – included total at least 700 facilities in the regions most affected by bombardment. And this is since 24 February, and inclusive of Ukraine’s full-scale invasion. There are other areas that were less affected or only more recently affected and that had lower agricultural outputs, and they – those may not be as reflected in this report.
In terms of being able to identify all of those, we used a variety of tools. This is a combination of open source imagery and, most importantly, looking at the Elevatorist. This is a resource that was originally put together by Ukraine and then taken offline in 2019 due to security concerns about the targeting of crop facilities across the country. That data was effectively captured in 2019 by the U.S. Department of Agriculture, and it was with the USDA that Yale HRL was able to pull a much more sensitive version of that data, provide this assessment, and then start to build on top of that using additional open source means.
So we run a combination of searches. We are looking for anything that may reference a facility and also anything that may reference specific damage to a facility, and that’s over a range of media, everything from user-generated social media to official reports to doing deep dives in agricultural media outlets and finding other scientific studies that have been done to date.
The NASA harvest data was also very important as a foundational layer for this report, and that covers everything inclusive within Ukraine, not only the eastern regions but also Crimea. It’s important to keep in mind that of course the Government of Ukraine has not had access to many of these regions, including Crimea, since 2014. So some data has been degraded due to that gap.
We then, working with Oak Ridge National Lab, built a machine learning model that was able to identify additional (inaudible) that were not included in that Elevatorist data or other parts of our research. And that’s a really key thing to understand here. We never take any single data source as comprehensive or simply take it on face value. We – our team assesses each data layer for its validity, for its accuracy, and they’re essentially building pieces and vetting each of those data sources as they go and checking them against additional sources. So the Oak Ridge machine learning model was a really crucial piece of being able to double-check all of the – sort all of the – all of the locations that we had originally identified and to see if there was anything missing, which in this case, there was.
It was also really important to be able to contrast field-level data. This is from the ground where people are able to assess layers of damage, and that helps to inform some of the remote sensing analysis to get a better idea of exactly what we’re seeing from the sky. That can be anything from burns, shrapnel damage, anything that might cause a change in the layer of detail that we see as we’re doing that multitemporal change detection.
One of the things that we would point you to in the various examples provided is that you’ll see a wide range in the types of buildings, as I previously mentioned. That’s also important to note here that we made sure that this report was fundamentally leaning on what we could visually identify and robustly confirm to high confidence. So it’s important to keep in mind that if we were not able to visually confirm detail, it was not included in this report and that makes this report’s findings more conservative than the likely extent of damage that was actually been (inaudible) to date.
Of all of those locations that were identified, there were also some really key limitations. About 50 percent of the locations that we wanted to be able to assess, we weren’t able to due to the fact that there wasn’t sufficient satellite imagery data in the more recent period. And that’s something that we struggle with sometimes in more remote regions, but it is something that we always work to try and come back to. So you will probably be seeing additional reporting coming out as new imagery is made available to us and that which we are able to contrast sufficiently to archival material.
And with that, I’m going to hand it back to Nathaniel Raymond, who is going to make sure that we understand the full impact of this report and its findings.
MR RAYMOND: So unlike previous reports, where we have looked at damage to hospitals and schools across five oblasts in the case of hospitals – and schools and hospitals in the case of Sievierodonetsk – that building stock is significantly different than what we’re looking at in this report. And I want to stress that, in terms of silos and grain elevators, even a small amount of damage, which may not take a school or a hospital out of commission, can be catastrophic for a crop storage facility.
As those of you who have covered grain in the past know, it has to be kept at a specific temperature and percentage of moisture. And so even smaller degrees of damage in these facilities can have significant effect, including knocking out the air dryers for the grain storage facilities – and because of the thing aluminum walls of silos, even a small amount of shrapnel can perforate the exterior wall and expose the grain to the elements. And so you’ll see that in the report, we have a damage scale which is damaged/not damaged. And that’s very important here because we’re dealing with a type of facility that is significantly fragile.
What we also discovered about these facilities is that 80 percent of them that were damaged appear to be at ports or critical railway spurs. And so that finding – 80 percent was near transportation infrastructure – was also important. Three oblasts – Mykolaiv, Donetsk, and Zaporizhzhia – contained the majority of the damaged facilities that we were able to see. And we were dealing with what we call a confounder factor, which is that there was a large amount of pre-existing dilapidation that we had to make sure did not skew our results. And so that’s why we had a very high standard in terms of visible conflict-related damage being required combined with validation, when possible, across two sets of imagery.
The closing line here is that Ukraine provides the World Food Program 40 percent of its emergency food supply that it uses to feed the world’s most food insecure people. The amount of damaged and destroyed grain that could be affected by the facilities that we’ve imaged in this report is equal to the total reported amount of grain that has left Ukraine under the UN-negotiated grain deal in the past few weeks, just to give you a sense of scale.
In closing, the legal headline here is that whether intentional or indiscriminate, the targeting of crop production and food storage infrastructure underneath the Amended Protocol of the Fourth Geneva Convention and under the 1998 Rome Statute can constitute a war crime and a crime against humanity in some cases. The evidence we have here is consistent with that alleged crime. And with that, I turn it back to Doris.
MODERATOR: Great, thank you so much for those remarks. We will now start the question-and-answer portion of the briefing. For journalists who would like to ask a question, please hit the raised hand icon, and we have questions already. We’ll take our first question from Dmitry Anopchenko, Inter TV News, Ukraine.
QUESTION: Good morning to you. Good morning, Doris. Very nice to see you. Thank you very much for (inaudible). I’d like to ask a question about these grain storage facilities damaged or destroyed. We were told on a different levels, even in State Department, that Russia is using the food as an arm, as it’s using the gas arm. Do you – may one suggest that this damage or – which were – which was given to storage facilities was made intentionally, that it’s the part of the wider plan? If yes, how to find the evidences? Because at the top you talked about the responsibility of the foreign governments of the war crimes, so if we need to make someone responsible, which evidence could be find during this process? Thank you.
MR RAYMOND: Dmitry, you’re asking the critical question. So I’m going to start by telling you what we’re doing now that’s not in the report. Right now, we are going through each of those 75 facilities in great detail to do what’s called typing, where we are looking at the damage at each of these facilities and trying to ascertain what potential munition from what weapons platform was used. We’re also going to do a second process to look at surrounding damage. When you’re trying to differentiate between indiscriminate and intentional bombardment, a critical indicator that can help you in that analysis is whether there is a large amount of damage in a common radius around that target.
Now, as you’ll see in some of the imagery included in this report, we have examples of facilities that were struck that could be seen as a – basically affected by indiscriminate fire. I’m thinking of one example where there’s artillery-consistent craters all around the facility. And then there are others that look like more precision strikes. Probably what we are going to find is it is a combination of indiscriminate and intentional. That said, we have to do that analysis on every single facility, and that takes a large amount of time. So that’s underway right now.
The second point here is the principle in the law of armed conflict of precaution. Regardless of whether any of these strikes were intentional, these were civilian objects. They begin in a legal state as protected civilian infrastructure. That means that Russia, that’s a party to the Geneva Convention and to the Hague Treaty, has an obligation to engage in what’s called precaution, meaning that they have to take steps to prevent hitting that infrastructure. We can say now, based on the data in front of us, that those steps of precaution were not taken.
If hypothetically there was no intentional bombardment in this data set that we have – which I’m not saying that but I’m saying let’s take that hypothetically – the fact of the matter is that the crime is the same whether or not it was intentional or indiscriminate. And for our assessment now, we can clearly say there was large-scale indiscriminate bombardment and there is indications that we are now verifying of incidents of intentionality that require further analysis. But we’re doing that analysis probably right now as we speak.
MS HOWARTH: I would just add one quick point, which is that of the various sites that we were able to get to this high-level confirmation on damage, 60 out of the 75 were transportation hubs. And that we found to be quite significant, given the importance of storage facilities at the transportation points in order to facilitate the actual delivery of these crops to their ultimate destinations. So that would be something I would keep in mind as you produce your copy.
MR RAYMOND: And we are going to – to Caitlin’s point, we are going to run an analysis of the other facilities that were not hit to see what percentage of those that were not hit were not near transportation hubs.
MODERATOR: Thank you. And we’ll take our next question from Yaroslavl Dovgopol with National News Agency of Ukraine.
QUESTION: Thank you so much. Can you hear me?
MR RAYMOND: Okay.
QUESTION: Thank you. First of all, I want to thank for all this. Of course, it’s essential for a lot of nations, not just for Ukraine, and for the whole international community. So my question is about some of your next steps. You said about gathering all the evidences, but are you considering also to make some proposals for the U.S. Government and international community on how to mitigate the harm to Ukrainian grain facilities effectively? Thanks.
MR RAYMOND: Well, that is a great question, and it’s important for us to stay in our lane of expertise. We’re diagnosticians as it relates to the impact of conflict on civilian populations. We are not agronomists nor crop storage specialists, but we have learned in the past eight weeks more than we ever thought we would know about silos in Ukraine. It’s been a education.
That said, we think – and we know – that the information in this report is supporting – inside the U.S. Government and outside the U.S. Government – those who can make those recommendations. So this information will be useful for World Food Program, Food and Agriculture Organization. It supports what’s happening with NASA Harvest in USDA’s Foreign Agriculture Service. And it also supports efforts by elements within the U.S. Government and around the world that are looking at sanctions and accountability actions right now, both shortterm and long-term.
And so we are not going to make recommendations based on this report about how to mitigate the effects, but we know those who can make those recommendations, based on their specific expertise, can utilize this data and this data is in their hands. So stand by, and we are continuing to support the people who can take those policy steps and make those recommendations to give them the best information we can.
MODERATOR: Okay. And Dr. Khoshnood, did you have any remarks?
MR KHOSHNOOD: No. I think Nat has summarized it well.
MODERATOR: Great. We’ll take our next question from Alex Raufoglu from Turan News Agency, Azerbaijan.
MR RAYMOND: Good to see you again, Alex. How you doing?
QUESTION: Thank you so very much for doing. This is Alex Raufoglu from Azerbaijan, Turan News Agency. My apologies for technical issues; I’m joining from an airport. My question is about, first of all, the very wide implications of these actions, particularly geographically. We just heard from Putin, again putting – pushing his own propaganda today, trying to divide the world, basically, blaming the West by saying that they are trying to exempt sanctions; basically, they only are trying to basically prioritize developing world versus world’s poorest countries. Could please explain how much this impacts – its implications, let’s say, touch the, let’s say, Global South and other parts of the world? We heard World Food Program today warn that the war in itself has pushed 70 million people closer to starvation.
Secondly, in terms of accountability, the report notes that any actual destruction of such facilities may constitute war crime and violation of the 1977 Additional Protocol I to the Geneva Convention. I also heard, Mr. Raymond, you mentioned 1986 Protocol, if I’m not mistaken. For those of us who don’t study this for making a living, could you please explain to us what’s the difference and what does it mean in terms of accountability and in terms of going after the perpetrators? The U.S. called for further investigation yesterday through appropriate mechanisms. What is the next step in this case? Thank you so much. Again, great to see you all.
MR RAYMOND: Great to see you. The next step here is that basically we’re going to do more analysis and be supporting our colleagues at International Criminal Court and domestic accountability leaders in Ukraine in the Office of the Prosecutor General to use this information to get to that next level of validation on each of these facilities, and we’re going to try to expand the number of facilities in that count from over 340 imaged to get it as close to our magic number, which is 800 approximately in the area of interest that we believe were most conflict-affected.
As it relates to your question on international humanitarian law, the fact of the matter here – let’s separate it out. So on the Geneva side, the Amended Protocols are clear, I believe in both Article 52 and Article 54, as it relates to the targeting of food production. And it also has language at the end basically saying in any other – and I’m not being precise here with my language, but basically in any other steps to disrupt the food supply is the way that language is read. And so in this case, whether through indiscriminate or intentional action, these steps – even if it doesn’t affect production itself but the ability to bring production to market – is a violation of those elements in the Amended Protocol.
As it relates to the Rome Statute of 1998 in terms of crimes against humanity, there is an element there of intentionality. And so back to Dmitry’s earlier question, to prove a crime against humanity in this case will require a determination of intention and command direction, unlike on the case of a war crime where it can be individual commanders in a situation that have failed in their duty of precaution. In the case of the Rome Statute, it requires a finding of intentionality, and to get to that finding of intentionality requires a level of investigation beyond what’s in our report. So it would require communications, it would require witness testimony from the chain of command, and would require a showing of an intent to disrupt the food supply.
MODERATOR: Great, thank you. We are just about out of time and I don’t see any other hands raised, so I think I’ll throw it back to our briefers for any closing remarks.
MR RAYMOND: A headline closing remark for me, and I’ll turn it over to my colleagues after that, just to talk about what happened in this assessment technically, so stepping back from the evidence of alleged crimes. In approximately eight weeks, the Yale team with our colleagues at Oak Ridge National Lab pulled off a – an engineering feat combined with an analysis achievement that represents arguably the largest-scale remote assessment of conflict damage to a country’s grain storage infrastructure that has ever been done. And it’s been done to a level of accuracy, a level of granularity – no pun intended – that deserves note.
We simply did not just look at satellite imagery. We created an underlying data set on top of the Elevatorist asset. We were able to – with Oak Ridge’s help – to validate and expand on that underlying map, and then we were able to confirm that against recently collected imagery at 340-plus locations, which is a scope of satellite imagery collection in Europe’s second-largest country which represents 25 percent approximately of Ukraine’s total grain storage capacity. And that was done basically from the beginning of July until a few days ago. So I just want to honor the teams, Yale and Oak Ridge, that was given an incredible challenge and they achieved it.
Over to my colleagues.
MR KHOSHNOOD: I just want to mention that, as public health professionals, we’re always try to prevent bad things from happening in the first place. We wish we could have prevented these kinds of conflicts to occur, but we don’t really have the power and the tools to do that. But we still want to step up and make sure we can address all the negative health consequence of armed conflict. I hope that my colleagues as well continue to do this work. Thank you.
MS HOWARTH: And final point for me. It’s also important to keep in mind that even producing this kind of analysis can create weaponizable data for the armed actors that are party to the conflict. And we’ve taken significant pains in the way that we’ve represented this data, including specific obfuscation of the locations of these many sites, to ensure that food production facilities cannot be effectively targeted using our report. That’s something that is critical to everything that the Humanitarian Research Lab produces, and something that we strongly urge among all of our colleagues to ensure that our material can’t be used to create disinformation, misinformation, or even used as a targeting asset.
MODERATOR: Thank you so much. I would like to take this time to thank our briefers, Dr. Khoshnood, Nathaniel Raymond, and Caitlin Howarth, for taking the time to brief with us today, and all of the journalists who participated as well. And that concludes today’s briefing. We will send a transcript later today at fpc.state.gov. Thank you all.